Morphology is visible Marc van Oostendorp Leiden University & - - PowerPoint PPT Presentation

▶
morphology is visible
SMART_READER_LITE
LIVE PREVIEW

Morphology is visible Marc van Oostendorp Leiden University & - - PowerPoint PPT Presentation

Dec 05, 2023 419 likes •1.06k views

Two theories of faithfulness Ineffability Relativized MP ARSE Allomorphy Morphology is visible Marc van Oostendorp Leiden University & Meertens Instituut Network on Morphological Exponence Two theories of faithfulness Ineffability

slide-1
SLIDE 1

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Morphology is visible

Marc van Oostendorp

Leiden University & Meertens Instituut

Network on Morphological Exponence

slide-2
SLIDE 2

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Morphology is visible

  • 1. I claim that one of the ‘functions’ of phonology is to make

morphology visible

  • 2. Many phonological anomalies can be understood from this

function

  • 3. I present an OT model in which underlying structures are

morphosyntactic feature bundles

  • 4. it is the function of Gen to interpret these bundles, among
  • ther things by lexical insertion
  • 5. This explains ineffability and allomorphy
slide-3
SLIDE 3

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Morphology is visible

Two theories of faithfulness Two theories of faithfulness Consistency of Exponence Ineffability Examples and possible analyses Ineffability in classical Containment Relativized MPARSE Background Not parsing the morphology for phonological reasons Allomorphy The nature of inputs A case study: Dyirbal Alternative analysis

slide-4
SLIDE 4

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Containment and Correspondence

  • 1. Correspondence Theory: There are separate input and
  • utput representations, as well as correspondence

constraints between elements of these (McCarthy and Prince 1995)

  • 2. Containment Theory: The input is contained in the output,

therefore all faithfulness constraints can be read off the surface representation (Prince and Smolensky 1993, Van Oostendorp 2005).

slide-5
SLIDE 5

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Correspondence

k k l u u k k u input

  • utput
slide-6
SLIDE 6

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Containment

  • Containment. Every element of the phonological input

representation is contained in the output. (There is no deletion.)

slide-7
SLIDE 7

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Containment: Prince and Smolensky 1993

  • PARSE: All elements should be ‘parsed’ in the phonological

structure (no deletion.)

  • FILL: Do not allow empty elements. (No insertion.)
slide-8
SLIDE 8

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Containment Representation

Φ k l u k ∅

slide-9
SLIDE 9

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Occam’s Razor and Containment

  • PARSE-C: Every consonant needs to be affiliated to

prosodic structure

  • FILL-V: (Nucelar) syllable slots need features.
slide-10
SLIDE 10

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Problems with the Prince & Smolensky Interpretation

  • features should also not be allowed to ever spread to an

epenthetic vowel

  • how do we prevent spreading from happening everywhere

in every language?

slide-11
SLIDE 11

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Morphology is visible

Two theories of faithfulness Two theories of faithfulness Consistency of Exponence Ineffability Examples and possible analyses Ineffability in classical Containment Relativized MPARSE Background Not parsing the morphology for phonological reasons Allomorphy The nature of inputs A case study: Dyirbal Alternative analysis

slide-12
SLIDE 12

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Consistency of Exponence

  • “No changes in the exponence of a

phonologically-specified morpheme are permitted.” (McCarthy and Prince 1993, 1994)

slide-13
SLIDE 13

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Consistency of Exponence

“[Consistency of Exponence] means that the lexical specifications of a morpheme (segments, prosody, or whatever) can never be affected by Gen. In particular, epenthetic elements posited by Gen will have no morphological affiliation, even when they lie within or between strings with morphemic

  • identity. Similarly, underparsing of segments — failure to endow

them with syllable structure — will not change the make-up of a morpheme, though it will surely change how that morpheme is realized phonetically. Thus, any given morpheme’s phonological exponents must be identical in underlying and surface form.”

slide-14
SLIDE 14

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

CoE Representation

Φ k l u k u M

slide-15
SLIDE 15

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Consistency of Exponence (Coloured version)

  • Gen does not affect morphological colours.
slide-16
SLIDE 16

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Faithfulness constraints (coloured versions)

  • PARSE-φ(x): The morphological element x must be

incorporated into the phonological structure. (No deletion.)

  • PARSE-µ(x): The phonological element x must be

incorporated into the morphological structure. (No insertion.)

slide-17
SLIDE 17

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Morphology is visible

Two theories of faithfulness Two theories of faithfulness Consistency of Exponence Ineffability Examples and possible analyses Ineffability in classical Containment Relativized MPARSE Background Not parsing the morphology for phonological reasons Allomorphy The nature of inputs A case study: Dyirbal Alternative analysis

slide-18
SLIDE 18

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Example: Dutch diminutives

base form diminutive form gloss man man-@tj@ man maan maan-tj@ moon raam raam-pj@ window dak dak-j@ roof

slide-19
SLIDE 19

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Example: exceptions to diminutive formation

base form diminutive form gloss lente

??lente-tj@

spring

∗lent-j@

schade

??schade-tj@

damage

∗schaad-j@

boete

??boete-tj@

fee

??boet-j@

Hilde

??Hilde-tj@

(name)

?Hilde-k@

slide-20
SLIDE 20

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Lexicalisation

  • These words tend to get better, when they are repeated;

for some speakers a name such as Hildetje has become perfectly acceptable.

  • We thus have a form of a derived environment effect
  • We find this more often in cases of ineffability: derived

forms of shape X cannot be generated, even if X exists underlyingly

slide-21
SLIDE 21

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Dealing with ineffability within OT (1)

  • 1. The ‘paradigmatic solution’: the Generator function does

not generate an individual form, but a paradigm. Ineffability

  • f an individual form means that this particular form is not

generated within the paradigm (Rice 2005, 2006).

  • 2. The ‘null parse’ solution: the Generator function generates

a candidate in the phonology which does not have a phonetic interpretation, and this is selected as the winner in certain cases (Prince and Smolensky 1993, McCarthy and Wolfe 2006).

slide-22
SLIDE 22

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Dealing with ineffability within OT (2)

  • 3. The ‘control’ solution: the Generator and Evaluator

function conspire to create a (pronounceable) candidate, but a grammatical component outside of the standard OT system then blocks this candidate (Orgun & Sprouse 1999)

  • 4. The ‘divergent meaning’ solution: we generate a

phonologically well-formed form, but one which does not have the intended semantics; the form is therefore

  • unusable. This solution is basically the one proposed for

syntax, and will be defended here for phonology.

slide-23
SLIDE 23

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Ineffability and faithfulness

  • In all of these solutions, the account is in the relation

between input and output structures of forms, i.e. in the theory of faithfulness.

slide-24
SLIDE 24

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Morphology is visible

Two theories of faithfulness Two theories of faithfulness Consistency of Exponence Ineffability Examples and possible analyses Ineffability in classical Containment Relativized MPARSE Background Not parsing the morphology for phonological reasons Allomorphy The nature of inputs A case study: Dyirbal Alternative analysis

slide-25
SLIDE 25

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

No Structure May be Optimal

/rˇ e/ FTBIN LX≈PR . . . PARSE

  • a. rˇ

e * . . . *

  • b. [ (rˇ

e)F ]PrWd *! . . .

slide-26
SLIDE 26

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

But Can We Ever Derive More Complicated Cases

  • Why would it be more optimal to derive ∅ from an input

structure { lent@, tj@ }, rather than, say, [lent@j@]?

slide-27
SLIDE 27

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Another attempt: MPARSE

“On this view, then, the underlying form of an item will consist of a very incompletely structured set of specifications which constrain but do not themselves fully determine even the morphological character of the output form. These specifications must be put in relation, parsed into structure, in

  • rder to be interpretable. ”

“Failure to achieve morphological parsing is fatal. An unparsed item has no morphological category, and cannot be interpreted, either semantically or in terms of higher morphological structure.”

slide-28
SLIDE 28

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Problems with MPARSE

  • If higher-order systems are also OT grammars, where is

the crash?

  • How can we maintain Richness of the Base?
slide-29
SLIDE 29

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Morphology is visible

Two theories of faithfulness Two theories of faithfulness Consistency of Exponence Ineffability Examples and possible analyses Ineffability in classical Containment Relativized MPARSE Background Not parsing the morphology for phonological reasons Allomorphy The nature of inputs A case study: Dyirbal Alternative analysis

slide-30
SLIDE 30

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Assumptions about morphology

  • Items-and-Arrangement.
  • The input can be either an unstructured set of morphemes,
  • r a complex word consisting of morphemes arranged into

some structure

  • The optimal output consists of a morphological word (just

like the optimal output consists of a phonological word), because

  • There are (M)PARSE constraints which require that

individual morphemes should be part of the morphological structure.

slide-31
SLIDE 31

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

MPARSE

  • MPARSE(M): Every morpheme M has to be parsed into a

morphological word.

slide-32
SLIDE 32

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Independent justification for MPARSE

  • LEXDIM: There is no diminutive of function words.
  • { aan ‘to’, DIM }

LEXDIM MPARSE(DIM)

  • a. aantje <+DIM>

*! ☞ b. aan *

slide-33
SLIDE 33

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Where is the phonology?

µ P

  • ❅

❅

DIM

  • a :

n tj @

❅ ❅

  • σ

φ

slide-34
SLIDE 34

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Where is the phonology?

µ P

  • ❅

❅

DIM a : n

❅ ❅

  • σ

φ

slide-35
SLIDE 35

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Theoretical implications

  • We have to assume that the input is a bunch of

morphological features

  • Phonological insertion works in parallel with the

phonological evaluation

  • This changes the implementation of Richness of the Base,

but not (necessarily) its spirit:

  • Any bunch of morphemes can be underlying
  • There can be no ‘inherent’ restrictions on the structure of a

morpheme.

Wolfe (2007) arrives at similar conclusions following a completely different line of thought.

slide-36
SLIDE 36

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Not filling in the segments

µ P

  • ❅

❅

DIM

  • a :

n tj @

❅ ❅

  • σ

❅ ❅

σ

✏ ✏ ✏ ✏ ✏

φ Technically, tj@ here behaves as a string of gratuitous epenthetic segments, violating PARSE-µ

slide-37
SLIDE 37

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Filling in the segments

µ

❍❍ ❍

N

✟ ✟ ✟

  • ❅

❅

DIM m a : n

❍ ❍ ❍ ❅ ❅

  • σ

φ REALIZEMORPHEME: morphological nodes should have an equivalent in the phonological representation

slide-38
SLIDE 38

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Morphology is visible

Two theories of faithfulness Two theories of faithfulness Consistency of Exponence Ineffability Examples and possible analyses Ineffability in classical Containment Relativized MPARSE Background Not parsing the morphology for phonological reasons Allomorphy The nature of inputs A case study: Dyirbal Alternative analysis

slide-39
SLIDE 39

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Not parsing the morphology for phonological reasons

{ /lEnt@/, DIM } PARSE-φ OCP(cor) MPARSE(DIM) ☞lEnt@ (DIM) * lEnt@tj@ *! lEnt<@t>j@ *! This form is now not a diminutive and will not be treated as such by any outside module.

slide-40
SLIDE 40

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Lexicalisation

/hIld@tj@/ } MPARSE(LEX) PARSE-φ OCP(cor) MPARSE(DIM) ☞hIld@tj@ * hIld@<tj@> *!** hIld<@t>j@ *!* This form is now not a diminutive and will not be treated as such by any outside module.

slide-41
SLIDE 41

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Swedish example

  • en r¨

add (MASC) pojke ‘a scared boy’

  • *et r¨

add-t (NEUTER) barn ‘a scared child’

slide-42
SLIDE 42

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Swedish example

/r¨ ad:/+NEUTER OCP(cor) PARSE(C) MPARSE (NEUTER) ☞r¨ ad: <NEUTER> * r¨ ad:t *! r¨ ad:<t> *!

slide-43
SLIDE 43

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Morphology is visible

Two theories of faithfulness Two theories of faithfulness Consistency of Exponence Ineffability Examples and possible analyses Ineffability in classical Containment Relativized MPARSE Background Not parsing the morphology for phonological reasons Allomorphy The nature of inputs A case study: Dyirbal Alternative analysis

slide-44
SLIDE 44

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

The nature of inputs

Under the model presented here:

  • inputs are abstract morphosyntactic (and semantic)

features

  • Gen selects items from the lexicon and puts them in a

phonological representation

  • The only ‘faithfulness’ strictu senso takes place in the

morphology

slide-45
SLIDE 45

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Allomorphy

  • this solves the problem of how to deal with allomorphy and

faithfulness.

  • e.g. Dutch plurals (simplified):
  • /@n/ combines with final stress
  • /s/ combines with prefinal stress
  • The input thus is necessarily something like { @n, s }
  • Some stems have stem allomorphy (Booij 1998):

professoren [profEs´

  • :r@n] / prof´

essors [profEsOrs] (*[profes´

  • :rs], *[prof´

Esor@n])

  • How do we formulate faithfulness? (In particular within

Containment?)

slide-46
SLIDE 46

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Morphology is visible

Two theories of faithfulness Two theories of faithfulness Consistency of Exponence Ineffability Examples and possible analyses Ineffability in classical Containment Relativized MPARSE Background Not parsing the morphology for phonological reasons Allomorphy The nature of inputs A case study: Dyirbal Alternative analysis

slide-47
SLIDE 47

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Dyirbal

  • Wolfe (2007), working in a framework similar to the present
  • ne (but with Correspondence), analyses the famous case
  • f Dyirbal
  • yaúa-Ngu/*yaúa-gu ‘man’
  • *yamani-Ngu/yamani-gu ‘rainbow’
  • *balagara-Ngu/balagara-gu ‘they’
slide-48
SLIDE 48

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Wolfe (2007) on Dyirbal (1)

  • These suffixes spell out the following case features (Halle

and Vaux 1998): [-oblique, +structural, +superior, -free]; there is an inclusion relation

  • /-Ngu/ spells out [-oblique, +structural, +superior, -free]
  • /-gu/ spells out [-oblique, +structural, +superior]
slide-49
SLIDE 49

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Wolfe (2007) on Dyirbal (2)

  • Because of this inclusion relation, and

REALIZEMORPHEME([-free]), there will be a preference for /-Ngu/, even if this is phonologically marked ‘man’, [-obl,+str,+sup,-fr] RM-[-fr] *Nasal ☞yaúa-Ngu * yaúa-gu *!

slide-50
SLIDE 50

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Wolfe (2007) on Dyirbal (3)

  • However, since [Ngu] needs to be aligned to a foot,

sometimes we might choose the non-preferred option

  • ALIGN-[Ngu]: The left edge of [Nku] coincides with the right

edge of the head foot ‘rainbow’, [-obl,+str,+sup,-fr] ALIGN-[Ngu] RM-[-fr] *Nasal yamani-Ngu *! * ☞yamani-gu *

slide-51
SLIDE 51

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Problems

  • The stipulation that [Ngu] lacks a morphological feature

[-free] is arbitrary

  • Richness of the Base: Why couldn’t we have an input

‘man’+ [-obl,+str,+sup] (without [-fr]), so that we would get yaúa-gu.

  • We stipulate that [Ngu] lacks [-fr] and that it is subject to a

specific constraint on alignment, without linking those specifications

slide-52
SLIDE 52

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Morphology is visible

Two theories of faithfulness Two theories of faithfulness Consistency of Exponence Ineffability Examples and possible analyses Ineffability in classical Containment Relativized MPARSE Background Not parsing the morphology for phonological reasons Allomorphy The nature of inputs A case study: Dyirbal Alternative analysis

slide-53
SLIDE 53

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Alternative analysis (1)

  • The only relevant property of [Ngu] is its phonological

property: it can go very well with a bisyllabic unit

  • The reason for this is that it is prespecified with its own

prosodic word structure, allowing the preceding word to be independent

  • However, the stem only wants to be an independent unit if

it is prosodically perfect, i.e. a bisyllabic unit

slide-54
SLIDE 54

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Alternative analysis (2)

  • Maybe the fact that [Ngu] is special, can be derived from its

phonological shape.

  • For now we assume however that it comes with its own

prosodic word prespecified, and [gu] does not.

  • Constraints:
  • WORDBIN: Words are bisyllabic units
  • FAITHFUL-P: Don’t add prosodic structure
slide-55
SLIDE 55

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Tableau: (yaúa)(Ngu)

‘man’, [-obl,+str,+sup,-fr] WORDBIN FAITHFUL-P *Nasal ☞(yaúa)(Ngu) * (yaúa-gu) *! (yaúa-Ngu) *! * (yaúa)(gu) *!

slide-56
SLIDE 56

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Tableau: (yamani-gu)

‘man’, [-obl,+str,+sup,-fr] WORDBIN FAITHFUL-P *Nasal (yamani)(Ngu) * *! ☞(yamani-gu) * (yamaniNgu) * *! (yamani)(gu) * *!

slide-57
SLIDE 57

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Conclusions

  • A Containment view of the relation between lexical

structure and phonological output allows a (relatively) restricted view of input-output relations

slide-58
SLIDE 58

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Conclusions

  • A Containment view of the relation between lexical

structure and phonological output allows a (relatively) restricted view of input-output relations

  • In many cases, this means that the wealth of theoretical

possibilities to solve a particular problem is severely reduced

slide-59
SLIDE 59

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Conclusions

  • A Containment view of the relation between lexical

structure and phonological output allows a (relatively) restricted view of input-output relations

  • In many cases, this means that the wealth of theoretical

possibilities to solve a particular problem is severely reduced

  • And I consider that to be a good thing
slide-60
SLIDE 60

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Conclusions

  • A Containment view of the relation between lexical

structure and phonological output allows a (relatively) restricted view of input-output relations

  • In many cases, this means that the wealth of theoretical

possibilities to solve a particular problem is severely reduced

  • And I consider that to be a good thing
  • In particular, it leads us to conclude that ineffability means

that Gen inserts morphological exponents

slide-61
SLIDE 61

Two theories of faithfulness Ineffability Relativized MPARSE Allomorphy

Conclusions

  • A Containment view of the relation between lexical

structure and phonological output allows a (relatively) restricted view of input-output relations

  • In many cases, this means that the wealth of theoretical

possibilities to solve a particular problem is severely reduced

  • And I consider that to be a good thing
  • In particular, it leads us to conclude that ineffability means

that Gen inserts morphological exponents

  • and it gives a restrictive view of lexical insertion in Dyirbal