Optimal Theory

Question 1

What is OT?

Answer 1

OT is:
a generative approach to phonology (like rule-based)
constraint-based (no more rules, no more phonological processes)
an approach that is based on computational modeling

Question 2

Rule-based vs. OT (i.e. constraint based)

Answer 2

Rule Based
Generative theory
Posits a UR and SR(s)
Limited to describing a process in a specific language
ex. Vowel epenthesis in Basque
ex. Nasalization in English

Optimality theory
Generative theory
Start with a UR and SR(s)
Considers all processes to be universal and seeks to explain language-specific differences

Question 3

Why OT?

Answer 3

Many problems with rule-based phonology
Rules don’t make predictions
Rules ‘fail’ us in a few ways (Zsiga gives many examples in 14.1)
I will focus only on 1 which is static generalizations
An attempt to understand cross-linguistic patterns
Apply a set of universal constraints to all words in all languages (with a computer, of course)
Depending on the SRs (and the changes from the UR to the SR) we can see how languages prioritize or ORDER their constraints- some are important, some are not

Question 4

Static generalizations vs. morpheme structure constraints (MSC)

Answer 4

Morpheme Structure Constraints (p. 306)
What we write rules for.
Process of nasals matching place of articulation of following stops.
In English, morpheme ‘-in’ (not). When we add it to a word, it changes.
Impossible (m)
Incongruous (ŋ)
Indelible (n)
Static generalizations (p.223, 306)
In English we have words like: ‘camp’ and ‘think’ and ‘wind’ where nasals match in place of articulation with the following stop
But, we can’t write ‘rules’ for these. There is no evidence of an underlying form (UR)… there are single morphemes and there is no change. They just are.

Question 5

The mechanics of OT

Answer 5

CON, GEN, & EVAL — not helpful. Ignore these terms.

How it works.
Start with the UR (aka ‘input’)
Create a list of ‘candidates’- (possible SRs, including the one that is the actual SR)
The actual SR is called the ‘optimal candidate’
Apply a set of constraints to these candidates
Determine which constraints are meaningful for that language

Question 6

Types of constraints

Answer 6

Markedness: phonological or phonetic based processes (what we’ve seen in the ‘rules’) 
Agreement (AGREE)
Consonant clusters (*CC)
Vowel rounding (VR)
…
Faithfulness: keep UR and SR identical
DEP (‘depend’): don’t allow epenthesis
MAX (‘maximal’): don’t allow deletion
IDENT (SR and UR should be identical in some specified feature: IDENT-PLACE, IDENT-VOICE, etc)

Question 7

2 possible paths in OT

Answer 7

1. We know the constraint ranking…. we should be able to accurately predict the surface form
2. We know the accurate surface form….we can determine the optimal ranking of constraints

Question 8

Optimality Theory (OT): Some Examples

Answer 8

Both languages don’t allow VV structures (lots of languages don’t)

In Yoruba, when 2 vowels are adjacent, one gets deleted
Deletion rule applies

In Basque, when 2 vowels are adjacent, a consonant gets inserted
Epenthesis rule applies

Can we explain this using rule-based phonology?
Can we explain this using OT?

Question 9

An Example

Answer 9

Vowel deletion in Yoruba:
[ri] ‘to see’
[aʃɔ] ‘cloth’
[r-aʃɔ] ‘to see cloth’

ri + aʃɔ => r-aʃɔ

V -> Ø / __ V

Question 10

Example: What constraints do we need?

Answer 10

Possible constraints
MAX: don’t delete any segments from the UR to the SR (faithfulness)
*VV: don’t have a sequence of 2 vowels (markedness)

Why are MAX and *VV the constraints we need?

• We are focusing on a deleted segment- MAX is a constraint that is based on deleting segments
• We also see (from our rule) that the 1st vowel likely gets deleted because there is a VV sequence (we know that many languages don’t allow 2 vowels together. So, this is a marked pattern. Hence, the markedness contraint *VV

Question 11

Compare

Answer 11

Yoruba: *VV, DEP&raquo_space; MAX
A MAX violation (i.e. deletion) is preferable to a *VV (2 V sequence) violation or DEP violation (i.e. epenthesis)
Basque: *VV, MAX&raquo_space; DEP
A DEP violation is preferable to a *VV violation or MAX violation

Question 12

Describe the results

Answer 12

2 options
In prose:

DEP is ranked lower than MAX and *CC. The language allows DEP to be violated but not the other 2

Using the ranking notation:
MAX, *CC&raquo_space; DEP

Question 13

Understanding the data

Answer 13

Input: /n-kopa/
Candidate set: [nkopa], [ŋkopa], [mkopa]
Constraints: AGREE-NASAL PLACE&raquo_space; IDENT- PLACE
AGREE-NASAL-PLACE = nasals agree in place with following stop
IDENT-PLACE = Consonants should have the same place of articulation from SR to UR (IDENT features are a matching from SR to UR)

what does the ranking mean?
This language values AGREE-NASAL-PLACE over IDENT-PLACE
The optimal candidate will violate IDENT-PLACE (because it isn’t important)