1
It is based around 3 principles
Gen
Con
Eval
Generate all possible candidates
Constrain the set of candidates
Evaluate the candidates with relation to the constraints available.
If language is innate, all constraints should be available for every language.
But obviously not all constraints are used in every language. The constraints a language uses are considered active.
If all the constraints can be considered, the generated candidates should be infinite.
BUT obviously this would be quite the task for the brain.
It's important to consider:
OT does not represent what linguists think happens in the brain.
The candidates are evaluated based on an ordering of the constraints.
are really more like preferences.
These
are not absolute.
are typically divided into 2 categories.
In order to understand these constraints: here's a brief breakdown of the components of a syllable.
A syllable can have 3 parts.
syl.a.ble
syllable boundaries are marked with a .
At minimum, a syllable must have a nucleus.
syl
Now, back to business.
Which constrain well-formedness.
Which constrain changes from the input.
Markedness Constraints
Syllables should have onsets.
Syllables should not have codas.
Onsets should not be complex.
- Onset
- NoCoda
- *Complex
Faithfulness Constraints
No segments should be deleted.
Don't epenthesize.
- Max
- Dep
These are just a small sampling of constraints.
Constraints
Select the least bad option.
How?
Each language uses a different ranking of its active constraints.
Candidates which violate highly ranked constraints are considered worse than candidates which violate more lowly ranked constraints.
Through OT we can select the candidate that violates fewer of the more valued constraints than the others.
We can illustrate OT with tableaux.
This is where we list our candidates from Gen.
/input/
When generating candidates, always include a candidate which is fully faithful to the input.
in.put
Next, we can create candidates which improve upon each of the constraints under consideration.
Violations are marked with * ! marks elimination
Since all the candidates violate Onset and NoCoda constraints at least once, they are considered equally bad and only an additional violation will eliminate them.
OT analysis is performed column...
by column.
Onset
*
*
**!
**!
*Complex
*!
NoCoda
**
*
**
*
Max
*
Although Max does have a violation, it is not necessary in this particular example, since in.ut was previously eliminated.
If the ranking of NoCoda were higher than that of Onset, the incorrect form would be selected.
Incorrectly selected candidates are indicated with a bomb.
The winning candidate is traditionally indicated by a pointing finger or arrow.
All of this
is
Optimality Theory
and
Onset
Coda
segments after the vowel
segments before the vowel
Nucleus
the vowel
i.nput
inp.ut
in.ut
/input/
in.put
Onset
*
*
**
**
NoCoda
**!
*
**!
*
i.nput
*Complex
*!
Max
*
inp.ut
in.ut