Optimality Theory - Phonological Theory I - Notes | LING 200A, Study notes of Linguistics

Download Optimality Theory - Phonological Theory I - Notes | LING 200A and more Study notes Linguistics in PDF only on Docsity! Oct. 14, 2008 1 Class 6: Optimality Theory, part I To do • Finish Prince & Smolensky excerpt (SQs due Thursday) • Start working on beginning OT problem (due Tuesday) 1. The “conceptual crisis” (Prince & Smolensky p. 1) Since Kisseberth 1970, constraints were taking on a bigger and bigger role. But… • What happens when there’s more than one way to satisfy a constraint? We need to prioritize the rules that could be triggered. • Why aren’t constraints always obeyed? • Relatedly, what happens when constraints conflict? What if one constraint wants to trigger a rule, but another wants to block it? We need a way of prioritizing constraints. • Should a rule be allowed to look far ahead in the derivation to see if applying alleviates a constraint violation? Or does the alleviation have to be immediate? • Can a constraint be against making a certain type of change, rather than against a certain structure? 2. Prince & Smolensky’s solution: Optimality Theory rule-based grammar with constraints OT grammar start with UR/input (from mental lexicon) apply rules in sequence—intermediate representation is known at all times apply all possible rules, producing a (large!) set of candidate outputs constraints may block or trigger rules constraints pick the best candidate look-ahead is nonexistent or sketchy since the candidate outputs are all potential surface forms, there is full look-ahead to the end of each possible derivation interaction of constraints is nonexistent or sketchy constraints interact through strict domination similarity to UR is the result of not applying too many rules and not having too many constraints similarity to UR is enforced by faithfulness constraints end with SR/output (send it to the phonetic system) 3. Gen() This is the function that creates the set of candidate outputs from the input. One way to think of it:1 apply all possible rules to the input, any number of times. Each underlying segment can be deleted or have its features changed; extra segments can be inserted anywhere; underlying segments can change their order 1 This is what P&S call ‘anharmonic serialism,’ except with a universal set of rules that’s broad enough that the result is the “all possible variants” that P&S propose. Ling 200A, Phonological Theory I. Fall 2008, Zuraw 2 Gen(/ab/) = {[ab], [a], [b], [ba], [], [ta], [at], [ae], …} /ab/ ab a b ba tab atb abt eab aeb abe ib ob ap am… Ø ta at ae ea i o … o Why is the resulting set of candidates infinite (assuming a finite alphabet of symbols)? 4. Constraints In standard OT, we can think of each constraint as a function from a candidate output to a natural number (the number of violations). NOCODA([bak]) = 1 NOCODA([tik.pad]) = 2 Alternatively, we can think of each constraint Ci as imposing a strict partial ordering
i (“is more harmonic than with respect to Ci”) on a set of candidates, with the following additional properties: • The ordering is stratified: If a  b and b a, then for any i
a, i
b too; and for any j such that a
j, b
j too. (We can say that if a  b and b a, a and b are of equivalent harmony.) • There exists some a such that there is no i
a. (That is, one or more candidates are the most harmonic; there are not necessarily one or more least harmonic candidates, though.) A strict partial ordering is transitive, irreflexive, and asymmetric: • Transitive: if a
b and b
c, then a
c. • Irreflexive: a  a. • Asymmetric: If a
b, then b a. (In Colin Wilson’s targeted-constraints variant of OT, the stratification requirement is relaxed.) NOCODA: ta.da bo ba.du.pi tak.do i.tek o.tek.lao tak.kat bad.ku.pit sik.lep.bu … o Why does assigning a non-unique natural number (0, 1, 2, …) to each candidate meet the ordering requirements above? o If you’re ahead in the reading, can you recall a case from P&S where numbers of violations weren’t used? o Why are there no least-harmonic candidates for NOCODA? Ling 200A, Phonological Theory I. Fall 2008, Zuraw 5 This tableau shows a ranking argument: we have two candidates that differ in that NOCODA prefers a (the winner), whereas DEP-V prefers b. If that’s the only difference between the candidates—there is no other constraint that prefers a over b—then NOCODA must outrank (>>) DEP-V. /at+ka/ NOCODA DEP-V
a [a.t.ka] * b [at.ka] *! Parts of the tableau: • input • output candidates • constraints (highest-ranked on left) • asterisks • exclamation marks • shading • pointing finger (or you can use an arrow) 9. How do I know which candidates and constraints to include in my tableaux? Here is a procedure that usually works reasonably well: • Start with the winning candidate and the fully faithful candidate. • If the winning candidate
the fully faithful candidate…  Add the markedness constraint(s) that rule out the fully faithful candidate.  Add the faithfulness constraints that the winning candidate violates.  Think of other ways to satisfy the markedness constraints that rule out the fully faithful candidate. Add those candidates, and the faithfulness and markedness constraints that rule them out. You have to use your judgment in deciding how far to take this step. • If the winning candidate = the fully faithful candidate, then you are probably including this example only to show how faithfulness prevents satisfaction of a markedness constraint that, in other cases, causes deviation from the underlying form.  Add that markedness constraint.  Add one or more candidates that satisfy that markedness constraint.  Add the faithfulness constraints that rule out those candidates. o Let’s try it for /atka/  [atka]. o One of the candidates below is unnecessary in arguing for the constraint ranking. Why? /at+ka/ *CC DEP-V
a [atka] * b [atka] *! c [atka] **! A candidate is harmonically bounded if it could not win under any constraint ranking. These three don’t add any new information, but are there for the convenience of the reader. Ling 200A, Phonological Theory I. Fall 2008, Zuraw 6 10. Comparative tableaux An innovation of Alan Prince. They convey the same information, but in a different form /at+ka/  [atka] *CC DEP-V a [atka] vs. [atka] W L b [atka] vs. [atka] W Comparative tableaux are nice because you can easily see if your ranking is correct: the first non- blank cell in each row must say W. We also see easily why [atka] is irrelevant to the ranking. 11. Exercise: Metaphony (the two easy cases) o Develop an OT account of these two metaphony systems. Foggiano/Pugliese pte ‘foot’ píti ‘feet’ móa ‘soft (fem.)’ múu ‘soft (masc.)’ kjéna ‘full (fem.)’ kjínu ‘full (masc.)’ gr ssa ‘big (fem.)’ grússu ‘big (masc.)’ Veneto védo ‘I see’ te vídi ‘you see’ kóro ‘I run’ te kúri ‘you run’ prte ‘priest’ prti ‘priests’ blo ‘beautiful (masc. sg.)’ bli ‘beautiful (masc. pl.)’ m do ‘way’ m di ‘ways’ gáto ‘cat’ gáti ‘cats’ 12. Another exercise: English regular plurals -z ‘peas’ -s ‘blokes’ 
-z ‘toes’ 
-s ‘coughs’ -z ‘dolls’ - ‘glasses’
-z ‘pans’ - ‘fizzes’ -z ‘dogs’ - ‘branches’  -z ‘labs’ - ‘badges’ 
-z ‘kilns’  - ‘wishes’ 
 -s ‘clasps’ - ‘garages’  -s ‘mitts’ Each line compares the winner to one losing candidate, and shows whether each constraint prefers the winner (W) or the loser (L)