Constraint Conjunction Ties Opacity Plan of this unit

Constraint Conjunction, Ties, Opacity Plan of this unit. Discussion of Optionality Introduction of Ties Technicalities Cumulative Effects Introduction of Constraint Conjunctions The Limits of the Approach 11/28/2020 1

Constraint Conjunction, Ties, Opacity We saw yesterday that the relation between input and output is not always one-to-one but that some inputs lead to ineffability. Today we will see more cases of more complex relations between inputs, outputs and candidates. . 11/28/2020 2

Optionality The reverse of ineffability is a case where several outputs emerge as optimal, not just one. This is a case of optionality or free variation. A phenomenon like optionality or free variation implies that the grammar must be flexible enough to allow competing expressions to emerge. . 11/28/2020 3

Multiple outputs: optionality Different ways to express ‘Who did you see? ’ in French: Tu as vu qui? Qui as-tu vu? Qui est-ce que tu as vu? C’est qui que tu as vu? 11/28/2020 4

Multiple outputs: optionality Expletive Insertion vs. Movement there was a moose shot a moose was shot Dative Shift I gave a book to Mary I gave Mary a book 11/28/2020 5

Multiple outputs: optionality Complementizer Deletion I think (that) she is intelligent Extraposition / HNPS / Particle Shift a man from India arrived a man arrived from India she looked up the interesting answer she looked the interesting answer up 11/28/2020 6

Multiple outputs: optionality Scrambling dass niemand das Buch gelesen hatte that nobody the book read had dass das Buch niemand gelesen hatte Topicalization das Buch hatte niemand gelesen hatte das Buch niemand hatte das Buch gelesen 11/28/2020 7

Multiple outputs: optionality Free variation in the phonology of German a. Segmental alternation [taç] and [ta: k] for Tag ‘day’ [g´reo: n] and [ge: reo: n] for Gereon ‘a name’ b. Alternation in stress position Télefon vs. Telefón 11/28/2020 8

Multiple outputs: optionality Free variation in the phonology of French a. Choice of an alternant ancien ‘old’: [ãsj´n] and [ãsj´~n] in the masculine liaison case (un ancien ami) b. Glide formation ouest [w´st] ~ [u. ´st] ‘west’, nuage [nu. aΩ] ~ [n¥aΩ] ‘cloud’, piano [pja. no] ~ [pi. a. no] 11/28/2020 9

Solutions 1. 2. 3. 4. Optionality is only apparent Identity of profiles Co-Phonologies (Co-syntaxes) Tied Constraints 11/28/2020 10

1. Apparent Optionality First, optionality may be only apparent. There IS a meaning difference between der Pfarrer kommt ‘the priest comes’ and es kommt der Pfarrer 11/28/2020 11

1. Apparent Optionality The same would seem to hold for a moose was shot there was a moose shot Scrambling Different Topicalizations One can encode such differences in terms of semantic/pragmatic features. 11/28/2020 12

1. Apparent Optionality If these are in the input, than structures do not really compete with each other in instances of apparent optionality. Likewise, dative shift may involve a difference in lexical composition. 11/28/2020 13

2. Identical violation profiles Second, EVAL will not always be able to differentiate between the options. . . Suppose e. g. that complementizers do not induce a violation of Full Interpretation. . . 11/28/2020 14

2. Identical violation profiles Then I think he will come I think that he will come can be derived from the same input, and they have identical constraint violation profiles (Grimshaw) --> both can be grammatical (Dative Shift) 11/28/2020 15

Solution 3 : cophonologies Within phonology, quite a different solution has been proposed: Co-phonologies are parallel phonologies for different parts of the phonology. (Co-phonologies as a way to account for free variation is usually dismissed because too permissive. Most cases which have been explained with co-phonologies can be explained otherwise. ) 11/28/2020 16

Cophonologies Typical cases imply the co-existence of two systems of stress patterns in a language (Turkish is a standard example, and German, too). We saw that in multi-stratal approaches, each stratum defines its own phonology. Remember class, classy, classic and levels. 11/28/2020 17

Cophonologies The most relevant case for OT is stratification of the lexicon. Different parts of the vocabulary can define different strata: some segments, stress patterns, phonotactic generalizations and the like can be specific to some strata and absent in others. In German, final full vowels are typical for nonnative words: Auto, Menü, Biologie… 11/28/2020 18

Cophonologies A widely accepted view is that the lexicon is organized concentrically (see Ito & Mester for Japanese, Féry for German). In the center there are the native words, obeying a strict phonology. A great deal of markedness constraints are active there. Going away from the center, words are less and less assimilated nonnative words. The less assimilated words are, the less markedness constraints they fulfill and the more faithful they are to their source language. 11/28/2020 19

Cophonologies 1. Native vocabulary 2. Assimilated foreign 3. Unassimilated foreign 1 2 3 11/28/2020 20

Cophonologies Each stratum formed by some part of the vocabulary (Germanic, Latinate, Sino-Chinese, unassimilated…) is a co-phonology. It must been observed that ideally the markedness constraints are organized in just one hierarchy. Words fulfill the constraints up to a certain point, a different one for each stratum. Thus cophonologies are just partial. 11/28/2020 21

3. Cophonologies Free variation as cophonologies implies that one word can be in one stratum for one speaker and in another stratum for another speaker (genre in English, city in German…) Example in Japanese Citybank (a Japanese bank) is pronounced [∫it∫i], [∫iti] or [siti] bank 11/28/2020 22

3. Cosyntax Syntax shows related phenomena, though they are typically ignored. Consider e. g. the Germanic co-syntax of English Restricted V/2: In the garden stands a fountain „I am sick“ said the ugly stranger Rules of English proper must not follow Engliman *Does in the garden stand a fountain? 11/28/2020 23

4. Ties are two (or more) constraints of the same rank. In case it is these constraints which decide on the optimality of candidates, the result of a tie is two or more different optimal outputs. This solution differs from the identity of profiles solution since the optional candidates have different profiles. 11/28/2020 24

4. Ties can be interpreted differently. In the first interpretation, two hierarchies define simultaneous grammars from a certain point up (this is a case of cophonology and cosyntax). The result is two or more different optimal outputs. 11/28/2020 25

First interpretation of ties C 2 a >> C 2 b >> C 3 … / …C 1 >> C 2 b >> C 2 a >> C 3 … or C 1 >> C 2 a >> C 2 b >> C 3 … C 1 >> C 2 b >> C 2 a >> C 3 … 11/28/2020 26

First interpretation of ties Given the hierarchy A. . B C 1/C 2 D. . . E If C 1 and C 2 are tied by hierarchy, then S is grammatical iff S is optimal with respect to A. . . B C 1 D. . . E or A. . . B C 2 D. . . E 11/28/2020 27

Multiple outputs: optionality 11/28/2020 28

A Tie between Constraints in Syntax Pesetsky-style treatment of complemetizers: A different solution for I think (that) he will come LE(CP): A CP must begin with a complementizer (Align (CP, COMP, left) TEL: Do no pronounce function words 11/28/2020 29

Complementizers in embedded clauses (1) (2) 11/28/2020 I think that he is a fool I think he is a fool 30

Relative Clauses (1) (2) (3) (4) a a man who that I like man who I like man that I like man I like Candidate (4) is eliminated because it violates both Tel and LECP. The other candidates violate either Tel or LCPC and are thus all optimal. Mixed case of identity of profiles and ties. 11/28/2020 31

Relative Clauses 11/28/2020 32

Possible drawback Since the candidates differ in their profiles it can be the case that a lower ranking constraint decide to eliminate one of the candidate which was chosen as optimal by the tie. 11/28/2020 33

Second interpretation of ties The tie is defined in the same way as before, but the remaining of the hierarchy is identical. The drawback identified above is eliminated. 11/28/2020 34

Second interpretation of tie C 2 a >> C 2 b / C 1 >> >> C 3 … / C 2 b >> C 2 a or C 1 >> C 2 a >> C 2 b >> C 3 … C 1 >> C 2 b >> C 2 a >> C 3 … 11/28/2020 35

Relative Clauses 11/28/2020 36

Third interpretation of ties In this concept of a tie, the number of violations of the tied constraints taken together is relevant for evaluation C 1 >> C 2 a + C 2 b >> C 3 … Such a definition is needed when more than two candidates emerge as optimal. 11/28/2020 37

Ties between Constraints Given the hierarchy A. . B C 1/C 2 D. . . E If C 1 and C 2 are tied cumulatively, then S is grammatical iff S is optimal with respect to A. . . B F D. . . E where F is the sum of violations of C 1 and C 2 11/28/2020 38

Pesetsky’s Concept of a Tie Wh-expletive insertion in German seems to be another case in point: any combination of Stay and FI violations yields a grammatical result 11/28/2020 39

A Tie between Constraints 1. Wen denkst du t dass sie meint t dass Fritz liebt who think you that she believes that Fritz loves 2. was denkst du wen sie meint t dass Fritz liebt 3. was denkst du was sie meint wen Fritz liebt 4. was denkst du t dass sie meint wen Fritz liebt 11/28/2020 40

A Tie between Constraints FI 1. 2. 3. 4. Wen … t … was … wen … was … t … wen … 11/28/2020 Stay ** * * * 41

A Tie between Constraints In interpretation 1 and 2, only the first and third candidates can emerge as optimal. 11/28/2020 42

The Need for a further concept Ich denke / I think 1. dass der Fritz nicht geschlafen hat the Fritz not slept has 2. der Fritz hat nicht geschlafen 1. 2. FI Stay (Comp) and Stay (prefield) A Pesetsky-style tie would favor 1. ! GENERATE TWO HIERARCHIES 11/28/2020 43

Another case in point Scrambling as an instance of multiple hierarchies (Uszkoreit) 1. nom > acc/dat 2. animate > inanimate 3. definite > indefinite NOM ANIM DEF A sentence is grammatical if it satisfies at least one constraint 11/28/2020 44

Ties between Constraints Dass der Mann ein Buch liest that the man a book reads NOM, DEF, ANIM *dass ein Buch der Mann liest *NOM, *DEF, *ANIM dass ein Mann das Buch liest that a man the book reads NOM, ANIM, *DEF dass das Buch ein Mann liest *NOM, *ANIM, DEF 11/28/2020 45

Ties between Constraints Dass ein Buch der Frau hilft that a book the woman helps NOM, *DEF, *ANIM dass der Frau ein Buch hilft *NOM, DEF, ANIM dass das Buch einer Frau hilft NOM, DEF, *ANIM dass einer Frau das Buch hilft *NOM, *DEF, ANIM 11/28/2020 46

Ties between Constraints A Pesetsky style concept of a tie would incorrectly predict that structures with n violations are blocked by structures with nk violations. What we need for such examples is a concept of ties in which complete hierarchies are tied. . . 11/28/2020 47

Lexicographic Conflict Resolution Recall that conflict resolution in OT is lexicographic: there is a hierarchy H of constraints, and C is better than D relative to H iff D violates the highest constraint on which C and D differ more often than C 11/28/2020 48

Lexicographic Conflict Resolution A number of proposals have been made which imply that conflict resolution is not always lexicographic 11/28/2020 49

Constraint conjunction Two cases of constraint conjunction: selfconjunction of one constraint and conjunction of different constraints. Universal ranking schema: C 1 & C 1 >> C 1 (self-conjunction) C 1 & C 2 >> C 1 , C 2 (conjunction of different constraints) 11/28/2020 50

Constraint conjunction 1. Self-conjunction of constraints: it is worse to violate the same constraints n times than to violate it n-1 times. 2. Conjunction of different constraints: we will see that some typical derivational effects have been accounted for with the help of constraint conjunction. When do we need local constraint conjunction? 11/28/2020 51

Constraint conjunction Chain shift A -> B, B –> C but not: A –> C In Western Basque (Etxarri dialect), mid vowels raise to high, and high to raised (Kirchner 1996) Indef e –> i seme bat o –> uasto bat i –> ij erri bet u –> uw 11/28/2020 buru bet Def semi-e‘son’ astu-e ‘donkey’ errij-e ‘village’ buruw -e ‘head’ 52

Constraint conjunction Raised {ij, uw}: [-low, +high, +raised] High {i, u}: [-low, +high, –raised] Mid {e, o}: [-low, –high, –raised] Low {a}: [+low, –high, –raised] In a serial approach, this is not a problem: raising from high to raised is ordered before raising from mid to high But in standard OT this is difficult to express. 11/28/2020 53

Constraint conjunction HIATUS-RAISING: In V 1 V 2, maximize height of V 1. This constraint is gradient: a is 3 violations, mid vowels 2, high 1 and raised none. Faithfulness: IDENT-IO(high): If an input segment id [ahigh], then its output correspondent is [ahigh] IDENT-IO(raised): If an input segment id [araised], then its output correspondent is [araised] 11/28/2020 54

Wrong results HIAT-RAIS IDENT(high) IDENT (rais) a. i e –> e*!* a. ii e –> i a. iii e –> ij b. ii i –> ij 11/28/2020 * * *! * 55

Right results with constraint conjunction [ID(high) & HIAT-RAIS ID(high) ID(raised)] e –> e *! e –> i * e –> ij *! * * i –> i *! i –> ij 11/28/2020 ID(rais) * 56

Constraint conjunction Another nice example: Rendaku in Japanese : /ore-kami/ –> [ore-gami]: voicing of the first obstruent in the second part of a compound The application of Rendaku is limited by Lyman’s Law: ‘Only one voiced obstruent per morpheme’ /kami-kaze/ –> [kami-kaze] *kami-gaze. 11/28/2020 57

Constraint conjunction /ore-kami/ *voic. Obstr 2 ore-gami ore-kami /kami-kaze/ kami-kaze kami-gaze 11/28/2020 Rendaku * *! *voic. Obstr * ** 58

Constraint conjunction 2. Conjunction of different constraints: only markedness, only faithfuness, both kinds (Lubowicz, Ito & Mester)? To imitate the effect of the strict alternant condition (only segments subject to allophony can be subject to a rule (or the effect of a markedness constraint), it seems that faithfulness and markedness must be conjoined. 11/28/2020 59

Chomsky on Barriers in „Barriers“ In his 1986 book „barriers“, Chomsky proposed that we measure the distance between a phrase and its trace in terms of the numnber of barriers that have been crossed. 11/28/2020 60

Chomsky on Barriers in „Barriers“ O barriers what do you fix 1 barrier what do you wonder how to fix 2 barriers ? ? what do you wonder how one should fix 11/28/2020 61

Chomsky on Barriers in „Barriers“ If it is true that languages/ construction types may differ as to how many barriers may be crossed, then we need to be able to rank e. g. the PARSE constraint between k barriers crossed > PARSE > k-1 barriers crossed CERTAINLY: k barriers crossed should not be an atomic constraint 11/28/2020 62

Self conjunction of constraints Recall for self-conjunction of constraints: We say that CONk is violated if CON is violated at least k times. . . It seems necessary to assume that CONk >> PRIN >> CONk-1 11/28/2020 63

Locality of self conjunction It seems more adequate to say that CONk is violated if CON is violated at least k times by the same element/in the same domain! 2 violations by SAME element *? What do you wonder who bought 2 violations by DIFFERENT elements what he wonders how to fix has an influence on what I wonder when to fix 11/28/2020 64

Conjunction of different constraints What do you wonder how to fix t *how do you wonder what to fix t Adjuncts have to fulfill stricter locality requirements than arguments. . . REF: a chain is not headed by an adjunct 11/28/2020 65

Conjunction of different constraints how do you think that she did it *how do you wonder when to fix the car what do you wonder when to fix BAR 1 & REF > Parse. Scope > BAR 1 11/28/2020 66

An obvious problem By allowing constraint conjunction, the weighting (compensatory) type of conflict resolution can be represented in OT -->> OT gets less restrictive 11/28/2020 67

An obvious problem A is more important than B, and A is more important than C, but B and C together outrank A B&C>A>B>C 11/28/2020 68

Compensatory Effects in NL? Q-Scope in German Pafel proposes the following principles PREF: a quantifier in the prefield takes wide scope NOM: A nominative quantifier takes wide scope DIST: Inherently distributive quantifiers take wide scope 11/28/2020 69

Compensatory Effects in NL? Jeder Pianist hat eine Fuge gespielt every pianist has a fugue played by pref, nom & dist Jede Fuge hat ein Pianist gespielt nom alone does not win over pref Ein Pianist hat jede Fuge gespielt 11/28/2020 dist alone does not win over pref 70

Compensatory Effects in NL? Eine Fuge hat jeder Pianist gespielt and dist and nom win over pref A factual problem jeden Studenten hatte ein Pianist aus Polen empfangen each-acc student had a pianist from Polen received Grammaticality vs. Parsing Ease? 11/28/2020 71