Gisbert Fanselow & Caroline Féry University of Potsdam Introduction to Optimality Theory 3/19/2018 1
Basic goals of the class • Introduction of the key concepts of Optimality Theory • both in phonology and in syntax • identification of its precursors • discussion of OT’s successes • extensions of OT • applications of OT in processing and acquisition 3/19/2018 2
The Plan for Today 1. 2. 3/19/2018 A brief sketch of OT Brief introduction of the issue: representational (constraint related) or derivational accounts? Three types of arguments for a constraint based view in phonology and syntax 3
A brief sketch of OT OT is a grammatical architecture. It involves a set of constraints on linguistic representations. The constraints may be in conflict with each other Conflicts are resolved by lexicographic constraint hierarchies. which are language-particular 3/19/2018 4
A brief sketch of OT Two plausible constraints from the domain of syntax: A category should begin with its head PP: in the park VP: kiss Mary in the park AP: proud of Mary Subjects should come first 3/19/2018 5
A brief sketch of OT The principles get in conflict with each other when a categories has both a head and a subject. Conflict resolution is lexicographic if the principles appear in a hierarchy, such that conflicts are always resolved in favor of the higher principle. 3/19/2018 6
A brief sketch of OT Irish: Heads always go first duirt Seán go-bhfuil Cathal ag rince said John that-is Charles –ing dance English: Subjects are initial John claims that he likes Mary *Claims John that likes he Mary 3/19/2018 7
The Plan for Today 1. 2. 3/19/2018 A brief sketch of OT Brief introduction of the issue: representational (constraint related) or derivational accounts? Three types of arguments for a constraint based view in phonology and syntax 8
Constraints play a crucial role in the grammatical architecture of OT. They govern the well-formedness of phonological output forms, and of syntactic surface structures. Grammars need not put so much emphasis on constraints. but those that do seem to fare particularly well in a number of domains. 3/19/2018 9
Derivational Approaches I The alternative to constraintbased/representational models is the derivational approach that figured prominently in the early days of generative grammar. A derivational account explains grammatical wellformedness by formulating rules that generate exactly the set of grammatical items. 3/19/2018 10
Derivational Phonology In derivational phonology (as in Chomsky & Halle 1968. The Sound Pattern of English, SPE) rules derive a surface representation from a deep structure. Deep structure Rules Surface structure 3/19/2018 11
Derivational Phonology Derivational models make use of explicit rules for generating the surface expressions of a language. The surface generalizations are however, expressed independently by the phonotactics of the individual languages. 3/19/2018 12
Derivational Phonology Typical domains of application of derivational rules are allophonies and alternations (in SPE also nonalternating surface structures). 3/19/2018 13
Derivation of long in SPE (p. 211) An example of derivation of a nonalternating form from an abstract underlying representation Underlying representation /long/ Tensing Rule l√€ng √€ _ + tief la€ng Diphthongization la€wng Glide Vocalization Rule la€ung 3/19/2018 14
Derivation of long in SPE (p. 211) Vowel Shift Rounding Adjustment Nasal assimilation g-Tilgung Surface form 3/19/2018 la€ong lø€√˜g lø€√˜ [lø√˜] 15
Why derivational accounts? Why do we need derivations in phonology at all? In some cases, a rule needs an environment to apply which is derived by another, independent rule. This is an instance of feeding. 3/19/2018 16
![A Feeding Example Chukchi (1) r –> t/ _ [+coronal] pkir-k ‘to arrive’ pkit-tk](https://present5.com/presentation/237e0bdfba9144e6740d1894f8171dc2/image-17.jpg "A Feeding Example Chukchi (1) r –> t/ _ [+coronal] pkir-k ‘to arrive’ pkit-tk")
A Feeding Example Chukchi (1) r –> t/ _ [+coronal] pkir-k ‘to arrive’ pkit-tk ‘you (pl) arrived qeper ‘glutton’ qapat-t∫˜-n ‘big glutton’ 3/19/2018 17
![A Feeding Example (2) [–sonorant] –> [+sonorant, nasal]/ –[+nasal] ©e-mne-lin ‘he ground’ pne-k ‘grind’](https://present5.com/presentation/237e0bdfba9144e6740d1894f8171dc2/image-18.jpg "A Feeding Example (2) [–sonorant] –> [+sonorant, nasal]/ –[+nasal] ©e-mne-lin ‘he ground’ pne-k ‘grind’")
A Feeding Example (2) [–sonorant] –> [+sonorant, nasal]/ –[+nasal] ©e-mne-lin ‘he ground’ pne-k ‘grind’ rmn- t ‘flesh sides of hides’ rpn ‘id’ 3/19/2018 18
A Feeding Example (3) kun-nin kur-k ‘he bought’ ‘to buy’ (3) shows that (1) must apply before (2). [r] of kur-k first becomes [t] which then can become a sonorant again (but this time a nasal as a result of assimilation). 3/19/2018 19
A Further Feeding Example Take the underlying form /lang/ ‘long’ in German Apply first the rule of nasal assimilation: +nasal –anterior +cons +coronal –> -coronal / – -ant +high -cor +high 3/19/2018 20
A Further Feeding Example Then apply the rule of g-deletion: g –> ø / ˜ _ Co +syllabic +stressed Surface structure: [la˜] (Wurzel 1980: 960 ) On the surface there is a dorsal nasal motivated by the presence of [g] in the deep structure. 3/19/2018 21
Derivationalism in Syntax Modern Syntax also started out as a derivational theory. The so-called Standard Theory has been one of its most influential versions LEXICON 3/19/2018 Deep Structure Surface Structure 22
From the Lexicon to D-Structures Explicit phrase structure rules S NP VP John kisses Mary NP Det A N the nice man VP V sleeps VP V NP kisses Mary VP V PP laughs at Mary VP V NP PP gave it to Mary VP V NP NP gave him a book VP V NP S‘ told him that. . . 3/19/2018 23
From deep structures to the surface Explicit transformations X NP V (P) NP Y 1 2 3 4 5 6 => 1 5 be V+en 5 by+2 6 that John laughs at Mary today that Mary is laughed at by John today 3/19/2018 24
Why derivations in syntax? Many theories believe that wh-phrase move to clause-initial position in a number of derivational steps. . . do you think that he claims that she will sell what do you think that he claims what that she will sell do you think what that he claims that she will sell what do you think that he claims that she will sell 3/19/2018 25
Why derivations in syntax? In more complex cases, each of these steps can be followed by the licensing of reflexive pronouns which is strictly bound to the next subject: Which picture of herself do you think that he claims that she will sell? do you think that he claims that she will sell which picture of herself ? 3/19/2018 26
Why derivations in syntax? But we also get: Which picture of himself do you think that he claims that she will sell? do you think that he claims which picture of himself that she will sell? It is hard to see how this can be explained with reference to tiny derivational steps. . . 3/19/2018 27
Representational aspects Derivational theories tend to have a „representational“ residue. . . Phonotactic Constraints (concerning e. g. the basic syllabic makeup of English or German) were stated in addition to rules --> constraints on the form of lexical entries No syllable-initial nasal followed by a stop (Co. lo. mbo) 3/19/2018 28
The Plan for Today 1. 2. 3/19/2018 A brief sketch of OT Brief introduction of the issue: representational (constraint related) or derivational accounts? Three types of arguments for a constraint based view in phonology and syntax 29
Problems of derivationalism In spite of the advantages discussed above, purely derivational models were criticized for at least 3 reasons 1) They were accused of being too unconstrained 2) They imply that information is duplicated in grammar 3) They lead to the rule conspiracy problem 3/19/2018 30
Why is derivational phonology unsatisfying? 1. Rules and their ordering are unconstrained. All kinds of processes can be accounted for: natural and unnatural, marked and unmarked, possible and impossible ones. Possible and natural rule: rule [coronal] –> [labial] / _ [coronal] (tp –> pp) Impossible and unnatural rule: rule [voice] –> [labial] –> _ [voice] (dd –> bd) 3/19/2018 31
Why is derivational phonology unsatisfying? The complexity of the two rules is the same. If features are unorganized, assimilation is predicted to happen between unrelated features. 3/19/2018 32
Why is derivational phonology unsatisfying? A specific rule ordering is determined by the need of a single language. Two languages or two dialects of a single language can have the same rules, but with a reverse ordering. 3/19/2018 33
Two dialects of German and g-deletion First dialect (Standard German) la˜g Assimilation of ˜ to g la˜ g-deletion – Final Devoicing /lang/ Second dialect (Northeren German) la˜g Assimilation of ˜ to g lank Final Devoicing – g-deletion /lang/ 3/19/2018 [la˜] [la˜k] 34
Impossible rule orderings Pseudo-Yokuts (Kiparsky) In real life, some rule ordering are not possible Underlying /maat/ R 1: Epenthesis maati (the trimoraic syllable is resolved) R 2: Palatalisition maatji ([t] –> [tj] before [i]: R 1 feeds R 2) R 3: Apocope maatj ([i] is deleted word finally) R 4: Shortening matj (the trimoraic syllable is resolved) Surface [math] (Mc. Carthy 2000) 3/19/2018 35
Constraining Rules in Syntax Typical syntactic rules of the early period were felt to be too unconstrained because they failed to express parallels between modules of grammar, and between the grammars of the world‘s languages. 3/19/2018 36
A set of early phrase structure rules VP VP VP 3/19/2018 V V V NP PP NP NP NP S‘ sleeps kisses Mary laughs at Mary gave it to Mary gave him a book told him that. . . 37
Problem 1 for phrase structure rules (All? ) Syntactic Categories are endocentric C is endocentric if the syntactic behavior of C (=the category of C) is determined by one part of C. X is a verb phrase because it contains a verb 3/19/2018 38
Problem 1 for phrase structure rules The early rule format is too unconstrained because it allows for non-endocentric categories as well: VP N PP VP Comp A N 3/19/2018 39
Solutions The X-bar-Scheme XP . . . X. . . where X is any category as a METACONSTRAINT on rules as A RULE + language particular statements such as the head must be left in VP, etc. 3/19/2018 40
Why is derivational phonology unsatisfying? Second problem 2. Duplication Problem: Rules and phonotactics replicate each other - Phonotactics says: 1) obstruent clusters share a single voicing specification; there are no geminates. 2) Rules trigger voicing agreement in obstruent clusters and eliminate geminates 3/19/2018 41
Why is derivational phonology unsatisfying? Second problem Plural in English: Plural morpheme is realized as [s], [z] or [iz] depending on the preceding segment. cats pets dogs cars judges churches In a rule approach, one allomorph is taken as underlying and the others are derived by rules: voicing agreement and [i] is inserted between two sibilants. 3/19/2018 42
The Subcategorization Problem The generative component of the syntax must specify what the simple syntactic configurations of a language are. . . The same type of information reappears in the lexical entries 3/19/2018 43
Some rules for German VPs VP VP VP 3/19/2018 V NP NP V PP V NP PP V 44
Positive results VP V VP NP V dass Hans that John weinte wept das Buch verfasste the book wrote VP NP NP V der Maria den Fritz vorstellte introduced Fritz to Mary VP PP V mit ihr stritt with her argued 3/19/2018 45
Negative results VP V dass Hans verfasste *wrote VP NP V das Buch rechnet the book reckons VP NP NP V der Maria den Fritz weinte wept Fritz to Mary 3/19/2018 46
Subcategorization Verbs fit into certain environments only befahren: requires an NP object „drive on“ vorstellen: requires two NP objects „introduce“ stellen: requires an NP and a PP „put“ 3/19/2018 47
Subcategorization existieren: does not combine with an object „exist“ This information must be specified in the lexicon 3/19/2018 48
Duplication And this information reappears in the phrase structure rules! What would happen if we enter a VP rule like VP V NP PP AP Comp A? Nothing! Because we have no verb to insert into such a structure 3/19/2018 49
Consequence: The rule can be trivial VP . . V. . because the rest is done by the lexicon Some theories claim: EVERYTHING is done in the lexicon 3/19/2018 50
Constraints on Variables: A related problem was discovered by Ross: Rules must obey the same or at least very similar constraints 3/19/2018 51
Coordinate Structures I love Mary and Bill *who do you love Mary and __ *John, I love Mary and __ *a man who I love Mary and __ *she is more intelligent than you are handsome and __ 3/19/2018 52
Subject Islands To invite Mary is fun *I wonder who to invite is fun *he said that John, to invite is fun *a man who to invite _ is fun *she is more intelligent than to be _ is fun 3/19/2018 53
Wh - Islands I wonder when I should kiss Mary *who do you wonder when I should kiss_ *John, I wonder when I should kiss _ *a man who I wonder when I should kiss _ *she is more intelligent that I wonder whether you are _ 3/19/2018 54
Adjunct - Islands I weep because I saw Mary *who do you weep because I saw _ *John, I wept because I saw _ *a man who I wept because I saw _ *she is more intelligent than I am envious because you are _ 3/19/2018 55
Ross‘s Proposal Movement rules are very simple They specify what can move to what kind of position They are constrained by a universal constraint component: No rule can involve X and Y if X and Y are separated by a wh-island. . 3/19/2018 56
Extensions of Ross‘s Proposal The movement rule is trivialized Move A from B to C! The Constraints check for the wellformedness of the output 3/19/2018 57
Why is derivational phonology unsatisfying? Third problem 3. Conspiracies of the rules (Kenstowicz & Kisseberth) Different rules have the same purpose, but this is not visible from the rules themselves. Consider hiatus in several languages. 3/19/2018 58
Avoidance of hiatus by vowel deletion French a. *le ami –> l’ami ‘the friend b. *je arrive cet après-midi –> j’arrive cet a. -m. Rule: V –> Ø / _ V (vowel deletion) • The rule is blocked if a morpheme would be left unrealized Elle a eu un enfant. 3/19/2018 59
Avoidance of hiatus by consonant epenthesis • German Chaot –> Cha? ot ‘chaotic person’, Ruin –> Ru? in • Axininca Campa i-N-koma-i –> i˜komati ‘he will paddle’ • Rule: ø –> C / V _ V (consonant epenthesis) • The rule does not apply in German if the second syllable is unstressed: Théo, Muséum 3/19/2018 60
Avoidance of hiatus by glide formation • Dutch: bioscoop –> bijoscoop ‘movie • Ilokano: da? o ‘kind of tree’ pag- da? o-an -> pagda? wan Rule: Ø -> j (glide formation) • The rule does not apply in some segmental environment 3/19/2018 61
Avoidance of hiatus by glide formation and compensatory lengthening • Luganda (Bantu) Kenstowicz & Kisseberth (1979) mu ami –> mw. ami Glide formation mw. ami–> mw aami CL ‘chief, pl. ’ • Vowel deletion + compensatory lengthening C high V V 1 2 3 –> 1+2 glide ø 3[long] 3/19/2018 62
Excursus: Ordering is crucial It must be observed that the ordering vowel deletion + compensatory lengthening is crucial. When the two rules are reversed, the context for lengthening is no longer present. mu ami –> mw. ami *[mw ami] 3/19/2018 CL is blocked Glide formation 63
Summary of conspiracy in phonology Rules like vowel deletion, consonant epenthesis and glide formation in the intervocalic environment ‘conspire’ to avoid hiatus. In other words, rules aim at something, but different rules can have the same aim. In a rewriting rule A – > B/ C _ D, the aim is CBD. 3/19/2018 64
Summary of conspiracy in phonology However, nothing in the format of the rules we just saw specifies what the rules try to achieve. Thus, one can formulate a rule B –> A/ C _ D, whose aim is CAD, the reverse of the rule above. If one looks at derivations in the SPE style it is sometimes difficult to figure out what the aim of some rules is. 3/19/2018 65
Conspiracy in Syntax For quite some time, linguists working in the Chomskyan paradigm refused to concede that there might be conspiracy phenomena in syntax The earliest concession was in the domain of the interpretation of pronouns The issue of why a rule applies was not studied systematically before the early nineties! 3/19/2018 66
The Fin Second Conspiracy 1 English: I kissed Mary NEGATION & DO-INSERTION *I not kissed Mary I did not kiss Mary 3/19/2018 67
The Fin Second Conspiracy 2 English: I must not kiss Mary Verb Movement to not have kissed Mary. . . I have not kissed Mary Verb movement & do-insertion must be applied such that the finite verb follows the subject 3/19/2018 68
The Fin Second Conspiracy 3 He believes that John loves Mary * __ was believed that John loves Mary Movement that John loves Mary was believed by many Expletive Insertion it was believed that John loves Mary 3/19/2018 69
The Fin Second Conspiracy 4 He shoots a moose *__ was a moose shot Expletive Insertion there was a moose shot __ Movement a moose was shot __ 3/19/2018 70
Fin Second in English IPs In the core of the English IP, the finite verb must be the second element Verb movement & do-insertion NP-movement & there/it insertion Short adverbs can be ignored 3/19/2018 71
Fin Second in Other Languages Breton Word Order: Preposing of Non-finite verb if nothing appears in focus position Lennet en deus Yann al levr read 3 -have John the book Yann en deus lennet al levr Al levr en deus lennet Yann *En deus lennet Yann al levr 3/19/2018 72
Fin Second in Other Languages Verb preposing/expletive insertion in Icelandic Fram hefur komidh adh fiskadh hefur veridh í leyfisleysi forth has come that fished has been illegally Fram hefur komidh adh thadh hefur veridh fiskadh í leyfisleysi forth has come that it has been fished illegally 3/19/2018 73
Fin Second in Other Languages Verb preposing in Croatian if the clitics would not be in second position otherwise dao mu ga je Ivan given him. it. is Ivan 3/19/2018 74
Fin Second in Other Languages At least six rules conspire Focus Movement Subject Movement Expletive Insertion do-insertion stylistic fronting strange verb preposing 3/19/2018 75
Conspiracies There are no successful accounts of why expletive insertion, stylistic fronting or subject placement apply in the way the do. . . Besides those that make, in one way or the other, reference to the fact that the finite verb must be the second element in the clause! 3/19/2018 76
Summary of conspiracy from phonology‘s perspective: From the three problems we identified with rule ordering, a representational model gives a better solution only to the first one, the unconstrainedness of the rules. The other two problems, duplication and conspiracy are not addressed by representational models. 3/19/2018 77
and from the perspective of syntax (GB-) Syntax became strictly representational, all explanation was done by constraints. . . In this sense, OT syntax is a direct offspring of GB. . . 3/19/2018 78
Brief summary Derivational theories face a number of problems. Grammatical accounts therefore need a representational/constraint-based component. 3/19/2018 79
Representations • Representational models express linguistic generalizations in terms of the results of the derivations. Some structures are possible, some others are not. • Representations just specify which rules and alternations are possible, not whether they apply. • Everything can be generated but unallowed structures are unlicensed and thus eliminated. The notion of filter is crucial. 3/19/2018 80
Representations (nonlinear) • Typical domains of application in phonology are • • Autosegmental : tonal phonology Metrical Phonology: stress Templates: Syllable structure, prosodic phonology Features geometry: allowed and unallowed feature representation, assimilation and dissimilation (OCP) 3/19/2018 81
Syllable templates • Syllable templates specify which kind of syllables are possible in a language. Unsyllabifiable segments are repaired by epenthesis or deletion. • Palestinian Arabic: Syllables are maximally CVC If an input cannot be syllabified straightforwardly, an epenthetic [i] is inserted before the unsyllabifiable consonant. /dars/ /dars + ha/ –> /÷akl/ 3/19/2018 –> da. ris. ha –> ÷a. kil ‘course ‘her course ‘food’ 82
Syllable templates • In PA, epenthesis is a direct consequence of the association of the segments to syllabic positions. The unsyllabifiable consonant has no slot in which to fit in (syllablles are projected from vowels), and, to avoid deletion, the only solution is to create a new syllable by insertion of a vowel. • Other languages choose to delete ‘unsyllabifiable’ consonants. • Notice the derivational component is still needed to account for epenthesis (or deletion). 3/19/2018 83
![Possible assimilations /b/ –> [p] /t/ / Laryngeal | [+voiced] | Laryngeal | [–voiced]](https://present5.com/presentation/237e0bdfba9144e6740d1894f8171dc2/image-84.jpg "Possible assimilations /b/ –> [p] /t/ / Laryngeal | [+voiced] | Laryngeal | [–voiced]")
Possible assimilations /b/ –> [p] /t/ / Laryngeal | [+voiced] | Laryngeal | [–voiced] Possible processes Assimilation of the whole laryngeal node or just the feature [voiced] 3/19/2018 84
![Impossible assimilation /b/ –>? [t] | Laryngeal Supralaryngeal | | | [+voiced] [–voiced] [coronal]](https://present5.com/presentation/237e0bdfba9144e6740d1894f8171dc2/image-85.jpg "Impossible assimilation /b/ –>? [t] | Laryngeal Supralaryngeal | | | [+voiced] [–voiced] [coronal]")
Impossible assimilation /b/ –>? [t] | Laryngeal Supralaryngeal | | | [+voiced] [–voiced] [coronal] Impossible process Assimilation of the feature [coronal] to the laryngeal node: the result ist ill-formed. 3/19/2018 85
Summary of representations • Representations are surface-oriented, since some forms are filtered out by not corresponding to the representational requirements. • In theories which heavily rely on representations, a derivational component is still present, to account for the kind of repair found in different languages. 3/19/2018 86
Problems for representations Here is a list of standard problems for representational accounts of syntax: subjacency effects: movement seems cyclic who do you think that she loves *who do you wonder who loves 3/19/2018 87
Problems for representations Derivational residues in Case theory wem hast du gesagt dass er meint dass er hilft who-dat have you said that he thinks that he helps wen hast du gesagt dass er meint dass er unterstützt who-acc have you said that he thinks that he supports 3/19/2018 88
Problems for representations Derivational residues in binding theory which pictures of herself does he say that Mary saw at the exhibition 3/19/2018 89
Problems for representations Derivational residues in quantifier scope theory UNAMBIGUOUS dass ein Mann jede Frau liebt that some man every woman loves AMBIGUOUS dass jede Frau ein Mann t liebt the sentence still has the scope option of the pre-scrambling structure 3/19/2018 90
Problems for representations Standard solution: Trace theory! Perhaps insufficient (see Chomsky 1993) OT faces a certain problem here - but only if it slavishly follows GBaccounts 3/19/2018 91
ENDE 3/19/2018 92
Скачать презентацию Gisbert Fanselow Caroline Féry University of Potsdam
237e0bdfba9144e6740d1894f8171dc2.ppt