Popular in Course
Popular in Linguistics
This 87 page Class Notes was uploaded by Kenna Muller on Saturday September 19, 2015. The Class Notes belongs to LING607 at University of Delaware taught by Staff in Fall. Since its upload, it has received 153 views. For similar materials see /class/207133/ling607-university-of-delaware in Linguistics at University of Delaware.
Reviews for PhonologyI
Report this Material
What is Karma?
Karma is the currency of StudySoup.
Date Created: 09/19/15
Phonology I Ling 607 University of Delaware February 12 2008 Acknowledgements These notes draw heavily on the notes of Colin Wilson7s Fall 2002 and especially on Kie Zuraw7s Fall 2005 UCLA graduate phonology Classes Although I have made some Changes7 in many oases7 they are word for word the same There are undoubtedly many typos Please email them to me as you nd them Jeff Heinz Feb 127 2008 Intro to Phonology Feb 12 2008 Contents 1 Introduction to Phonology 11 12 13 14 15 Goals of cognitive science Generative phonology 121 Overview 122 123 Course methodology Generalization in phonology 131 Phonotactic generalizations 132 What about knowledge of morphological alternations 133 Substance in phonology Overview of the next few weeks Exercise if we have time Course issues 2 Phonetics 21 22 26 Sounds in Human Languages 211 Articulatory Characteristics 212 Acoustic Characteristics Feature Theory 221 What are features 222 So what are is the feature system we should use 223 Major Glass Features 224 The Sonority Hierarchy Vowels Consonants 241 Dependencies 242 Consonantal features Few comments on feature values Other Phonological Units 261 Syllables 262 Stress 19 19 23 4 CONTENTS 3 Phonological Rules 31 31 List and Rule 31 311 Basics 31 312 More than one rule 32 32 Examples 32 321 English nasal assimilation 32 322 Mohawk 33 323 Sierra Miwok 34 33 Rule Formalizations 34 331 Format 34 332 Left side of arrow 35 333 Right side of arrow 36 334 Redundancy 37 335 Right side of slash context 37 336 Expansion Conventions 37 337 Transformational rules 40 34 Rule Application 41 341 Simultaneous application 41 342 Ordered Rules 41 343 lntrinisic vs Extrinsic rule ordering 42 35 Rule Orderings 43 351 Feeding 43 352 Bleeding 44 353 Transparent vs opaque interactions 45 354 Counterfeeding opacity 46 355 Oounterbleeding opacity 47 356 Noninteraction 48 357 Other issues with rules and orderings 49 36 Phonological Opacity 49 361 Review 49 362 Polish Revisited 50 363 Yawelmani Yokuts 53 364 Mini Typology of Opacity 53 37 Multiple Application Problem 53 371 Overview 53 372 Possible solution I directional application 55 373 Possible solution ll from Anderson 1974 55 374 Minimal vs maximal application 57 38 Intrinsic Ordering Proposals 57 381 Review and Preview 57 382 Some relevant things to keep in mind 58 383 Goal of Koutsoudas et al 58 Intro to Phonology Feb 12 2008 CONTENTS 5 384 Interlude The Elsewhere Condition 61 385 Anderson 1974 and natural orderings 61 39 The Cycle 64 310 Lexical phonology 66 3101 Overview 66 3102 Cyclicity in the lexical component 67 3103 Strict Cycle Condition 67 3104 Bare bones lexical phonology bibliography 69 3105 Strict Cycle Condition revisited 69 3106 lcelandic 70 311 Summary 72 3111 Data 72 3112 Theories 73 3113 Preview 74 4 Representations 77 41 Representing words with features 77 411 How can we decide 78 42 Tonal Association 78 43 Review 80 44 CV and melodic tiers 81 45 Assimilation as spreading 81 46 Geminate lnalterability 82 5 Constraints 87 51 Duplication and Conspiracy Problems 87 511 Dynamic vs static phonology 87 512 Conceptual remarks 87 52 The Duplication Problem 88 521 Example Estonian 88 53 The Conspiracy Problem 89 531 Example Shortening agrammar 89 532 Kisseberth functional unity of rules 90 533 Evidence in Yawelmani Yokuts for conspiracies 90 54 Solution Constraints 91 541 More conspiracies 91 542 A small example of international conspiracy 93 55 Morpheme Structure Constraints7 Surface Phonetic Constraints7 or Both 95 551 Conceptual Arguments for Surface Phonetic Constraints 96 552 Evidence for Surface Phonetic Constraints 97 56 Rule Triggering 99 Intro to Phonology Feb 12 2008 CONTENTS U I 58 U H o 511 512 513 514 517 518 52 H Intro to Phonology A frequently used constraint the OOP Obligatory Contour Principle 100 1 571 Extended example in Japanese 00 572 Discussion Applying Myers7 idea 105 Overview 106 McCarthy 1986 106 Seri and OOP Triggering 107 English 108 5111 More on English 109 Anti antigemination and OOP 109 Overview 110 What is stress 111 5141 What it is 111 5142 Stress as a feature 112 5143 Detecting Stress 112 5144 Examples of simple stress systems from Kager 113 Grid theory 114 5151 Basics 114 5152 Typological Predictions 115 5153 Alternations 115 Alternatives to elaborations of the grid 116 5161 More Parameters 117 Principles and Parameters 118 Extrametricality 119 5181 Moras 120 5182 Cairene Arabic 120 5183 Italian 121 Feet 123 5191 Bedouin Hijazi Arabic 123 5192 Winnebago again 124 5193 Other arguments 126 Arguments for Feet 127 5201 Review 127 5202 Minimality 127 5203 Other arguments for feet 129 5204 Do we need feet 129 Metrical Stress Theory Hayes 1995 130 5211 Quantity insensitive syllabic trochees 131 5212 Quantity sensitive moraic trochees 132 5213 Quantity sensitive uneven iambs 134 5214 Why the asymmetric inventory 134 lambicTrochaic Law 135 5221 lambic lengthening 136 Feb 12 2008 CONTENTS 7 5222 Trochaic Shortening 136 523 Unbounded Stress Patterns 138 524 Sorne Further Reading on Stress 139 525 Transitioning to Optimality Theory 139 6 Optimality Theory 143 61 Optimality Theory 143 611 Overview 143 612 Gen 144 613 Constraints 144 614 Eval 146 62 OT in Practice 147 621 Two types of constraint 147 622 Exposition the tableau 148 63 Review 150 64 The duplication and conspiracy problerns 151 65 Constraint violability 152 66 Extrinsic Rule Ordering 152 67 Multiple sites for rule application 154 68 Directional application 154 69 Repeated application self feeding 155 610 Elsewhere Condition 155 611 Look ahead 156 612 Cyclicity 157 613 Lexical phonology 157 614 Autosegrnentalisrn 157 615 Learnability 158 616 Constraint learning 160 Intro to Phonology Feb 12 2008 8 CONTENTS Intro to Phonology Feb 12 2008 Chapter 1 Introduction to Phonology 11 Goals of cognitive science 1 Characterize the knowledge that humans possess in various cognitive do mains perception categorization language reasoning compe tence theory by a generative grammar I mean simply a system of rules that in some edplicit and well de ned way assigns structural descrip tions to sentences Obviously every speaker of language has mas tered and internalized a generative grammar that eccpresses his knowledge of his language This is not to say that he is aware of the rules of the grammar or even that he can become aware of them or that his statements about his intuitive knowledge of the language are necessarily accurate Similarly a theory of visual perception would attempt to account for what a person ac tually sees and the mechanisms that determine this rather than his statements about what he sees and why though these statements may provide useful in fact compelling evidence for such a theory Chomsky 19658 9 Aspects 2 Discover the methods by which knowledge is acquired including a charac terization of innate restrictons on the form and acquisition of knowledge learning theory A theory of linguistic structure that aims for explanatory adequacy incorporates an account of linguistic universals and it attributes tacit knowledge of these universals to the child It proposes then that the child approaches the data with the presumption that they are drawn from a language of a certain antecedently well de ned type his problem being to determine which of the humanly pos sible languages is that of the community in which he is placed 10 CHAPTER 1 INTRODUCTION TO PHONOLOGY Language learning would be impossible unless this were the case The important question is What are the initial assumptions con cerning the nature of language that the child brings to language learning and how detailed and speci c is the innate schema the general de nition of quotgrammar that gradually becomes more ecc plicit and di erentiated as the child learns the language For the present we cannot come at all close to making a hypothesis about innate schema that is rich detailed and speci c enough to ac count for the fact of language acquisition Consequently the main task of linguistic theory must be to develop an account of linguis tic uniuersals that on the one hand will not be falsi ed by the actual diversity of languages and on the other will be su ciently rich and epplicit to account for the rapidity and uniformity of lan guage learning and the remarkable complepity and range of the generatiue grammars that are the product of language learning Chomsky 1965 Aspects 27 on explanatory adequacy see Chom sky 1965 30 32 34 37 ch 1 etc 3 Understand how knowledge is applied in particular behaviors and what limitations prevent performance from being coextensive with competence performance theory There seems to be little reason to question the traditional view that investigation of performance will proceed only so far as un derstanding of underlying competence permits To my knowl edge the only concrete results that have been achieved and the only clear suggestions that have been put forth concerning the theory of performance outside of phonetics have come from studies ofper formance models that incorporate generatiue grammars of speci c kinds Chomsky 1965 Aspects 10 in ch 12 Toward A Theory of Performance 12 Generative phonology 121 Overview 4 Knowledge of linguistic sound patterns well be mainly talking about this a Representations features feature geometry autosegmental representations natural class metrics maps of articulatory dif culty and perceptual similarity b Generalizations Intro to Phonology Feb 12 2008 12 GENERATIVE PHONOLOGY 11 linguistically signi cant generalization rules rule ordering rule iter ation cyclic application constraints constraint con ict status of ex ceptions 5 Learning sound patterns a b Partial theories with signi cant innate cornponents eg Ellison 1992 1994b Gildea and Jurafsky 1996 Tesar and Srnolensky 2000 Al bright and Hayes 2003 Hayes and Wilson 2008 Heinz 2007 Largely negative or simply absent results concerning the computa tional feasability of the theories that have been proposed eg Eisner 2000 ldsardi 2006 6 Phonological performance 7 considered mostly as evidence for theories a b d Traditional sources descriptive grarnrnars eldwork phonetic experi ments also poetry verse etc Psycholinguistics wug7 testing and rating novel forms for rnorphologi cal andor phonological well forrnedness eg Albright and Hayes 2003 Zuraw 2000 induced speech errors eg Goldrick 2002 Child phonology increasingly detailed analyses largely in response to development of computational learning theories eg Tessier 2006 Brain imaging studies have investigated localization of natural class phonological feature representations eg Phillips et al 1995 Relatively little work in the opposite direction 7 incorporating phono logical cornpetence into theories of speech productionperception word recognition and sentence parsing 122 Course issues 7 Course issues a Intro to Phonology What methods can be used to de ne linguistically signi cant general ization And how does generalization7 in phonology relate to gener alization7 in psychology In what form are individual phonological generalizations internalized by the native speaker rules rnorpherne structure conditions constraints lters protototypes How do generalizations about phonological structure interact with each other rule ordering vs constraint ranking vs rnixed ruleconstraint theories cyclic interaction What innate restrictions or biases are imposed on phonological gener alizations And how do these biases relate to the phonetic properties of sounds and sound sequences rnarkedness naturalness formal uni versals Feb 12 2008 12 CHAPTER 1 INTRODUCTION TO PHONOLOGY e What role do representations play in restricting or classifying phonolog ical generalizations if the representation is right then the rules will follow7 McCarthy 1988 privativity underspeci cation feature geome try linking condition 123 Course methodology 8 Course methodology aka instructor7s biases a Develop awarenessknowledge of the empirical phenomena b Simpli cation and approximation i Develop reasonably simple methods of analysis eg small set of possible rules ii Main goal is understanding the techniques and their failings not simulation of speakers Focus rst on basic cases then on more dif cult problems with the aim of approximating the important qualitative aspects of a phenomenon iii V Although the ability of a connectionist network or other compu tational device to reproduce certain aspects of human performance is interesting and impressive this ability alone does not qualify the net work as a theory and does not amount to epplaining the performance McCloskey Michael 1991 Networks and Theories The Place of Con nectionism in Cognitive Science Psychological Science 26 388 the term numerical simulation makes many of us uncomfortable It is easy to build models on computers and watch what they do but it is often unjusti ed to claim that weve learned anything from such epercises The fact that some computer code produces a pattern that looks like a snow ake or aflower or an earthquake doesnt necessarily tell us that the ingredients of the code have anything to do with the natural phenomenon it seems to depict Langer James S 1999 Computing in Physics Are We Taking It Too Seriously Or Not Seriously Enough Physics Today July 1999 p12 It can be argued that getting a machine to perform intelligently is more important than understanding how it does so If a magic procedureisay for learningidid in fact lead to the level of performance desired de spite our inability to understand the resulting computation that would of course be a landmark accomplishment But to eppect this kind of breakthrough is naive We now have enough disappointing epperience Intro to Phonology Feb 12 2008 13 GENERALIZATION IN PHONOLOGY to eppect that any particular insight is going to take us a very small fraction of the way to the kind of truly intelligent mechanisms we seek The only way to reasonably eppect to make progress is by chaining to gether many such small steps And the only way to chain together these steps is to understand at the end of each one where we are how we got there and why we got no further so we can make an informed guess as to how to take the nept small step A magic step is apt to be a last step it is ne as long as it takes you emactly where you want to go emphasis original Smolensky Paul 1986 Information Processing in Dynamical Sys tems Foundations of Harmony Theory In David E Rumelhart James L McClelland and the PDP Research Group Parallel Distributed Pro cessing Vol 1 Foundations Cambridge Mass MIT Press p220 9 Theories as composite entities a Every serious theory of phonology has many many components i SPE rule notation ordering morpheme structure rules linking exception features ii OT richness of the base universal constraints strict domination factorial typology b Goal of theory evaluation is therefore very rarely wholesale endorsement or rejection c Focus instead on separating chaff from wheation identifying the com ponents that lead to speci c incorrect predictions and on replacing them with better alternatives lmpossible without rst understanding the individual components and their interactions 10 Suspension of commitment and disbelief a Do not commit fully to any particular theory nor reject theories on general grounds Focus instead on working out the speci c predictions of any given theory looking for new connections among facts as well as internal contradictions and falsi cations b 13 Generalization in phonology 131 Phonotactic generalizations 11 Knowledge about the sounds and sound sequences that are permitted in surface forms a English All syllable or word nal NC clusters are homorganic Intro to Phonology Feb 12 2008 CHAPTER 1 INTRODUCTION TO PHONOLOGY 12 a 132 13 a b 14 a b Intro to Phonology English All morpheme internal obstruent clusters are voiceless cf adze7 rugby7 asbestos Russian Voiced affricates eg and voiced velar fricative occur only as the result of voicing assimilation Halle 1959 Malayalam Non geminate velar nasal occurs only as the result of place assimilation Evidence for knowledge of phonotactics Perception of phonetically ambiguous stimuli Pitt 19987 PampP7 More ton 20027 Cognition7 word spotting in continuous speech Suomi7 Mc Queen amp Cutler 19977 JML7 McQueen 19987 JML7 McQueen7 Otake amp Cutler 20017 JML7 well formedness ratings of novel words Pierrehumbert 19947 LabPhon llL Coleman amp Pierrehumbert 19977 ACL7 P risch7 Large amp Pisoni 20007 JML Frisch amp Zawaydeh 20017 Language memory for novel words P risch7 Large amp Pisoni 20007 JML7 naming latency of existing and novel words Vitevitch 20027 JEP See also studies on experimental phonotactics transitional proba bilities Dell7 Reed7 Meyer amp Adams 20007 JEP7 Goldrick 20027 others What about knowledge of morphological alternations Some relevant work Albright amp Hayes 2002 comparison of rule and exemplar based models Wilson 2006 testing for biases in acquisition of experimental alter nations See also Finley 2007 Brief review of English past tense Regular weak verbs d7 t7 ad conditioned by the nal con sonant of the stem lrregular strong verbs Various degrees of irregularity irregulars with same change often bear family resemblance7 to one another see Pinker amp Prince 19897 p185 Feb 12 2008 13 GENERALIZATION IN PHONOLOGY 15 15 16 17 Intro to Phonology C a No change hit slit bid rid let cut put burst hurt b Laxing bleed breed feed meet hide slide bite shoot c t ending burn 7 bend spend lose deal feel mean creep leave d d ending ee say hear e isw 000 freeze speak steal swear tear wear bear f Iwae ring sing spring drink shrink sink stink swim begin g INA cling ing sling sting string swing wring stick dig h orqu blow grow know throw draw y i GINO take shake j aINOU rise write ride drive dive k Suppletion go buy Previous theories of speakers knowledge of the past tense a Halle amp Mohanan 1985 Ll multiple rules assignment of verbs to lexical strata b Rumelhart amp McClelland 1986 PDP connectionist network with no rules See also MacWhinney amp Leinbach 1991 Daugherty amp Seidenberg 1994 others c Pinker amp Prince 1989 single regular rule associative network for irregulars Theories evaluated by Albright amp Hayes a AampH theory Multiple detailed rules that describe surface changes for both regular and irregular verbs each rule interpreted stochastically weighted by success on the lexicon hitsdomain b a modi ed version ofthe General Context Model GCM Nosofsky 1990 etc i Components of the GCM Training phase exemplars of the categories are stored in memory Testing phase stimuli are categorized by computing their similarity to stored exemplars ii Past tense of a verb calculated by summing the similarity of the verb to all existing verbs that undergo a particular change Similarity metric alignment matching provided by AampH Learning rules in the AampH theory based on intuition of Pinker amp Prince 1989 a Begin with maximally speci c rules eg Q a d ainpast tr b Construct more general rules by minimal generalization compare stems that undergo the same change aligning them at the point where Feb 12 2008 16 CHAPTER 1 INTRODUCTION TO PHONOLOGY the change is made 93 w d l ammalast Q a d knfa1npast Q a d X soncont voicea1nHum c General rules do not replace speci c rules all are available to form past tenses 18 Novel ndings of AampH a Well known that irregular verbs fall into family resemblance classes Bybee amp Slobin 1982 Language Pinker amp Prince 1988 Connections and Symbols Ex Many verbs that end in 113 form their past tense with the 1gtmst change b AampHs learner discovered the same type of structure in the set of regular verbs Ex All verbs that end in voiceless fricatives f6s are regular c AampHs experimental results generation of past tenses and ratings sug gest that speakers are sensitive to such sub regularities for both regular and irregular changes 19 Comparison with the GCM a Overall the AampH theory ts the data somewhat better than the GCM b The AampH theory places an innate restriction on the set of rules that can be learned i Only context local to the change can be used to generalize a rule ii In other words internal variables are not allowed eg Q a d an is impossible c But the GCM computes the global similarity of a verb to all existing verbs i The model is not biased to use contextual information local to the change ii Therefore it draws bizarre analogies eg rsndarad based on similarity of render7 to rend end vend mid fend mend tend round dread 20 Conclusions a Correct morpho phonological generalization depends on structured similarity Concepts such as locality natural classes etc provide the structure for generalization Many ofthese concepts were fundamental concerns of early generative phonology b Several dimensions on which traditional psychological and phonologi cal theories differ Most interesting proposals such as AampHs theory set universal restrictions on generalization within a framework that is Intro to Phonology Feb 12 2008 14 OVERVIEW OF THE NEXT FEW WEEKS 17 more inductivist than traditional generative phonology 133 Substance in phonology 21 22 23 24 Intro to Phonology The substance question What role do substantive relations among phonological entitiesirelations such as similarity7 enhancementantagonism7 and ease of transitioniplay in phonological systems Synchronic grounding Substantive relations place restrictions andor biases on the generalizations that are acquired by language learners See SPE ch 97 Natural Phonology Stampe 19797 phonetically driven phonology eg7 Steriade 2001720027 Hayes Kirchner7 and Steriade eds 2004 and Wilson 2006 for related proposals Diachronic grounding Substantive relations play a role in determining likely phonological changes7 but do not restrict or bias synchonic generalizations See Anderson 19817 Ll7 Hyman 19987 Blevins amp Garrett 19987 Language7 Buckley 20007 Blevins 2004 for related proposals General conclusions a Determining how innate restrictions7 or analytic biases7 structure the generalizations that are induced from data is arguably the central prob lem facing cognitive science b Phonological theory provides a framework of hypothesized representa tions and analytic biases7 both closely related to phonetic substance7 within which to address this problem eg Wilson7s 2006 substantive bias hypothesis Overview of the next few weeks Writing phonological rules7 notation7 principles of analysis7 etc Dimensions of complexity a Unnaturalness No synchronic substantive relation between a property of Oz and the context in which it is conditioned random rules77 b Non locality The conditioning element is not strictly local to a The distance be tween the two may be unbounded7 possibly with restrictions on inter veners Feb 12 2008 18 CHAPTER 1 INTRODUCTION TO PHONOLOGY c Morphologicalrestrictions Conditioning applies only when 04 and the relevant element of the context stand in a certain morphological relation with one another d Opacity The conditioning element is not present in the surface representation although it may also condition effects on elements other than 04 e lnteraction Another generalization prevents 04 from having a property that is con ditioned by an element in its environment 15 Exercise if we have time Kasem Class C noun plurals Based on discussion in SPE Ch 87 data originally from Callow7 plus additional discussion in Halle 1978 singular plural bakada bakadi boy sada sadi grass mat mimina mimini thin fala fali white man tula tuli granary kukuda kukudi dog fana fani knife cana cani moon bakala bakali shoulder kambia kambi cooking pot pia p1 yam buga bwi river di room malaa male chameleon kabaa kabe slave Zizaa Zize grass roof laija le song naga ne leg pia pe sheep cf yam babia babe brave nanj ua nanj we y yua ywe hair koga kwe back 6013a cwe path gtlt ltcdmwneodBwmermqshmouocrm 9 on 33 Intro to Phonology Feb 12 2008 Chapter 2 Phonetics 21 211 Sounds in Human Languages Articulatory Characteristics 212 Acoustic Characteristics 22 Feature Theory 1 Feature Theory is the theory whose subject is subsegmental structure Major issues include a b c d e f g How are features de ned What are the relevant features What values can features take Are there dependency relations between features and if so what are they What is the role of feature underspeci cation Where do features come from How do we nd answers to these questions 2 Goals of Distinctive Feature Theory a Determine which units can be minimally contrastive in the worlds lan guages i The idea is each language draws its distinctive features from this universal set Another idea is that the features bridge the gap between the abstract phonemic symbol and the physical realization of the soundithey are concrete instructions to articulators7 for example 3 Of course there is always a tension 19 20 CHAPTER 2 PHONETICS a The more features posited7 the more phonetic detail is accounted for7 and potentially more phonological processes can be explained with ap peal to phonetic naturalness b However7 smaller inventories are simpler and make describing phono logical processes easier T a If two sounds contrast minimally in some way in some language7 then a feature must be posited to account for the difference b If a phonological process applies to some class of sounds then those sounds c The more common the distinction is used across languages7 the more languages which refer to the same class of sounds7 is evidence for that feature andor natural class d Consequently7 the phonemic inventories and phonological processes found in the worlds languages constitute the major sources of evidence for distinctive feature theory e lncreasingly detailed phonetic work providing articulatory via ultra sound or other more invasive means or acoustic measures can help clarify which speech sounds meet which de nitions U D methodology 221 What are features 5 6 Intro to Phonology The central idea behind features is that each segment is actually de nable in terms of a set of properties These properties are the features and this set is often called a feature bundle Thus words7 which we often think of as a string of segments7 can actually be represented as a string of feature bundles Here is one set of features we might use to illustrate this idea with the word mop syllabz39c isyllabz39c isyllabz39c 50n0rant 50n0rant isomorant istop st0p st0p mop inasal nasal l0w inasal LABIAL back LABIAL vozce ivmce fround The features can be thought of as members of a set of elementary categories which combine to form the speech sounds of language Where do the features come from a It is standardly assumed that they are universal and innate which Feb 12 2008 22 FEATURE THEORY 21 9 10 222 11 12 13 223 14 15 Intro to Phonology would explain why the same distinctions and processes are found in so many unrelated languages However it has recently been argued that features are emergent ie learned and are not innate Mielke 2004 In principle features can be de ned articulatorily or acoustically more on this below This is another area of debate Phonological rules appear to apply to classes of sounds which follows nat urally from de ning a segment as complexes of properties It has already been pointed out that the feature system pro vides us with a means not only for designating individual speech sounds but also for designating particular classes ofspeech sounds p 185 Halle 1972 So What are is the feature system we should use There have been three major proposals a Preliminaries to Speech Analysis Jakobson et al 1952 b SPE Chomsky and Halle 1968 c Uni ed Feature Theory Clements and Hume 1995 Note that SPE and Preliminaries use the linear feature model outlined above however more recent feature theories employ a non linear or 3 dimensional approach to features called feature geometry we may spend time on this at the end of the semester Among other things feature geometry encodes dependencies between features a Reasons to think features ought to be de ned articulatorily a The rules that we propose refer to natural classes which are best de ned with articulatory features as opposed to acoustic ones b Some acoustic gestures have unclear acoustic correlates c Some acoustic gestures have different acoustic correlates depending on their environment d But as we will see the issues are more complex than this Major Class Features AcknowledgementThis section is adpopted from a 2003 draft of an intro ductory text to phonology written by Bruce Hayes We have been introduced to the following terms as designating certain groups ofsimilar sounds vowels glides liquids nasals fricatives affricates Feb 12 2008 22 16 17 and stops CHAPTER 2 PHONETICS Bruce Hayes de nes these features in terms of sonority7 an acoustic measure7 which roughly corresponds to acoustic loudness a b Note to be 1 Although There is no clear single acoustic correllate of sonority that further distinctions can be made le lower vowels are said more sonorous than higher vowels Others de ne these sonority based features in terms of STRIC TURE7 ie how much air ow is allowed to pass Different artic ulations constrict the ow of airs in different ways contributing to different degrees of sonority we could use a feature system with features like Hstop7 this misses the point that these different groups of sounds pattern together in many rules in natural language 1quot Can you give an example from any of your HW assignments Hint think of 18 19 Intro to Phonology lndoensian We can reconstruct the categories above with the following features vowel glide liquid nasal fricative affricate stop syllabic but see below syllabic7 consonantal consonantal7 approximant approximant7 sonorant sonorant7 continuant continuant7 delayed release continuant7 delayed release l l il l l l l we add delayed release to Syllabic fricatives and stops are quite rare they occur in Berber languages7 but can be de ned analogously We remarked earlier that English has syllabic nasals and consonants Feb 12 2008 distinguish affricates from stops 22 FEATURE THEORY 23 20 224 21 22 23 24 25 26 Thus strictly speaking syllabic is outside of the system given above since all of the categories in the set glide liquid nasal fricative stop can be syllabic a However in the majority of the worlds languages only vowels can be syllabic segments and in these languages it is reasonable to take syllabic as de ning the extreme sonorous end of the hierarchy The Sonority Hierarchy Acknowledgement This section is taken somewhat more directly from a 2003 draft of an introductory text Chapter 4 to phonology written by Bruce Hayes This feature system with cut off points along a continuum is a way of formally characterizing the so called sonority hierarchy which is precisely the continuum shown in 15 The features corresponds to cut off points in the continuum a Thus consonantal means at least as sonorous as a glide b approximant means at least as sonorous as an approximant c sonorant means at least as sonorous as a nasal d continuant means at least as sonorous as a fricative One major source of evidence for the above feature system comes from examining which segments may be syllabic peaks Typologically in many languages syllabic sounds must be vowels But there are some languages like Czech and Serbo Croatian in which syllabic sounds may be vowels or liquids but no other sound English allows vow els liquids and nasals to be syllabic but no other sound Some Berber languages allow absolutely any sound to occur as a syllabic nucleus This is called an implicational hierarchy if in a given language a segment of a certain type can be a syllabic nucleus then all segments of greater sonority may also serve as a syllabic nucleus 1quot How do the proposed features capture this implicational hierarchy 27 Intro to Phonology Another application of the Sonority Hierarchy involves sonority sequencing within the syllable In a language which permits clusters of consonants to occur at the margins of syllables typically the situation will be this the most sonorous segment in the syllable serves as its nucleus Then Feb 12 2008 24 CHAPTER 2 PHONETICS proceeding outwards from the nucleus towards the edges of the syllable7 one tends to encounter segments of progressively decreasing sonority What are some counter exarnple to the Sonority Sequencing Principle in En glish 23 Vowels Go to Bruce7s notes on vowels 24 Consonants 241 Dependencies 28 Place of articulation LABIAL CORONAL DORSAL round anterior high labiodental distributed low strident front lateral back 29 The fact that certain features depend on values of other features ie labiodental iff labiall suggests that the features might be organized hierarchally Place LABIAL CORONAL DORSAL round labiodental high low from back anterior distributed strident lateral Intro to Phonology Feb 12 2008 2 4 CONSONANTS 25 so Then phonologlcal pxocesses f0 example ale sad to apply to nodes as opposed to lndwldual feacules p T 7 5 place i o place rcontmuant he noLaLlon In che zule above ls a shonhancl of saylng all feacules clonnnalecl by che place node change to macch che feacule values of che followmg slop oz a ncace o m 242 Consonantal features 31 Lawgeneal feacules a Volce some bleak chls up to le vocal colds and slack vocal cozdsD b consumed glows gloccahzecl sounds ejectlves c spxead gloms aspnacecl sounds 32 CORONAL a anlenol sounds chal ale ploduced mole fzonL of che alveolax udge dlsLubuLed mole conlacl of Longlue blade lbllants nolsy glooved longue acelal baslcally dlsLlngulshes 1 fzom olhel Colonels Data fzom Khan 2006 see also zefezences Lhelem mu as LABIAL a zound b lablodencal 34 DORSAL a ack Intm tn Phnnnlngj Rab 12 2008 26 CHAPTER 2 PHONETICS b front 0 high d low Look at Bruce7s notes on subdivisions of LABlAL and DORSAL 35 Other features a trill7 ap b ejective c irnplosive 25 Few comments on feature values 36 Most features are binary valued ie 37 Some are considered to be rnonovalued7 typically nasal The idea is here is that non nasals do not seem to form a natural class in any language 38 Other features which are dependent on other features eg the features distributed which is dependent on the CORONALD may have a zero value in non coronal segments a Forrnally7 0 values are nondistinct from both values b It can be interpreted as phonetic indifference le the value of dis tributed will be determined by other factors lled in77 by context for example Can you think of a domain where it might be natural to have more than two values for a feature 26 Other Phonological Units 261 Syllables 39 Although we havent discussed syllables rnuch7 plenty of evidence has been cited in favor of their existence phonological processes are best stated in terms of the syllable dornain7 or edge7 some language garnes target syllables7 and it is generally claimed that native speakers have clear intuitions about how many syllables are in a word and See Blevins 1995 for more 40 ls syllabi cation of a string something that can be constrastive in a lan guage Intro to Phonology Feb 12 2008 26 OTHER PHONOLOGICAL UNITS 27 a Eg could a language contrast mistik from mistik7 b Generally7 it has been observed that syllabi action is predictable ie does not need to be speci ed underlyingly7 and a very simple algo rithm serves to syllabify a string 41 Syllabi cation Algorithm rst approximation a Scanning from left to right identify the vowels in the word7 and project a syllable node from them b For each syllable node7 link to a consonant to the left of the vowel if there is one c Then7 for each syllable node7 link to a consonant to the right of the vowel if there is one d Unlinked segments are then either linked to existing syllable nodes7 respecting the Sonority Sequencing Principle wherever possible7 and the process is repeated with other sonorous sounds like glides and liquids now serving to project syllable nuclei Intro to Phonology Feb 12 2008 28 CHAPTER 2 PHONETICS 1k Let7s apply the algorithm to nostradamAs 42 43 44 262 45 46 47 48 49 50 Intro to Phonology The segments which project the syllable node are called the nucleus The prenucleic elements make up what is called the onset7 and the post nucleic elements make up what is called the coda Note there arent any known phonetic corellates to syllables7 or their con stituent parts Note also that under a theory of extrinsic rule ordering7 we may expect syllab cation to be ordered with respect to rules Stress The rhythmic properties of words have been well studied over the past thirty years beginning with Lieberman 1975 Also the typology of stress has also been well studied Hyman 19777 Hayes 19817 Prince 19837 Halle and Vergnaud 19877 ldsardi 19927 Bailey 19957 Hayes 19957 Hyde 20027 Gordon 2002 We have described stress so far as a property of the vowel hence we have used a feature like stress However7 most people consider stress as a property of the syllable7 and not of a segment For example7 no language seems to make a distinction between tautosyllabic VV and tautosyllabic VV Tautosyllabic means in the same syllable77 As with syllabi cation7 we can ask whether languages allow words to con strast with only a difference in stress7 eg tAntAn vs tAn tAn However7 for now let us consider some of the known factors in languages in which stress is predictable One factor is called syllable weight Sometimes in order to state the stress rule7 it must make reference to how heavy a syllable is a In the same way that we could think of sonoruty as a property of segments7 we can also think of it as a property of syllables Thus we have a scale b CV lt CVN lt CVC lt CVCC lt CVVC c Again we could make ner distinctions syllables with low vowels in their nuceli are more sonorous than syllables with high vowels in their nuclei d Some languages apply a cut off point to distinguish two kinds of syl lables heavy and light Some make a third distinction superheavy Feb 12 2008 26 OTHER PHONOLOGICAL UNITS 29 51 Consider the classic case of Latin Jacobs 19897 Mester 19927 Hayes 1995 a Light syllables are those with a short vowel and no coda consonant7 ie CV Heavy syllables are all other syllable types in Latin7 eg CV7 CVC7 CVCC The stress rule is now given below b In words at least three syllables in length7 stress the penult if it is heavy7 otherwise stress the antepenult ln shorter words7 stress the initial syllable c a arnixkus L H friend kind7 b gubernaxbunt L H H they will reign7 c inirnixkitia L L H L L L hostility7 d dornestikus L L H belonging to the house7 e mandax H entrust 2sgirnp7 f kams L H dogj g h ri L yesterday7 d Also note that the examples e f show that stress falls initially in disyllabic words 52 Here we havent written the rule forrnally7 and we postpone that discussion for later in the semester when we discuss rnetrical stress theory Intro to Phonology Feb 12 2008 30 CHAPTER 2 PH ONETI CS Intro to Phonology Feb 12 2008 Chapter 3 Phonological Rules 31 311 1 it List and Rule Basics When we analyze the phonetic forms of words in a language sometimes it is possible to nd certain regularities In such cases it is often possible to predict some aspect of the phonetic form in particular contexts a Information about the phonetic form that is not predictable is idiosyn cratic and must be memorized along with linguistic information like meaning in the lexicon ie in a list of forms and their meanings b Information that is predictable can be described with a rule Why is it preferable to store predictable information about phonetic forms in a rule rather than in the lexicon Two different phonetic sounds are contrastive if there exists a minimal pair in the language eg English baI pal The fact that one word begins with b and the other with p is information that must be stored in the lexicon Eg the underlying phonemic representations are ba1 pa1 Phonemes can have different phonetic realizationsithese are called allo phones Sometimes these are predictable but other times they may occur in free variation Similarly morphemes may have more than one surface realizationithese are called allomorphs 32 CHAPTER 3 PHONOLOGICAL RULES 312 More than one rule 5 Typically more than one rule is necessary to adequately account for some data Questions arise a How do we apply multiple rules In some order or all at once 6 Similarly we may nd that two sets of possibly ordered rules account equally well for the same set of data How do we decide which one is the right one a Think about what the rules predict in forms that are not present in the data Where the two accounts make different predictions7 get data le solicit from a native speaker7 consult dictionaries7 do an experiment7 etc b Even so7 it is possible to have different analyses that make the same predictions everywhere In this case7 we need an evaluation metric to decide between competing7 equivalent analyses Generally7 go with the simpler set of rules 7 What does it mean to be simple a The evaluation metric suggested in SPE is brevity the proposed gram mar with the fewest number of symbols 8 How do we write a rule l Go to Rules and Notation Handoutl 32 Examples inkonceivable7 i13konce inglorious7 iijglorious 321 English nasal assimilation 9 Nasal place assimilation in English in 7 Adj a Adj7 with basically negation semantics a nallterable impossible intolerable inadvisable 7 indecent 1n8dible imeasurable inumerable in8xpensive infertile ins atiable ilnolthentic inviolable 7 b Other examples inoffensive7 inapt7 impalpable7 impartial7 impassive7 impatient7 impenetrable7 impulsive7 intangible7 intemperate7 intransi gent7 inde nable7 indescribable7 indiscriminate7 indistinguishable7 in conspicuous7 incompressible7 immoderate7 informal7 infrequent7 insolu ble7 insurmountable7 invalid see also ilnhlospitable7 irational7 ilrlelevant illlegal7 ilogical 10 Nasal place assimilation rule English Intro to Phonology Feb 12 2008 ivable 32 EXAMPLES 33 Con8 cons nasal aplace 7 icont aplace Condition Nasal is part of the rnorpherne in Can you think of other English rnorphernes that undergo the same alterna tion 11 Autosegrnentalgestural interlude assimilation as regressive backward7 extension of a place autosegrnent or gesture cons icont cons nasal I place 12 Questions left unanswered a Why does the nasal assirnilate to the following stop7 and not vice versa b Why does the nasal assirnilate to stops7 but not to fricatives c Why does every place participate in the alternation d How is place assirnilation related to other alternations eg illlegal 13 Place Features a Major divisions labial7 coronal7 dorsal b Subclassi cation of coronals front anterior anterior anterior anterior distributed distributed distributed distributed interdental dentalalveolar77 alveo palatal77 retro ex ley l ltydysyzynl 07371737111 t CL a ml 322 Mohawk 14 When two surface sounds A and B are in complementary distribution7 it means A only occurs in environments where B does not occur in and vice versa 15 Consider the distribution of long and short vowels in Mohawk Intro to Phonology Feb 12 2008 34 CHAPTER 3 PHONOLOGICAL RULES a wisk ve7 rayfxthos he plants7 y kreks 1 push it7 rak xtas he scrapes7 rehyaxraTa he remembers7 raxkAs he sees her7 k xsaks 1 look for it7 roy i te he works7 irraks he eats it7 nikanuhzakeh houses7 wahoy i dA he worked7 ranuzweTS he likes it7 b Source Kenstowicz amp Kisseberth 197943 based on Postal 1968 Describe the distribution of long and short vowels Are long and short vowels contrastive in Mohawk lf not7 determine the underlying forms and write a rule which explains the positions of the different vowels 323 Sierra Miwok 16 Placement of stress in Sierra Miwok a haxna axmaji kawaxtji jaxjaxli wataksa hanna kala ijpax palattata wittapi huffezpi imtejj a patkayi b Source Kenstowicz amp Kisseberth 197943 based on Freeland 1951 ls stress predictable in Sierra Miwok Examine the distribution of stressed and unstressed vowels Write the stress assigning rule 33 Rule Formalizations 331 Format A a B X Y 17 This means XAY is rewritten as XBY or7 in other terms7 A is rewritten as B when preceded by X and followed by Y syl 18 Example flow a High 7CC Intro to Phonology Feb 12 2008 33 RULE FORMALIZATIONS 35 19 A is the affected segment focus or target of the rule 13 is the structural change that the rule requires L Y is the context of the rule XAY is the structural description 20 If X and Y are null7 then the rule is said to be context free7 Otherwise7 it is context sensitive7 992C793 21 A rule applies to a form ifthe form contains a string nondistinct from XAY nondistinct7 explained precisely below 332 Left side of arrow 22 A can be a feature matrix sometimes called a feature bundle or Q a If A is a feature matrix7 like J 7 then the rule looks for any segment that is nondistmct from that matrix b The segments that a feature matrix picks out7 are a natural class 23 Two feature matrices are distinct iff there is some feature F whose value is different in the two matrices 1quot Which of the following are distinct from iii 7 syl isyl flow Tmmd 7 flow 7 flow back fround hzgh 24 Note this means that if A doesnt mention some feature F7 it doesn7t care77 about itithe rule accept segments that are either F or F 25 Sometimes7 if A is meant to pick out a single sound7 we use an lPA symbol instead u a lmy1 C a This is a good idea for readability7 but keep in mind that7 in order to determine how long the rule is for purposes of applying a length based evaluation metric7 you have to expand the IPA symbol into a feature matrix7 namely the smallest feature matrix that picks out just that sound from the languages sound inventory 1quot What does the u above abbreviate to if the languages vowel inventory is i7 17 u If its i7 17 u7 0 If its i7 17 17 u7 0 Intro to Phonology Feb 127 2008 36 CHAPTER 3 PHONOLOGICAL RULES 26 Sometimes we also use C to abbreviate syl or or V to abbreviate syl Again7 this is good for readability Be careful when reading7 though7 be cause some authors7 following SPE7 use C and V to abbreviate voc7cons and voc7 cons 27 If A is 07 youve got an insertion rule the idea is that insertion changes nothing into something Q a i C C Why dont we use the empty matrix instead of Q 28 Note A is at most a single segment If you need a rule which affects more than one segment at one time7 chek out transformational rules in 337 333 Right side of arrow 29 B also can be a feature matrix or Q 30 If B is a feature matrix7 then any of the affected segment7s features that are mentioned in B are changed to the value given in B All other features are left alorze 1 What does a high do to o to u 31 If B is 07 then the segment that A matches is deleted a C Q C b why not 32 Again7 we sometimes use an lPA symbol as an abbreviation for all the feature changes necessary to change anything that could match A into that lPA symbol a i silly rule What does the i above abbreviate if the languages vowel inventory is i a u If its i a u o If its i y a u o 33 If A is 07 then the IPA symbol abbreviates the features needed to pick it out of the languages phoneme inventory Q a i C C Intro to Phonology Feb 12 2008 33 RULE FORMALIZATI ON S 37 334 Redundancy 34 The principle that shorter rules are preferred over longer rules means that unnecessary features should be eliminated from A and B 1quot What is wrong with each of the following rules 335 35 36 37 38 39 336 40 41 Intro to Phonology 1 Syl a r0und fround nas Woke antemor assume the phoneme inventory of English Right side of slash context X and Y are strings made up of ofeature matrices othe boundary types and treated in SPE like segments oat their outside edges7 category boundaries Feature matrices in X and Y match segments in the same way that A does ie7 they match a segment if not distinct from it Boundaries7 word boundary and morpheme boundary7 are treated in SPE as feature matrices7 but we wont get into that here Category boundaries like lem and Verbl can also be used7 but only at the edges of X Y If both edges have brackets7 the labels must match More recently it has become common to indicate left and right morpheme and word boundaries with category boundary like notation Eg lwd indi cates a right word edge boundary Expansion Conventions Devices like parentheses7 curly brackets7 and angle brackets are used to collapse related rules into a single rule schema whose length is shorter cost is lower Rather than adjusting the de nition of nondistinctness7 SPE gives expan sion conventions to turn those schemata into sets of rules that can then be applied using the simple de nition of nondistinctness Feb 12 2008 38 CHAPTER 3 PHONOLOGICAL RULES Lowercase Greek letters 42 Variables that stand for 7 7 or whatever values the theory says some feature can take could be 1723 for some featuresican you think of any good candidates C a avoice avoice avoice expands into C a voice voice voice C a voice voice voice Parentheses 43 Parenthesis are used to indicate optionality 44 For exarnple7 the rule schema V a Q VC is expanded into these two rules in that order v a Q 7 v0 V S Q 7 C Do you ever need parenthesis in a feature matrix 45 The rules that a schema expands into are disjunctively ordered That rneans7 inforrnally7 that you try to apply the rst one if its structural description is rnet7 you apply that rst rule and dont try any of the other rules from the same scherna lf not7 move on to the next rule and proceed in the same fashion In other words7 you never apply two rules of the same scherna to a single word How does the rule above apply to bauk7 46 This is a bit too crude7 because it doesn7t give the right result for cases where different rules of a schema apply to different parts of a wordiin those cases7 we want rnultiple rules of the schema to apply to the same word7 just in different places Braces curly brackets 47 Braces are used to indicate rnultiple possibilities Intro to Phonology Feb 12 2008 33 RULE FORMALIZATIONS 39 48 For example7 the rule schema Z a Q V is expanded into these two rules in this order i 6 Q V 0 0 V Can you imagine a way to translate parentheses into braces Try it with V w VC Super and subscripts 49 50 51 Used when you want to specify a sequence of common elements Eg Cn means n or more Cs77 most common is CO C 60 700 expands to o w CCCC o w CCC o m CC 360 C 080 73 The tricky thing about this is that we apply the longest rule whose struc tural description matches How would the schema above apply to tabskt7 Parentheses with star 52 53 54 Intro to Phonology means that the material in parentheses can occur zero or more times For example7 V a stress CVCVC expands to V a stress C V a stress CVCVC V a stress CVCVCVCVC 7 With 7 disjunctive ordering does not apply Every version of the rule that can apply does applyithey apply simultaneously Feb 12 2008 40 CHAPTER 3 PHONOLOGICAL RULES 1quot How would the stress rule above apply to badupidome How would C a Q 7 C apply to tabskt7 Angled brackets 55 Like parentheses7 but when the optional information is in more than one place A schema with angle brackets expands into two rules the rule with the information in the angle brackets and the rule without that information 56 For example7 C a Q VC 7 CV silly example7 I know expands to C a Q VC CV C a Q V V Expand the following schema and apply it to putod luged and fesz39l Syl syl ltbackgt a high C back C ihz39gh 57 SPE proposes a nifty notation to take care of both parentheses and angled brackets and allow more than one pair of angled brackets in a rule7 but it doesn7t appear used often7 so we wont learn itisee pp 394 395 of SPE 337 Transformational rules 58 Transformational rules are used when more than one segment is a eeted at a time Typical phenomenon which require transformational rules are a Metathesis Metatheis occurs when segments switch positions eg ask a aks b Coalesence Coalesence occurs when two segments combine into one7 typically preserving some features of each eg mail a me 59 Typically these rules are of the form A a B7 where A and B are strings of equal length and each feature matrix is indexed vov NW 60 Example CV Metathesis 1 2 3 4 a 1 2 4 3 61 Coalesence Intro to Phonology Feb 12 2008 34 RULE APPLICATION 41 ihz39gh flow ihz39gh ihz39gh back flow flow iATR back back l0ng Q 1 2 a 1 2 Apply these two rules in the order presented above to the Citation forrn rn ko srnell7 in Kwara7ae Austronesian to obtain its pronunciation in nor rnal7 discourse form 34 Rule Application 62 If a language has more than one rule and they all do7 the rules have to nd a way to get along lt7s usually assumed that they are ordered and apply one by one7 but we can imagine other scenarios 341 Simultaneous application 63 Say we7ve got two rules labialization labial a round u V harrnony u a i iC What happens to the underlying forms below if each rule just nds any seg ments in the underlying form to which it can apply and performs the struc tural change dalbuge diburnpo griluda 342 Ordered Rules 64 If rules apply instead one by one in ordered fashion7 so that one rule7s output is the next rule7s input7 there are two possible outcomes with the same two rules Intro to Phonology Feb 12 2008 42 CHAPTER 3 PHONOLOGICAL RULES Fill in the derivations dalbuge dibumpo griluda labialization harmony dalbuge dibumpo griluda harmony labialization 343 Intrinisic vs Extrinsic rule ordering 65 Can we tell just from looking at a list of rules what order they should apply in a There have been proposals to do just thatithat is7 to impose an in trinsic rule ordering an ordering that is determined by properties of the rules themselves7 or maybe properties of the rules and the UR b But if each language can order the rules the way it likes7 rule ordering is extrinsic 66 Evidence for extrinsic rule ordering a What we need is languages or dialects that form a near minimal pair for the ordering of some rules Let7s try an example from SPE 67 Canadian raising rule of some English dialects a17aeo become A17 so before voiceless consonants a lald vs JAIt gaeodgl vs khsot ride7 right7 gouge7 couch7 b Does anyone in the class have this rule in their everyday speech 68 Pig Latin rule of children7s English language game lnitial consonants7 if any7 are moved to the end of the word7 and 61 is added to the end a phig lSBTn becomes Igphei eBTnleI Write the rule informally we havent yet reviewed the transformational notation that is needed 69 If you have Canadian raising and are reasonably adept in Pig Latin7 trans form the following words into Pig Latin and have your neighbor carefully transcribe them Intro to Phonology Feb 12 2008 35 RULE ORDERINGS 43 ice try might sigh 70 Now lets compare notes and see which dialects7 we7ve gotido we nd both orderings of Pig Latin movement7 and raising 35 Rule Orderings 71 72 73 351 74 75 Intro to Phonology Major research questions a What types of phonological patterns can be described with linear rule systems b What types of rule interactions does the linear system predict 5 types of rule interaction a Feeding b Bleeding c Counterfeeding d Counterbleeding e None For reasons that will become apparant7 patterns describable with feeding and bleeding rule interactions are often called transparant7 whereas pat terns describable with counterfeeding and counterbleeding rule interactions are called opaque7 Feeding Rule 1 feeds Rule 2 if 2 is applicable to some form that has undergone 1 but wouldnt be applicable to the same form if it had not undergone 1 lnformally7 Rule 1 creates a suitable input for Rule 2 Example Guinaang Kalinga dialect of Lubuagan Kalinga7 Austronesian language from the Philippines with 127000 157000 speakers a Assume that there are lots of examples like a7 where the rst stem vowel is not unstressed 0 Feb 12 2008 CHAPTER 3 PHONOLOGICAL RULES 44 a da bo b dopa c goba d Tom s e beta f Todaw g bosat h ponu i to6p j sog b k dog l 1 T0161 rn Towa hypothetical fathorn7 ring pots7 bath7 broken piece7 requesting7 sudden break7 lling7 satisfaction7 burning7 report7 tightening7 doing7 rnaking7 dinabo dimpana gimbana Timm osna bintaTna Tindawna binsatna pinnuna tin6pna sigg bna digg lna Till tna Tigwana hypothetical he measured by fathorn7 she red7 she bathed7 she broke7 he requested7 he snapped7 she lled7 he satis ed7 he burned7 he heard7 he made tight7 he rnade7 did7 Account for the different allornorphs of the in x in Give a derivation for dirnpana Getting the features right in l and is hardidon7t worry much about it Can we get a feeding interaction with simultaneous application Let7s try it on dirnpana x A variant on sirnultaneous application is all rules that can apply sirnultane ously to the input7 then again7 all rule that can apply simultaneously to the resulting forrn7 and so on until no more rules are applicable How would that work for dirnpana 352 Bleeding 76 Rule 1 bleeds Rule 2 if 2 is not applicable to some form that has undergone 17 but would have been applicable to the same form if it had not undergone 1 lnforrnally7 Rule 1 destroys a suitable input for Rule 2 Intro to Phonology Feb 12 2008 35 RULE ORDERINGS 45 77 Example English plural pi Z peas7 thOU Z toes7 dal z dolls7 phaen Z pans7 dag z dogs7 laeb z labs7 khIlDZ kilns7 khaesp s clasps7 mIt s rnitts7 blook s blokes7 khaf s coughs7 glas Iz glasses7 fIZ Iz zzes7 blaen iz branches baedg Iz badges7 w1 IZ wishes7 98103 1Z gar ages7 Account for these three allornorphsithat includes choosing an underlying forrn Give a derivation for Wif IZ Can we get a bleeding interaction with simultaneous application repeated sirnultaneous application Try them for for Wif Izl 353 Transparent vs opaque interactions 78 Feeding and bleeding interactions are called transparent7 because7 if we think of the two rules in declarative rather than procedural terrns7 they are both satis ed in the resulting forrn7 and ii this is achieved without super uous changes a Example It is evident from the string dimpana that it satis es both rules i dont have unstressed o in the environment VC CV77 ii nasal must match following consonant in certain features77 Intro to Phonology Feb 127 2008 46 CHAPTER 3 PHONOLOGICAL RULES b Likewise7 both rules below are obeyed by wifiz i adjacent obstruents must agree in voice77 ii dont have adjacent sibilants77 79 In opaque interactions7 this is not so 354 Counterfeeding opacity 80 Rule 2 counterfeeds Rule 1 if Rule 1 is ordered before Rule 27 and Rule 1 doesn7t apply7 but would have applied if Rule 2 had been ordered rst Rule 2 would have fed Rule 1 if the order had been reversed 81 In the simplest cases7 a rule A a B X Y has been counterfed if there exist surface representations containing XAY 82 Example Palauan Austronesian language from the Republic of Palau with about 157000 speakers X hisherz39ts X a rakth rakt l sickness7 b s sab sasab l re7 c b tkh batk l operation7 d final ragal l pain7 e kuk kaku l nail7 f r xkh rsk l rustling sound7 g Bak xl BakOl l cigarette7 h Tixs is l escape7 i bull s buT l betel nut7 Account for length and quality alternations you7ll need 2 rules Why does this interaction not qualify as transparent How is it different from bleeding Can we capture this case with simultaneous rule application Repeated si rnultaneous application Try it for is l 83 Patterns which can be described with a counterfeeding rule order are also said to exhibit underapplication7 Intro to Phonology Feb 12 2008 35 RULE ORDERINGS 47 355 Counterbleeding opacity 84 Rule 2 counterbleeds Rule 1 if Rule 1 is ordered before Rule 27 and Rule 1 applies7 but wouldnt have if Rule 2 had been ordered rst Rule 2 would have bled Rule 1 if the order had been reversed 85 In the simplest cases7 a rule A a B X Y has been counterbled if there exist surface representations containing B that was derived by the rule in environments other than L Y 86 Example Polish lndo European language from Poland with about 43 mil lion speakers sg pl a trup trupi horse7 b wuk wuki bow7 c snop snopi sheaf7 d kot koti cat7 e nos nosi nose7 f sok soki juice7 g klup klubi club7 h trut trudi labor7 i grus gruzi rubble7 j wuk wugi lye7 k Zwup Zwobi crib7 l lut lodi ice7 rn vus vozi cart7 n ruk rogi horn7 Account for the voicing and vowel height alternations you7ll need 2 rules Why does this interaction not qualify as transparent How is it different from feeding Can we capture this case with simultaneous rule application Repeated si rnultaneous application Try it for ruk 87 Patterns which can be described with a counterbleeding rule interaction are also said to exhibit overapplication7 Intro to Phonology Feb 12 2008 48 CHAPTER 3 PHONOLOGICAL RULES 356 Noninteraction 88 Alternations in Zoque rst person pre x N N rny N gloss N rny N gloss a parna rnbarna clothing7 e tatah ndatah father7 b burro rnburru burro7 f disko ndisko record7 c tsirna ndzirna calabash7 g kaju ijgaju horse7 d OTngoja p oTngoja rabbit7 h gaju ijgaju rooster7 faha belt7 rny belt7 SAk beans7 rny beans7 apun soap7 rny soap7 lawus nail7 rny nail7 ranth ranch7 rny ranch7 Sources Wonderly 19517 International Journal of American Linguistics 172105 123 Padgett 1995 Stricture in Feature Geornetry7 Stanford CSLl Publications 89 Rules for Zoque a Nasal Place Assirnilation Con8 cons nasal aplace 7 icont aplace Condition First person pre x In fact other pre xes undergo the alternation too b Voicing Assirnilation i cons cons a V01ce nasal c Nasal deletion cons cons nasal 0 cont What is unsatisfactory about the Voicing Assirnilation rule Can you suggest a way to write the rule which addresses this problem No ordering relation can be established between Zoque nasal place assirnila tion and voicing rules Why 90 Questions left unanswered a Why is a stop consonant voiced after a nasal Intro to Phonology Feb 12 2008 36 PHONOLOGICAL OPACITY 49 b Why is the nasal deleted before a fricative 91 Zoque ll monomorphemic words camdAhkaTOpya he meditates7 kendYOpra he wants to look7 minba he comes7 maijba he goes7 Why doesn7t the Nasal Assimilation rule apply above 92 Sources Wonderly 19517 International Journal of American Linguistics 172105 123 Padgett 1995 Stricture in Feature Geometry7 Stanford CSLl Publications See also Kenstowicz amp Kisseberth 197935 357 Other issues with rules and orderings 93 In SPE7 all these types of rule interaction represented as different order ings have equal statusino ordering is more or less favored than any other a Consequenlty7 rule ordering of this type predicts that all types of in teraction should exist7 perhaps with equal commonality b This might also be taken to mean that learners should be able to learn transparant and opaque patterns equally well 94 There have also been proposals to a rule out certain types of interaction entirely requires reanalysis of purported cases b make certain types of interaction the default7 with others needing to be stipulated by the grammar 95 What is the functional purpose of the rule a Eg in Zoque above7 the rule eliminate nasal Voice sequences 96 If rules repair7 bad structure7 are there other ways to repair them a Could there be another language that eliminates nasal Voice se quences by deletion metathesis denasalization 36 Phonological Opacity 361 Review 97 5 types of rule interaction a Feeding Intro to Phonology Feb 12 2008 50 CHAPTER 3 PHONOLOGICAL RULES 98 99 100 101 362 102 Intro to Phonology b Bleeding c Counterfeeding d Counterbleeding e None Phonological processes which are transparant7 are ones in which the declar ative statement or goal7 of the rules are satis ed in the surface form a Eg processes describable with feeding and bleeding rule interactions as in Guinaang Kalinga m in xation and in English plurals7 respec tively Phonological process which are opaque7 are ones in which the declarative statement or goal7 of the rules are not satis ed in the surface form a Phonological processes which are said to underapply7 are often asso ciated with patterns that are describable with a counterfeeding rule interaction as in Paluan b Phonological processes which are said to overapply7 are ofen associated with patterns describable with a counterbleeding rule interaction as in Polish Sometimes there is no interaction between two rules and it does not matter in which order they apply as in Zoque Recall the two kinds of hypothesis KampK discussed in the beginning of their book a The null hypothesis says Everything is memorized Eg there is no grammar b A rule based account separates the idiosyncratic from the predictable There is a grammar which is productive and allows us to apply the grammar in new ways Polish Revisited Polish Sanders 2003 Final devoicing counterbleeds Vowel Raising a Final Devoicing This generalization holds true of all Polish words7 regardless of language of origin7 morphological features7 or grammat ical category 77 p 48 Feb 12 2008 36 PHONOLOGICAL OPACITY 51 ward medial ward nal klubI klup club PLSC tQ8kaVI taskaf ready regularshort forrn kol nda kol nt Christmas carol NOM SGGEN PL dva raZI ras twice once talszts tales plate PLSC grizs s grig bite 2SGlMP btsgu bztsk edge GENAGO b Raising a u p 49 stem UR Nam 89 Nampl glass dVr dvur dvori mansion bl bul bole ache pkj pokuj pokojs roorn stW stuW stoWI table 2Lur ztur ZLHH a kind of sour soup ul ul uls beehive Vuj vuj vujs uncle muw muW muWI rnule c Raising is not triggered by voiceless stops or nasals sorne diacritics not shown p 50 stem UR Nam 89 Nam pl glass t gtp t 0p t opl peg kt kot koti cat VWs vwos VWOSI hair WZG wag wags elk sk sok soki juice grx grox grOXI pea drn darn dorm ho use t n tson tsoni trunk kp kop ipakOpE horse d Opaque interaction of Raising and Devoicing Intro to Phonology Feb 12 2008 52 CHAPTER 3 PHONOLOGICAL RULES stem UR Nam 89 Nam pl glass bb bup bobI bean rv ruf roVI ditch ld lut lodI ice dVz dovus dOVOZI supply n nug nos knife rg ruk rogi horn 103 Rules a o Raising 6 u C 7 nasal7voice b Final C Devoicing sonorant a Voice 7 c Final C Devoicing counterbleeds o raising xce tions oans 104 E p 7 1 stem UR Nam 89 Nam 89 Nam pl glass klr kolor gtkkolur kOlOH card suit Xl XOl gtkxul xols lobby parasl parasol gtkparasul parasols umbrella kvbj kovboj gtkkovbuj kovbojs cowboy grut W grutgow gtkgrutguw grut bw1 gland glb glop gtkglup globi globe snb snap gtksnup snobI snob 8pjizd spjizot gtkspjizut spjizodi episode kd kot gtkkut kodi code n8krlg nskrolok gtknskrz uk nskrologi obituary prlg prolok gtkproluk prologi prologue rkrd rskort gtkrskurt rskordi record fjrd fjort gtkiiiurt fjordi fj ord 105 The number of lexical exceptions7 especially in recent loanwords7 strongly suggests that the back vowel alternation is not synchronically produc tive 77 p 53 106 A wug test Intro to Phonology a 2 subjects b Stimuli plurals of Polish sounding non words like szlapagy leapogs embedded in sentences c Taskform the singular7 where the voiced C is in nal position d Example stimuli Feb 12 2008 107 363 364 MULTIPLE APPLICATION PROBLEM Bardzo ladnc inabody daly Jankuwi kzm39c nic llcrbalc Bardm ladne S7lagogx dal Janknwi kawe nie herbale cm quotl hc very many iunbods szlapogs gum Julm coffee nut ica 9 Example target production sentences leden bardzc ladny poz yczyl Jankowi i pleniqdzc i koszule One ven preuy lent John both money and 2 shin f Results Inspection of the vowel height in the nonce forms is in the same category as the vowel height of the plurals It appears raising does not apply to nonce forms Sanders concludes Polish oeraising is not productive and not a part of the synchronic grammar Yawelmani Yokuts MiniTypology of Opacity See C Wilson s handout 37 37 1 108 109 Intro to Phonology Multiple Application Problem Overview The basic problem to be dealt with today is what to do with a form that7 for some rule X Y ontains multiple instances of XAY7 either because XAY straightforwardly occurs twice in the form7 or because there are multiple ways of interpreting XAY eg7 it contains parentheses A B Multiple7 noneoverlapping matches a P p 344 To apply a rule7 the entire string is rst scanned for segments that satisfy the environmen al constraints of the rule Afe ter all such segments have been identi ed in the string7 the changes required by the rule are applied simultaneously b Example Remember Palauan vowel reduction from last time Add to our previous rules one that stresses the pen t Feb 12 2008 54 CHAPTER 3 PHONOLOGICAL RULES stress V stress COVCO V reduction ilong a a stress shortening a a ilong istress How would those rules apply to an underlying representation like abi13al7 110 Multiple matches one instance7s target is another7s environment Example optional schwa deletion French data originally from Dell 19701 suvanir a suvanir or suvnir to remember7 pasara a pasara or pasra will pass7 parvanir a parvanir parvnir to reach7 su ara a su ara gtksu ra will blow7 aridavepartir a aridavepartir or aridvepartir Henri had to go7 3akdavepartir a 3akdavepartirl 3akdvepartir Jacques had to go7 Write a rule for schwa deletion assuming that these data are correctl What does the quote from SPE above predict for this form tydavane you were becoming7 Actual result is supposedlyil7ve heard that French speakers react differ ently tydavane or tydavne or tyaff dvane7 but not gtktya Edvneldiscuss 1Dell7 Franois 1970 Les regles phonologiques tardives et la morphologie derivationelle du francaisi MIT dissertation Intro to Phonology Feb 12 2008 3 7 MULTIPLE APPLICATION PROBLEM 55 xamp e rom om 1son seen in your s u y ques ions o ea1an 111 E lf Cl W l td t Wl z Austronesian language from the Federated States of Micronesia with 1631 speakers mata mate eye7 matai metai my eyes7 matamami matemami our excl eyes7 yafar yefar shoulder7 yafarai yaferai our incl shoulders parasa perase switch7 parasarasa peraserase splash intrans marama merame moon7 maramali maremali moon of7 112 Two rules a Final V raising V a low 7 b Dissimilation V a low 7 CO cons7low 1quot What does the quote from SPE above say should happen to maramali7 372 Possible solution I directional application 113 Left to right Scan the string for the leftmost eligible segment and ap ply the rule to it Then scan the resulting form for the leftmost eligible segment7 etc 1quot Does this work for Woleaian French 114 Right to left Same thing but start with the rightmost eligible segment 1quot Does this work for Woleaian French 373 Possible solution II from Anderson 1974 115 1Find all segments eligible for the rule and circle them 2For each circled segment7 underline the smallest environment that lets the segment meet the rules structural description 2Data originally from Sohn7 Ho Min 1975 Woleaian Reference Grammar Honolulu Uni versity Press of Hawaii Intro to Phonology Feb 12 2008 56 CHAPTER 3 PHONOLOGICAL RULES 3lf the rule is optional7 you may uncircle some of the eligible segments and de underline their environments 4lf any circled segment is contained in some other circled segment7s underlined environment7 uncircle de underline the environments of as few segments as possible to get rid of these overlaps 5Now apply the rule simultaneously to the remaining circled segments Of course7 circling and underlining themselves have no theoretical statusthis is just a convenient way to say that we are identifying two different types of thing7 targets and environments3 1quot What does Anderson7s proposal predict for the French string tyvudrekasakalabado you would like that what the beadle7 1quot Does Anderson7s proposal help with Woleaian s s is s s is S S S S a H H H is 116 Tonkawa revisited you read about it in KampK Coahuiltecan language once spoken in Texas Today7s Tonkawa people are based in Oklahoma picena picnoT he cuts it7 picnanoT he wepcenoT he cuts them7 wepcenanoT he kepcenoT he cuts me7 kepcenanoT he icen castrated one steer7 notoxo notxoT he hoes it7 notxonoT he wentoxoT he hoes them7 wentoxonoT he kentoxoT he hoes me7 kentoxonoT he notox hoe7 netale netlo he licks it7 netlenoT he wentaloT he licks them7 wentalenoT he kentaloT he licks me7 kentalenoT he naxace naxcoT he makes it a re7 nxacenoT he wenxacoT he makes them a re7 wenxacenoT he kenxacoT he makes me a re7 kenxacenoT he 117 Recall KampK7s syncope rule Ch 3 V a Q CVC CV 3Anderson7 Stephen 1974 The Organization of Phonology New York Academic See chapter 13 Intro to Phonology Feb 12 2008 cutting it7 cutting them7 cutting me7 hoeing it7 hoeing them7 hoeing me7 licking it7 licking them7 licking me7 making it a re7 making them a r making me a re7 38 INTRINSIC ORDERING PROPOSALS 57 1quot If we simplify the rule to V a Q VC CV 7 what problems do we run into 374 Minimal vs maximal application 118 Something to think about in these and other cases of potential multiple application is the rule applying as often as possible or as seldom as possible Is this something we might want a theory to make reference to 38 Intrinsic Ordering Proposals 381 Review and Preview 119 Last time7 we looked at some proposals for how to apply rules the struc tural description is met multiple times in some form 120 Today7 we examine some proposals that address rule orderings Are there universal principles by which we can determine the order of rules 121 We7ve mainly assumed that a language can impose any order it wants on rules Many researchers have proposed that this is not the caseithat rules are intrinsically ordered and extrinisic rule ordering is required only in special circumstances 122 Vocabulary a Extrinsic Ordering The ordering ofrules for each language is language speci c b Intrinisic Ordering The ordering of rules for each language is deter mined by universal principles From a learning learning point of view7 what is attractive about intrinsic ordering 123 We look at two proposals with an interlude on the Elsewhere Condition77 a Koutsoudas7 Sanders and Noll 1974 simultaneous repeated applica tion plus proper inclusion precedence77 b Anderson 1974 natural orders c We will also ask What is the difference 7 Intro to Phonology Feb 12 2008 58 CHAPTER 3 PHONOLOGICAL RULES 382 Some relevant things to keep in mind 124 We have described different kinds of rule interactions a feeding 7 transparent 7 both rules apply b bleeding 7 transparent 7 one rule applies c counterfeeding 7opaque 7 one rules applies underapplication d counterbleeding 7opaque 7 both rules apply overapplication e none 125 Kiparsky 1968 observes that when languages change over time a We can describe these changes in terms of the grammars7 as opposed to just surface sound change b When we look at how the grammars change7 many changes can be described in terms of a re ordering of the rules i In particular7 rules tend to reorder themselves so that they can apply to more forms ii le Feeding and counter bleeding orderings appear to be fa vored 126 We might state this as rules want to apply maximally77 383 Goal of Koutsoudas et al 127 Adopt repeated simultaneous application KampK7s Free Reapplication Hypothesis7 FRH as a universal principle ofrule application and deal with the bleeding and counterfeeding cases with another principle 128 We have already observed that repeated7 simultaneous application of rules allows us to let rules apply maximally Feeding 129 Recall Karok Palatalization and Vowel Deletion a V deletion V 7 Q V b Palatalization s 7 f iC 130 niuksup V deletion niksup Palatalization nikfup How does simultaneous7 repeated application work FRH Intro to Phonology Feb 12 2008 38 INTRINSIC ORDERING PROPOSALS 59 Counterbleeding 131 Recall Counterbleeding opacity in Polish a o Raising 6 u C nasal7voice b Final C Devoicing sonorant a Voice 132 VZ o Raising vuz Final C Devoicing vus How does simultaneous7 repeated application work FRH Proper inclusion precedence 133 Latin American varieties of Spanish7 extrinsically ordered and rather ab stract analysis akez akezos 1 A a l 7 akel 7 2 A j 7 akejos that7 those7 What kind of rule ordering is this Try to apply these rules simultaneously and repeatedly to akeAwhat7s the problem 134 Koutsoudas et al propose following Sanders 1970 For any representation R7 which meets the structural descrip tions of each of two rules A and B7 A takes applicational prece dence over B with respect to R if and only if the structural description of A properly includes the structural description of B p 9 135 the structural description ofA properly includes the structural description of B means a you can match B7s structural description up with part of As that it is nondistinct from7 and still have part of A s structural description leftover Intro to Phonology Feb 12 2008 60 CHAPTER 3 PHONOLOGICAL RULES How does the de nition apply to the two Spanish rules Which rule is A and which is B Bleeding 136 Schaffhause dialect of Swiss German example originally from Kiparsky 1968 Singular Plural urnlaut context underlying boga boda bogaPL bodaPL surface bogs bode bogs bode 137 Analysis boga boda bogaPL bodaPL V fronting V a back umlaut contact me plurals bQSga bode cons o rounding 0 0 7 007 4 bode flat Why is this ordering crucial What happens if we use the Koutsoudas et al approach 138 Koutsoudas et al propose that in all apparent cases of bleeding and counterfeeding7 the rules need to be revised In this case7 they propose a context free rule 08 7 Apply this solution to bodaPL What additional fact needs to be true in Schaffhause for this to work 4Actually7 in the original it s not cor but graveli Grave is an acoustically based feature roughly7 lower frequencies are stronger for grave segments7 not much used these days Labials and velars are grave dentals and alveolars are grave aikiai acutei Intro to Phonology Feb 12 2008 38 INTRINSIC ORDERING PROPOSALS 61 384 Interlude The Elsewhere Condition 139 Anderson 19697 Kiparsky 1973 and fl5 140 Kiparsky argues that disjunctive ordering doesnt really have anything to do with expansion conventions though it happens to occur those cases He proposes that what really drives disjunctive ordering is 141 the Elsewhere Condition p 94 Two adjacent in the ordering rules of the form A a B P Q C a D R S are disjunctively ordered if and only if a the set of strings that t are nondistinct from FAQ is a subset of the set of strings that t RCS7 and b the structural changes of the two rules are either identical or incom patible 142 For a Malagasy like stress rule schema7 consider V a stress CVC a if we write it as two rules i V a stress CVC ii V a stress C how does the Elsewhere Condition say that these rules should apply Let7s discuss How does the elsewhere condition compare to proper inclusion precedence Are there cases where the two conditions apply differently 385 Anderson 1974 and natural orderings Selffeeding 143 Takelrna example from Anderson Chapter 9 Penutian language that was once spoken in Oregon 5Kiparsky7 Paul 1973 Elsewhere in phonology ln Stephen Anderson amp Paul Kiparsky eds A Festschrift for Morris Halle New York Holt7 Rinehart amp Winston Intro to Phonology Feb 12 2008 62 CHAPTER 3 PHONOLOGICAL RULES a a becomes if followed by i alx1xamis a alxiximis one who sees us b and any preceding als follow suit ikumanananinkh a ikumininininkh he will x it for him lohonananin a lohonininin I caused him to die for him c unless a voiceless C intervenes lohonananhi a lohonananhi 7 alsegesakhsanikh a alsegesakhsinikh we keep nodding to one an other Recall the rule that simultaneously applies to all the eligible vowelswhy was Anderson against it and what was his solution 1 7 isyll a 1 75mm lt 110239ce ha 1quot ls Anderson7s solution different in this case from Koutsoudas et al7s pro posal isyll voz39ce 1quot Something to think about do cases in which rules cant be allowed to apply to their own output have anything principled in common Natural Orderings 144 lcelandic lndo European language from Iceland with 2507000 speakers 145 umlaut and syncope barn child7 bo39rnum child datpl7 svant hungry neutnomsg7 sVongu hungry neutdatsg7 kalla lll call7 ko39llum we call7 hamar hammer7 hamri hammer datsg7 fi ll dandelion7 fi i dandelion datsg7 morgunn morning7 morgni morning datsg 146 Analysis Intro to Phonology Feb 12 2008 38 INTRINSIC ORDERING PROPOSALS 63 a Urnlaut a a o Cou syll b Syncope istress a Q C CV lf syncope precedes urnlaut7 what kind of ordering results for the UR katilurn kettle datpl For jakule glacier datsg77 What about urnlaut before syncope for katilurn7 For jakule7 147 As we observed with Karok HW 27 whether a rule ordering is feeding7 bleeding7 etc depends on the particular forms involved 770 um katil ketill kettle7 ko39tlurn kettle datpl7 ragin regin gods7 rognurn gods datpl7 alen alin ell of cloth7 olnurn ell of cloth datpl7 ulr ule bagg boggull parcel7 boggli parcel datsg7 jak jokull glacier7 jokli glacier datsg7 jat jotunn giant7 jo39tni giant datsg7 ulan 3ag bogull taciturn7 boglan taciturn rnascaccsg7 148 If the rules are right7 we have an ordering paradox How does Anderson resolve it 149 Anderson7s de nition of natural order Intro to Phonology Feb 12 2008 64 CHAPTER 3 PHONOLOGICAL RULES where only one of the two possible orders for a given pair of rules is feeding7 the feeding order is the natural one and that where only one of the two possible orders is bleeding7 the other order ie counterbleeding is the natural one In all other cases no natural order is yet de ned p 147 150 Anderson proposes that at least some pairs of rules are left unordered by a language7s grammar and so apply in their natural order in each case See Anderson ch 12 for some amendments to this proposal 1quot Again7 is this different from the Koutsoudas et al proposal 1quot So if a grammar consists of a list of rules and some statements about their orderings7 what does a change of the type observed by Kiparsky involve 1quot Can you think of other ways to deal with lcelandic 39 The Cycle 151 Now we can shift gears and consider the classic case of Palistinean Arabic 152 SPE the transformational cycle We assume as a general principle that the phonological rules rst apply to the maximal strings that contain no brackets ie7 morpheme or word boundaries7 and that after all relevant rules have applied7 the innermost brackets are erased the rules then reapply to maximal strings containing no brackets7 and again innermost brackets are erased after this application and so on7 until the maximal domain of phonological processes is reached p 15 153 Classic example Palestinian Arabic data originally from Brame 1974 Intro to Phonology Feb 12 2008 39 THE CYCLE 65 a Verbs without objects subject study7 understand7 2sg rnasc darast fhirnt 2sg fern darasti fhirnti 3sg rnasc daras fihirn 3sg fern darasat fihrnat 1p1 darasna fhirnna 2p1 darastu fhirntu 3p1 darasu fihrnu Whats the stress rule for this language7 based on the study7 paradigm Give a rule for the VQ alternations Determine the ordering of the two rules 154 Verbs with objects object he understood X7 she understood X7 You understood X7 1sg hirnni hrnatni fhirntni 2sg rnasc fihrnak fihrnatak fhirntak 2sg fern fihrnik fihrnatik fhirntik 3sg rnasc fihrnu fihrnatu fhirntu 3sg fern hirnha hrnatha fhirntha 1p1 hirnna hrnatna fhirntna 2p1 hirnkurn hrnatkurn fhirntkurn 3p1 hirnhurn hrnathurn fhirnthurn Lets do some derivations Intro to Phonology Feb 12 2008 66 CHAPTER 3 PHONOLOGICAL RULES Which forrns would be different if we did all the morphology rst and then applied the phonological rules 310 Lexical phonology 3101 Overview 155 As you read7 Kiparsky argues that this is not enough see Pesetsky 1979 for an earlier proposal along the same lines Different sub grammars apply at different levels of morphology in the lexical cornponent7 and an additional sub grammar postlexical applying after the syntax Lexicon English exarnple Root LEVEL 1 P rules stress7 trisyllabic shortening l T WFR7 if any prirnary in ection urnlaut7 ablaut7 irreg ular past tense and derivation al7 ous7 th7 irn LEVEL 2 P rules cornpound stress l T WFR7 if any secondary derivation un 7 ness7 er and compounding LEVEL 3 P rules laxing l T WFR7 if any past te nse SYNTAX Postlexical phonology postlexical rules apping7 aspiration7 156 Not clear Intro to Phonology Should the root pass through the Level 1 rules rst or go straight to WFR Feb 12 2008 secondary in ection regular plural and 310 LEXICAL PHONOLOGY 3102 Cyclicity in the lexical component 157 Within each level7 the phonological rules apply after each morphological operation thus the bidirectional arrows above 158 Evidenceexamples a WFRs can be sensitive to derived phonological properties eg ize7 which don7t apply to stems with nal stress eg public vs publicize Kiparsky7s interpretation is that stress rules apply to the stem on the previous cycle 159 Internal brackets are erased after each level7 so WFRs and phonological rules dont have access to morphological information from the previous level Postlexical rules dont have access to any bracketing 160 Evidenceexamples a Postlexical rules are automatic in the sense that they dont admit of lexical exceptions7 and dont care about morphological information 3103 Strict Cycle Condition 161 The idea is to allow lexical rules at least those that change feature values7 rather than lling in underspeci ed ones to apply only to environments newly made7 by either a morphological operation or a phonological rule in the same cycle This phenomenon is known as non derived environment blocking NDEB 162 Lexical phonology7s attempts to deal with NDEB were always kind of a mess7 and its not clear we7ve done better since then7 so rather than review the details of the proposals7 here are two classic examples7 Finnish and Sanskrit7 from Kiparsky Finnish 163 Ignore various other rules vowel harmony7 degemination7 ao to X haluta neta pietae lmata lla ajaa puhua Intro to Phonology Let himher X halutkoon noetkoon pietk n lmatkoon lkn ajakn puhukn aetive instructive in nitive U shehe was Xing haluten noeten pieten lmaten llen ajaen puhuen halusi nokesi pikesi lmasi li aji puhui Feb 12 2008 68 CHAPTER 3 PHONOLOGICAL RULES 164 So t a s i Can we modify the rule to deal with these cases tila roorn7 lahti Lahti7 valtin public7 aeiti rnother7 rnaeti roe7 silti however7 lirnnaati lernonade7 paasi boulder7 sinae you sg7 kuusi six7 165 Another rule is needed to account for this vowel alternation jokena river7 essive sg joki river7 nornsg rnaekenae river7 essive sg rnaeki hill7 norn sg aeitinae rnother7 essive sg aeiti rnother7 nornsg kahvina coffee7 essive sg kahvi coffee7 nornsg 166 How should the two rules be ordered7 given these data ignore hk alternation vetenae water7 essive sg vesi water7 nornsg kaetenae hand7 essive sg kaesi hand7 norn sg yhtenae one7 essive sg yksi one7 norn sg 1quot Whats the problem in vesi Sanskrit 167 ruki7 rule of Sanskrit s a s r7 u7 k7 i dadaxsi you give bibhar i you carry kramsjati he will go vak jati he will say 168 Now consider bisa lotus busa rnist barsa tip ablaut saxs instruct7 sasta a sista a si a participle V deletz39zm ghas eat gaghasanti a d3aksanti a d3ak anti 3 pl 1quot How is this like Finnish Intro to Phonology Feb 12 2008 310 LEXICAL PH ON OLO GY 3104 Barebones lexical phonology bibliography Mascaro7 Joan 1976 Catalan phonology and the phonological cycle Indi ana University dissertation Follows Kean7 Mary Louise 1974 The strict cycle in phonology Linguistic Inquiry 5 179 203 in extending Chomsky7 Noam 1965 Aspects of the Theory of Syntax Cambridge7 MA MIT Press7s syntactic Strict Cycle Condition to phonology to produce NDEB a phenomenon pointed out by Kiparsky 1973a in cyclic rulesand only in cyclic rules Includes now classic Catalan examples Pesetsky7 David 1979 Russian morphology and lexical theory MIT ms Addresses the problem that morphology seems to demand bracket erasure after each WFR later WFRs are blind to information from earlier stages7 but phonological rules need those brackets His solution was to interleave WFRs with the cyclic phonological rules7 instead of starting with the full morphological output and then erasing brackets All this happens in the lexicon Postcyclic rules come later7 after the syntax Kiparsky7 Paul 1982 Lexical phonology and morphology In I S Yang ed7 Linguistics in the Morning Calm Seoul Hanshin Pp 3 91 Shows how Pe setsky7s proposal explains various differences between lexical and postlexical rules Proposes levels7 and uses identity rules to capture NDEB Mohanan7 KP 1982 Lexical phonology MIT dissertation Revised as Mo hanan7 KP 1986 The Theory of Lexical Phonology Dordrecht Kluwer Like Kiparsky7 proposes levels7 but argues that some rules have to apply in more than one level as long as those levels are adjacent Kiparsky7 Paul 1985 Some consequences of lexical phonology Phonology Yearbook 2 83 138 And Booij7 Geert and Jerzy Rubach 1987 Postcyclic versus postlexical rules in lexical phonology Linguistic Inquiry 18 1 44 Propose that an additional level of postcyclic lexical rules word level applies before syntax 3105 Strict Cycle Condition revisited 169 170 Intro Strict Cycle Condition SCC a Cyclic rules apply only to derived representations b A representatin b is derived wrt rule R in cycle j iff b meets the structural analysis of R by virtue of a combination of morphemes introduced in cycle j or the application of a phonological rule in cycle j The SCC accounts for Non Derived Environment Blocking to Phonology Feb 12 2008 70 CHAPTER 3 PHONOLOGICAL RULES a As we saw in Finnish7 the rule t a s i applies in cases like haluti a halusi shehe was wanting7 and in cases where there is word nal front vowel raising as in vete a veti a vesi water nornsg77 but not in underived environments such as tila a tila roorn7 171 Kiparsky argues the 800 can be derived from the Elsewhere Condition7 which states that if two rules A and B are adjacently ordered7 and rule A de nes a subset of the contexts of rule l37 then rule A applies and not rule B a lntuitively only apply the more speci c rule b This requires an identity7 rule for lexical entries What does the Elsewhere Condition say should happen to the following ad jacently ordered rules tila a tila t a S i 3106 Icelandic 172 Application of Lexical Phonology to lcelandic frorn Kiparsky 1984 173 u epenthesis dagur day rnnornsg7 baer farrn rnnornsg7 tekur take 23sgpresind7 naer6 reach 23sgpresind7 174 j deletion byljar snowstorrn gensg7 krefji request 2pl7 byljir snowstorrn nornpl7 krefja request 3pl7 bylji snowstorrn accpl7 krefjurn request 1pl7 bylja snowstorrn datpl7 kref request 1sg7 byljurn snowstorrn datpl7 krefur request 23sg byl snowstorrn accsg7 byls snowstorrn gensg7 bylur snowstrorn nornsg7 How should we order these two rules Intro to Phonology Feb 12 2008 310 LEXICAL PHONOLOGY 71 175 u urnlaut har6um hor um hard datpl7 kallaurn kollurn call 1sg7 sagaur sogur sagas nornpl7 dagr dagur day nornsg7 1 Ordering 176 Syncope7 roughly certain unstressed Vs a Q lrn6sV Additional facts a syncope applies before case and derivational endings7 but not before the enclitic articles inn and i6 b Stress in Icelandic is almost always on the intial syllable hamar harnrner akur acne hofu nornsg7 nornsg7 hamri harnrner akri datsg7 hof6i datsg7 datsg7 hamra to harnrner7 hamarinn the hammer akurinn the acne hofu6i6 nornsg7 nornsg7 okrum acne datpl7 f ur lining nornsg7 dagur day nornsg7 f66ri lining datsg7 dagri day datsg7 f66ra to line7 f66uri6 the lining dagurinn the day sta6num nornsg7 nornsg7 177 Recall the problem from Anderson we have to order u urnlaut before syncope bagguli a bogglilicounterbleeding but we also have to order syncope before u urnlaut alinum a olnurnifeeding Shifting to Lexical Phonology7 is syncope lexical or postlexical u urnlaut u epenthesis Let7s try to resolve the ordering paradox using Lexical Phonology We should do derivations for dagur dagurmn bylur hamarf nn akur 6krum boggli sta6num Intro to Phonology Feb 12 2008 head nornsg head datsg the he nornsg the pl datsg 72 CHAPTER 3 PHONOLOGICAL RULES Some more dataiare they consistent with our analysis Nikulas Nicholas7 dagrinn dagurinn the day nornsg7 lifrinn lifrin the 7 nornsg7 311 Summary 3111 Data 178 We looked at phenomenon in the following languages in varying degrees of detail Kongo allophonic obstruent palatalization Sierra Popoluca Kasern English Mohawk Guinaang Kalinga Kwara7ae Paluauan Zoque Polish Malagasy Yawelrnani Yokuts Woleaian French Tonkawa lndonesian Karok Tangale Swiss German Takelrna lcelandic Hakha Lai Palistninean Arabic Finnish Intro to Phonology allophonic aspiration vowel deletion7 coalesence7 nasal assirnilation7 canadian raising some dialects7 lax ing7 trisyllabic shortening7 tapping7 phonology in plural forrnation7 stress conditioned allophonic vowel length syncope7 nasal assirnilation CV rnetathesis7 coalesence stress conditioned vowel neutralization voicing assirnilation7 nasal assirnilation7 nasal deletion nal devoicing7 raising consonant neutralization7 vowel hiatus7 vowel harrnony7 vowel lowering7 shortening7 nal vowel raising7 vowel dissirnilation optional schwa deletion syncope nasal assirnilation7 13 deletion palatalization7 glottal insertion syncope7 voicing assirnilation7 nal vowel devoicing urnlaut7 laXing ai alternation but voiceless consonants block urnlaut7 syncope7 j deletion7 phrase inital tonal falling7 tonal leveling quantity sensitive stress7 i deletion ts alternation7 nal e raising7 Continued on new page Feb 12 2008 311 SUMMARY Sanskrit 73 ruki rule7 3112 Theories 1 3 9 7 U 03 Intro to Phonology SPE a Language particular rules A a B C D b Language particular orerings Features and Natural Classes a Which groups of sounds are commonly affected by rules across world7s languages b Articulatory and acoustic de nitions of features Extrinsic Rule Ordering ordering relations that do not follow from general principles a Relationship to opacity counterfeeding and counterbleeding b Opaque orderings account for certain systematic exceptions Multiple application problems a How to apply a rule to a form that contains more than one instance of its SD b Can a rule apply to its own output c Theories simultaneous everywhere application Andersons no overlap proposal directional or linear rule application iterative minimal constraint violation Alternatives to extrinsic ordering a Proper lnclusion Precedence Principle b Natural Orderings c Elsewhere Condition Phonology and Morphological interaction a SPE Cycle b Lexical Phonology Feb 12 2008 74 CHAPTER 3 PHONOLOGICAL RULES Morphological Levels Word Formation Rules Phonological rules for each level7 plus a postleXical component Strict Cycle Condition iv Appeals to morphology account for systematic exceptions to phono logical rules7 eg NDEB 7 Brief introduction to syllabi cation and stress a sonority sequencing principle b quantity sensitive vs quantity insensitive distinction in stress rules 8 Distinguishing among proposed grammars a An observationally adequate grammar accepts the utterances that a typ ical speaker has been exposed to b A descriptiuely adequate grammar behaves the same as a typical speaker when confronted with novel utterances 0 When confronted with multiple descriptively adequate grammars7 select the simplest7 one often interpreted to mean shortest one c An emplarzatorz39ly adequate theory generates a descriptively adequate gram mar7 given a set of typical learning data 3113 Preview 1 Non linear representations a Autosegmental analysis of tone7 geminates7 assimilation b Do they simplify our rules and lead to new insights 2 Theories of constraints a Conspiracies7 surface constraints7 and linguistically signi cant general izations b Many logically possible ways to integrate theories of rules and constraints i Constraints trigger rule application ii Constraints block rule application iii do both c Constraints replace rules altogether Optimality Theory 3 Metrical Stress Theory Intro to Phonology Feb 12 2008 311 SUMMARY 75 a Given the logically possible ways7 languages can place stress on wordls7 what are the attesed patterns b How does metrical stress theory account for these patterns i iambictrochaic law ii syllable weight and prominence distinctions iii 4 Goals of contemporary theory a Eliminate language particular rules in favor of a limited set of basic operations b Eliminate abstract representational assumptions by appealing to pho netic properties c ldentify universal aspects of phonology Intro to Phonology Feb 12 2008 76 CHAPTER 3 PH ONOLOGICAL RULES Intro to Phonology Feb 12 2008 Chapter 4 Representations 41 of feature bundles So majab Representing words with features So far we have worked with linear representations7 where words are strings nas cons lab nas Cons low nas Cons high nas Cons low nas cons labial However there are other logically possible ways we might represent words For example7 we might put the feature nasal on its own tier7 nas nas cons Cons Cons Cons cons lab low high low labial 3 We might put every feature on its own tier nas nas cons Cons cons lab labial low l low l low high l high high 77 We might even adopt a skeletal structure like the following 78 CHAPTER 4 REPRESENTATIONS nas inas labial labial cons icons cons 411 How can we decide 5 Changing the theory in this way is a good idea only if the new theory does a better job than the old at correctly distinguishing highly valued from lowly valued grammars or grammar fragments 6 As before7 the claim is that rules that can be expressed in a simple form though we will not spend time spelling out how rule simplicity is to be calculated are highly valued So7 we are interested in orules that look relatively complicated relative to other rules7 that is in the old theory but relatively simple in the new oneinew theory predicts they are highly valued orules that look relatively simple in the old theory but relatively com plicated in the new oneinew theory predicts they are lowly valued 42 Tonal Association 7 Kikuyu Niger Congo language from Kenya with about 53 million speakers discussed here based on Goldsmith 19907 whose data come from Clements amp Ford to r51 ir we looked at7 ma r r ir they looked at7 to mio r51 ir we looked at him7 ma m r51 ir they looked at him7 to m r r ir we looked at them7 ma m r r ir they looked at them to tom ir we sent7 ma tom ir they sent7 to mio tom ir we sent him7 ma m tom ir they sent him7 to m tom ir we sent them7 ma m tom ir they sent them7 1quot Take a minute to ascertain the basic factsion what does the tone ofthe tense suf x ir i r depend On what do the tones of the two verb roots in bold depend On what do the tones of the object suf xes underlined depend Intro to Phonology Feb 12 2008 42 TONAL ASSOCIATION 79 Ideas for how we can account for this with linear representations and rules assume a feature hi tone 8 In the autosegmental77 notation proposed by Goldsmith7 we can write a rule thus T77 stands for any tone7 such as H or L in this language peninitial association W 00 V CoV T 9 Yes7 this is a rule Its structural description is wd Co V 00 V T ie7 everything except the dashed line and the structural change it requires is insertion of the association line 10 We need two more rules for the rest of the tones association convention V 00 V l TT initial association W 00 Q 00 V T The circle is part of the structural description7 and means not associated to anything on the other tier 11 For Goldsmith7 association conventions actually derive from universal prin ciples7 and dont need to be speci ed on a language particular basis Let7s apply this grammar fragment to derive we looked at them7iwhat must we assume about the association of tones in underlying forms Intro to Phonology Feb 12 2008 80 CHAPTER 4 REPRESENTATIONS 12 All three rules are typical of the kind of thing you see in tone languages7 and all three rules are some of the simplest that could be written in this notation 1quot Compare this to the linear analysis we developed above do the linear rules look simple compared to other7 less plausible linear tone rules we could write 43 Review 13 Last time we saw the following kind of metric to decide when faced with two possible grammar formalisms The formalism which highly values7 natural processes7 is to be preferred 14 Of course we have to be explicit about what highly valued7 and natural processes7 mean a lntuitively7 highly valued7 means simple b lntuitively7 natural processes7 means those we nd across many un related languagesiie those processes that make us realize language variation is not arbitrary c Rigrous notions are hard to nd eg see discussion in Kracht forth coming 15 This idea above helps justify the autosegmental representation of tone be cause a the tonal rules needed in this formalism to explain typical tonal pat terns eg Kikuyu are more highly valued than other logically possi bly rules we could write which describe less common7 or unattested patterns b the SPE style rules needed to explain typical tonal patterns are less highly valued than other logically possibly rules we could write which describe less common7 or unattested patterns 16 Autosegmental representation of tone a tonal features exist on a tier7 separate from the words b They are associated with particular vowels by virtue of being linked7 with them c One facet of this representation is that an element on a tier can be linked to more than one element on another tier d No crossing line7 convention 17 Today we consider another autosegmental approach7 where the tiers are the CV tier and the melodic7 tier Intro to Phonology Feb 12 2008 44 CV AND MELODIC TIERS 81 44 CV and melodic tiers 18 The idea rather than being a just string of matrices7 a representation con tains two strings7 one of C and V positions the skeletal tier or level and one of matrices the melodic tier or level7 plus a relation listing which skele tal positions are associated with which matrices indicated by association lines when we draw these representations C V C V C V l l l l l l b a d u p i 19 The association relation need not be one to one CVVCV C V l l ts l l b a d u i 45 Assimilation as spreading 20 We can ask what the difference is between the following representations of geminates C C C C l l false t t true t 21 Obligatory Contour Principle OCP a At the melodic level7 adjacent identical elements are forbidden77 Mc Carthy 19861 b This is a constraint7 argued to be universal 22 Consequences of OCP a in monomorphemic words7 geminates are true b fake geminates may occur across morpheme boundaries i In some languages7 a rule can convert fake geminates to true ones McCarthy 19812 23 Two theories of assimilation a Assimilation as spreading 1McCarthy7 John 1986 OCP effects gemination and antigeminationi Linguistic lnquiry 20 7199 2McCarthy7 Johni 1981i The Represenation of Consonant Length in Hebrewi Linguistic lnquiry 12 322327 Intro to Phonology Feb 12 2008 82 CHAPTER 4 REPRESENTATIONS C C C l a produces true geminates t t f eature change C C C l a l l produces fake geminates C l n b Assimilation as C l n t t t 24 Hayes7 strategy determine whether assimilation is spreading or feature change by examining the behavior of geminates ln Palestinean Arabic7 a rule of epenthesis inserts in between certain con sonant clusters futt a futit entered lp sg But this rule fails to apply to tautomorphemically sitt a sitt How does this fall out from the autosegmental representations above 25 This property of true geminates has been called integrity 46 Geminate Inalterability 26 Note that the rst half of a geminate often behaves differently from other consonants see7 eg7 Hayes 1986a3 oJapanese non nasal coda is OK if rst half of a geminate oPersian v 6 w V C7 unless rst half of geminate 1novruxz a nowruxz New Year7 2 aev a aevv barley7 3but aevvael rst77 qolovv exaggeration7 27 This is explainable using a C V skeleton a Japanese C V C C V V l l escapes prohibition gtkcons7 nascons g a k 0 b Persian C C V C l l l aVOidS requirement v a w unless followed by 8e v 8e 1 co nsl 3Hayes7 Brucei 1986a lnalterability in CV phonology Language 62 Intro to Phonology Feb 12 2008 46 GEMINATE INALTERABILITY 83 ut it won aso e exp a1na e w1t 1near representations7 1 we a ow 28 B ld l b l bl h l i if ll the feature long how7 ons1 er t e 1near versions 0 two rues rom o a ata 7 rom ayes 29 C d h l i f l f T b B k f H 1986b4 i 75071 a glottal formation 760m a T 7 C ganup taon a ganu taon every year7 dht lali i a dohz lali i and the hen harrier7 halak batak a hala batak Batak person7 lap piggol a la piggol wipe off the ear7 ma13ihut taon a magihu taon according to the year7 halak korsa a hala korsa Korean person7 b denasalization 607 6 7m 607 nas 7110206 7110206 magmum tna a maman tna rin pa m w1ne 39 k 39 k d k l i 7 mana 11 en a mana 11 en or a en 13 p 1p k p 1p p 7 holom saotik a holop saotik somewhat dark7 mananom plug a mananop plrllj ury a is 4b di h7 mamsrs a u a mamsrs a a 11 00 at a mosqu1to net 13 kal bb k k l bb l k i 7 ow 0 we now t ese are t e un er ying orms n care u speec 7 a 30 H d k h h d l i f 7 l fl h ll these rules are optional 1 Order Glottal formation applies within morphemesiit7s not a derived environment ruleiyet it doesnt apply to a morpheme internal geminates Can we patch up the linear account to explain this diktatr a rtrt a but d8kk8 a pittu a and aijsa a dthator r0er dskks pit tn aksa dictator7 to knock down7 sh7 door7 sh7 4Hayes7 Brucei 1986b Assimilation as spreading in Toba Bataki Intro to Phonology Feb 12 2008 84 CHAPTER 4 REPRESENTATIONS What about these cases across a rnorpherne boundary adatta a adaYta our custom7 suddutta a sudquta our generation7 31 Hayes7s argument yes7 we can capture the Toba Batak facts with linear rules But7 in the linear theory a glottalization rule that fails to apply in just these environments owhere denasalization has applied or n assirnilation not shown Denasalization can be thought of as a regressive assirnilation pro cess oto a tautornorphernic gerninate is not given a higher value than a rule that applied in some other corn bination of circumstances 32 Hayes contends that treating tautornorphernic gerninates and clusters that have undergone assirnilation the same wayiresistant to rules that would apply to the rst halfiis a cornrnon7 highly valued behavior Therefore7 we prefer the theory that can express this situation sirnply 33 In order to reproduce Hayes7 result7 let7s assume that the features are split onto two tiers a ocentral tier lips and tongue sonorant7 continuant7 labial7 coro nal7 dorsal7 anterior7 hi7 operipheral tier velurn and larynx nasal7 voice7 spread glottis7 constricted glottis b Another issue addressed is how to intepret the structural description SD of autosegrnental rules Hayes assumes that SDs are to be inter preted emhaustz39velyiie absence of a link is signi cant How could we write the denasalization rule as a spreading type rule Intro to Phonology Feb 12 2008 46 GEMINATE INALTERABILITY 85 34 The glottal formation rule Hayes provides is more complex and we see the SPE style rewrite rules being used in conjunction with the autosegmental analysis 75071 cont 0 Central Tier l C v C i 0 cv Tier l l H HOOHSEJl Peripheral Tier Why does glottal formation fail to apply just where it does 35 Moral assimilation creates a special relationship between two segments involved7 which in uences how they behave with respect to later rules Autosegmental representations can capture this directly7 but linear repre sentations can7t a grammar that displays the phenomenon is valued no more highly than a grammar that doesnt Intro to Phonology Feb 12 2008 86 CHAPTER 4 REPRESENTATIONS Intro to Phonology Feb 12 2008