PERSISTENCE OF THE LATIN ACCENT IN THE NOMINAL SYSTEM OF CASTILIAN, CATALAN AND PORTUGUESE
By
SONIA RAMÍREZ WOHLMUTH
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
2008
1
© 2008 Sonia Ramírez Wohlmuth
2
To the memory of my parents, Manuel D. Ramírez and Estelle López Ramírez.
3
ACKNOWLEDGMENTS
I would like to thank the University of Florida for funding through the University of
Florida Alumni Graduate Fellowship initiative which enabled me to become a full time student
while completing coursework toward the doctorate. My home institution, the University of
South Florida, where I am a full time instructor, permitted flexible scheduling of my classes and
reduced employment during the time I was taking classes. As I began the dissertation writing
phase I received a Personnel Development Leave for spring semester 2008 at the University of
South Florida which gave me the necessary impetus to write the preliminary chapters and begin
data analysis.
Coursework at the University of Florida provided a good foundation for the work on my
dissertation. I would like to acknowledge perspectives gained through coursework in historical
French linguistics with Jean Casagrande and George Diller, and in particular coursework taken
with members of my dissertation committee: Hispanic linguistics with David Pharies;
phonological theory with Caroline Wiltshire; historical and Indo-European linguistics with Gary
Miller; and applied linguistics with Joaquim Camps. Interaction with fellow students with whom
I share common interests provided an opportunity to test ideas and pursue new directions. I am
grateful for the ongoing dialog with Gary Baker and the path forged by his 2004 dissertation on
palatal phenomena in Spanish phonology.
I would like to express my appreciation to my husband, Enrique, for his unwavering
support, encouragement, and acceptance of my continued distraction as concomitant of my
pursuit of educational goals. My son, Sebastian, has been a willing participant in conversations
on various phonological phenomena and made me realize that enthusiasm for my discipline can
be contagious.
4
TABLE OF CONTENTS page
ACKNOWLEDGMENTS ...............................................................................................................4
LIST OF TABLES...........................................................................................................................8
LIST OF FIGURES .......................................................................................................................14
ABSTRACT...................................................................................................................................20
CHAPTER
1 INTRODUCTION: THE PROSODIC WORD .....................................................................22
Approaches to Word Level Accent.........................................................................................22 Metrical Theory and Its Precursors.........................................................................................23 Early Metrical Theory.............................................................................................................24 Autosegmental Theory............................................................................................................33 Metrical Grid ..........................................................................................................................35 Metrical Theory and Language Typology ..............................................................................37
Moraic Theory .................................................................................................................39 Extrametricality ...............................................................................................................42
Optimality Theory ..................................................................................................................44 Architecture of the Syllable.............................................................................................46
Syllable onset ...........................................................................................................47 Syllable coda ............................................................................................................49 Syllable peak (nucleus) ............................................................................................51 Onset and coda clusters ............................................................................................52
Faithfulness and Repair Strategies ..................................................................................53 Word Accent....................................................................................................................56 Prosodic Foot...................................................................................................................57 Alignment Constraints.....................................................................................................59
Optimality Theory, Multiple Outputs, and Analogy ..............................................................60
2 THE NATURE OF THE LATIN ACCENT ..........................................................................68
Polemic of Pitch versus Stress................................................................................................68 Evidence of Prosodic Change.................................................................................................68 Early Latin ..............................................................................................................................73
Syncope and Accent ........................................................................................................74 Iambic Shortening and Cretic Shortening .......................................................................81 Vowel Reduction .............................................................................................................85 Treatment of Words of Greek Origin ..............................................................................87
Strategies of Moraic Preservation...........................................................................................90 Classical Latin Accent: From the Metrical Grid to the OT Tableau .....................................94
5
Learnability of the Latin Accent.............................................................................................98 The Transition from Classical Latin to Late Latin .................................................................99
3 FROM QUANTITY SENSITIVE TO STRESS ACCENT..................................................101
Chronology of Latin .............................................................................................................101 Stress Accent as Catalyst of Phonological Change ..............................................................103
Ambisyllabification and Mora Sharing .........................................................................119 Stop + liquid clusters in Romance.................................................................................123 Weakening of Consonants in Coda Position .................................................................125 Stop + liquid clusters in Romance.................................................................................127 Apocope.........................................................................................................................129 Hiatus and Onset Glides ................................................................................................130 Motivation for Vowel Loss ...........................................................................................133
4 ACCENTUAL PATTERNS IN THE NOMINAL SYSTEM OF LATIN: OUTCOMES OF THE FIRST DECLENSION ..........................................................................................139
OT Constraints and the Latin Accent ...................................................................................139 The Latin Nominal System...................................................................................................141 Construction of the Data Set.................................................................................................145 An Overview of the First Declension ...................................................................................147 Two-Syllable Nouns .............................................................................................................147 Three-syllable Nouns............................................................................................................164
First Declension Trisyllables with Penultimate Accent ................................................165 First Declensions Trisyllables with Antepenultimate Accent .......................................183
Four-syllable Nouns..............................................................................................................194 Four-syllable Nouns with Penultimate Accent..............................................................195 Four-syllable Nouns with Antepenultimate Accent ......................................................213
Five-syllable Nouns ..............................................................................................................221 Five-syllable Nouns with Penultimate Accent ..............................................................221 Five-syllable Nouns with Antepenultimate Accent.......................................................227
Summary of Prosodic Patterns in Nouns from the First Declension ....................................230
5 VOWEL LOSS AND THE RISE OF ULTIMATE ACCENT IN IBERO-ROMANCE.....233
Synchronic and Diachronic Vowel Reduction .....................................................................233 Second/Fourth Declension Nouns ........................................................................................234 Third Declension Nouns .......................................................................................................235 Two-syllable Nouns..............................................................................................................237
Second and Fourth Declension Disyllables with Penultimate Accent ..........................238 Third Declension Disyllables with Penultimate Accent................................................250
Three-syllable Nouns............................................................................................................259 Trisyllables with Penultimate Accent............................................................................259
Trisyllables with accent on a penultimate HC type syllable ..................................261 Trisyllables with penultimate accent on HV type syllable.....................................272
Trisyllables with Antepenultimate Accent ....................................................................280
6
Four-syllable Nouns..............................................................................................................286 Four-syllable Nouns with Penultimate Accent..............................................................286
Four-syllable nouns with HC type initial syllable and HV type penultimate syllable................................................................................................................292
Four-syllable nouns with HV or L type initial syllable and HV type penultimate syllable ............................................................................................295
Four-syllable Nouns with Antepenultimate Accent ......................................................301 Five-syllable Nouns ..............................................................................................................304
Five-syllable Nouns with Penultimate Accent ..............................................................304 Five-syllable Nouns with Antepenultimate Accent.......................................................315 Summary of Effect of Apocope on Prosodic Outcomes of Nouns from Declensions
2/4 and 3.....................................................................................................................321
6 PREFERRED PROSODIC TEMPLATES: CONCLUSIONS............................................323
Patterns with Penultimate Accent and /-a/ Class Marker .....................................................323 Patterns with Antepenultimate Accent and /-a/ Class Marker..............................................331 Patterns with Penultimate Accent and /-o/ Class Marker or No Class Marker ....................338 Patterns with Antepenultimate Accent and /-o/ Class Marker or No Class Marker.............348
APPENDIX
A APPENDIX PROBI..............................................................................................................367
B APPLICATION OF PERFECT GRID .................................................................................370
C VERSIFICATION: FROM METER TO RHYTHM ..........................................................374
D DATABASE OF NOUNS IN CATALAN, CASTILIAN, AND PORTUGUESE WITH COMMON LATIN ETYMON.............................................................................................382
LIST OF REFERENCES.............................................................................................................503
BIOGRAPHICAL SKETCH .......................................................................................................520
7
LIST OF TABLES
Table page 1-1 Parameters for the construction of a metric foot................................................................37
1-2 Edge Marking Parameter in Latin......................................................................................44
1-3 Possible syllable types .......................................................................................................46
1-4 Typology of light and heavy syllables ...............................................................................50
1-5 Sonority scale of phonological segments...........................................................................52
1-6 Ranking of ONSET and PARSE above FILL .........................................................................54
1-7 Foot form constraints for Latin ..........................................................................................57
1-8 Alignment constraints in Latin...........................................................................................60
1-9 Free variation with unordered OT constraints ...................................................................62
1-10 Foot pattern of hypocoristics in Portuguese, Spanish, and Catalan...................................65
2-1 Characteristics of stress-timed and syllable-timed languages ...........................................70
2-2 Syllabic weight of the dactylic hexameter .........................................................................72
2-3 Early Latin syncope ...........................................................................................................75
2-4 Pre-classical Latin word initial accent ...............................................................................77
2-5 Early Classical Latin word accent vs. Classical Latin word accent...................................78
2-6 Iambic Shortening in Early Classical Latin .......................................................................83
2-7 Word Accent and cretic shortening....................................................................................85
2-8 Prosodic configurations in words of two and three syllables ............................................95
2-9 Parsing of three-syllable end of word ................................................................................97
3-1 Latin historical periods ....................................................................................................101
3-2 Syncope of post-tonic short vowels in Latin....................................................................104
3-3 Cases of syncope in the Appendix Probi .........................................................................106
3-4 Changes in prosodic form as a result of syncope.............................................................108
8
3-5 Vulgar Latin constraints on accent ..................................................................................109
3-6 Evaluation of faithful and syncope candidates ................................................................113
3-7 Post-tonic syncope ...........................................................................................................114
3-8 Degree of difference in sonority between medial consonants (Geisler) ..........................118
3-9 Resolution of medial consonant clusters produced by syncope.......................................118
3-10 Frequency of occurrence of Latin medial clusters: stop + liquid.....................................123
3-11 Ibero-Romance reflexes of Latin stop + liquid clusters...................................................124
3-12 Diphthongization before stop + liquid clusters in Romance languages...........................128
3-13 Ibero-romance reflexes of medial consonant + yod sequences ......................................132
3-14 Vowel loss and faithfulness constraints...........................................................................134
3-15 Post-tonic vowel loss in Early Western Romance ...........................................................134
3-16 Latin words of 2 to 5 syllables marked for possible syncope ..........................................137
4-1 Nominal inflection in Latin..............................................................................................141
4-2 Distribution of H/L syllable type in first syllable of Latin 2-syllable nouns of the first declension ........................................................................................................................152
4-3 Contingency table for H/L in first syllable of Latin first declension disyllables.............152
4-4 Contingency table for H/L in first syllable of Catalan nouns from Latin first declension disyllables ......................................................................................................154
4-5 Contingency table for H/L in first syllable of Castilian nouns from Latin first declension disyllables ......................................................................................................155
4-6 Contingency table for H/L in first syllable of Portuguese nouns from Latin first declension disyllables ......................................................................................................155
4-7 Preservation of HC initial syllables in Catalan, Castilian, and Portuguese nouns from Latin first declension disyllables .....................................................................................155
4-8 Accented syllable in nouns from Latin first declension disyllables with HV initial syllable .............................................................................................................................160
4-9 Outcome of Latin first declension ('L) nouns ...........................................................160
4-10 Duration of vowels in European Portuguese in stressed CV syllables ............................162
9
4-11 Heavy/Light (H/L) outcomes of accented syllables in Catalan, Castilian, and Portuguese based on nature of tonic syllable in Latin first declension disyllables..........163
4-12 Contingency table for H/L outcomes in tonic syllable from (HC)('HC), (HV)('HC), and L('HC) first declension nouns ..................................................167
4-13 Distribution of heavy/light syllables in outcomes of first declension trisyllables with penultimate accent of the type (HC)('HC) ................................................................167
4-14 Tableau for output of first declension trisyllable with penultimate accent on HC type syllable .............................................................................................................................171
4-15 Tableau for output of first declension trisyllable with penultimate accent on HC type syllable with PARSE-σ and ALIGNR constraints ...............................................................172
4-16 Tableau for output of first declension trisyllable with penultimate accent on HV type syllable with PARSE-σ and ALIGNR constraints ...............................................................175
4-17 Distribution of heavy/light syllables in outcomes from first declension trisyllable patterns (HV)('HV) and (L)('HV) ......................................................................176
4-18 Resolution of hiatus in output of Latin first declension trisyllables ................................177
4-19 Tableau for output of first declension trisyllable with penultimate accent on HV type syllable with PARSE-σ and ALIGNR constraints ...............................................................180
4-20 Tableau for selection of hiatus or rising diphthong in Catalan........................................182
4-21 Tableau for selection of falling diphthong in Catalan .....................................................183
4-22 Prosodic templates of first declension trisyllables with antepenultimate accent.............183
4-23 Outcomes of syncope in first declension trisyllables with antepenultimate accent: ('HV)L type ...............................................................................................................184
4-24 Output of first declension ('HV)L.............................................................................185
4-25 Output of first declension ('L.L) ...............................................................................187
4-26 Output of first declension ('HV)L and ('L.L).....................................................188
4-27 Output of first declension ('HC)L .............................................................................190
4-28 Contingency table for H/L outcome in Catalan, Castilian, and Portuguese for the accented syllable from ('HC)L and ('HC) ..........................................................192
4-29 Tableau to assign accent in Fijian loanwords ..................................................................194
4-30 Prosodic templates for first declension four-syllable nouns ............................................195
10
4-31 Possible heavy/light syllable configurations in four-syllable first declension nouns with penultimate accent and HC1 penultimate syllable...................................................196
4-32 Nature of first three syllables in outcomes of first declension L.L.HC1.L and L.HV.HC1.L inputs..........................................................................................................197
4-33 Proto-Catalan sonority ranking ........................................................................................197
4-34 Phonotactic restrictions for codas in Catalan, Castilian, and Portuguese ........................199
4-35 Tableau for first declension input type L.L.HC1.L .........................................................200
4-36 Possible heavy/light syllable configurations in 4-syllable first declension nouns with penultimate accent and HV1 penultimate syllable...........................................................201
4-37 Cases of syncope in 4-syllable first declension nouns with penultimate accent..............202
4-38 Nature of first three syllables in outcomes of first declension HC.HC.HV1.L ...............204
4-39 Nature of the first three syllables in outcomes of first declension HC.HC.HV1.L and HC.L.L1.L........................................................................................................................206
4-40 Nature of the first three syllables in outcomes of first declension HV/L.HV/L.HV/L1.L ......................................................................................................211
4-41 Nature of the first three syllables in outcomes of all first declension tetrayllables with penultimate accent ...........................................................................................................212
4-42 Syllable count in output of X.HC1.L.L............................................................................214
4-43 Outcomes of X.HC1.L.L first declension nouns by syllable type ...................................215
4-44 Tableau for first declension input type HC.HC.L.L ........................................................216
4-45 Outcomes of X.L1.L.L first declension nouns by type of accented syllable and word level syllable count ..........................................................................................................217
4-46 Outcomes of five-syllable first declension nouns with penultimate accent.....................226
4-47 Output of first declension pentasyllables with penultimate accent by language and syllable count ...................................................................................................................227
4-48 Outcomes of five-syllable first declension nouns with antepenultimate accent. .............228
5-1 Vowel sonority hierarchy.................................................................................................237
5-2 Preservation of HC initial syllables in Catalan, Castilian, and Portuguese nouns from Latin second and fourth declension disyllables............................................238
11
5-3 Prosodic outcomes of Latin disyllables of the second and fourth declension with initial heavy syllable ........................................................................................................239
5-4 Differences in constraints governing word accent in Latin and Spanish.........................240
5-5 Constraints on complex codas in Catalan ........................................................................241
5-6 Constraint against vowel deletion in Early Romance: R-ANCHOR-V..............................242
5-7 Proto-Catalan sonority ranking ........................................................................................243
5-8 Nature of accent bearing syllables derived from Latin 2nd and 4th declension disyllables with L or HV initial syllable........................................................246
5-9 Accented syllable by type in the output of 3rd declension disyllabic nouns with initial HC syllable ............................................................................................................250
5-10 Comparison of prosodic outcomes of Latin disyllables with HC initial syllable from the 3rd and 2nd/4th declensions.......................................................................................251
5-11 Accented syllable by type in the output of 3rd declension disyllabic nouns with initial HV/L syllable ........................................................................................................254
5-12 Comparison of prosodic outcomes of Latin disyllables with HV and L initial syllable from 3rd and 2nd/4th declensions...................................................................................256
5-13 Constraints governing apocope in Ibero-Romance..........................................................260
5-14 Output of second/fourth declension (HC)('HC) by number of syllables and nature of accented syllable...............................................................................................261
5-15 Comparison of heavy/light syllables in initial and accent bearing syllables in output of second/fourth declension (HC)('HC)....................................................................265
5-16 Tableau for output of declension 2/4 trisyllable with penultimate accent on HC type syllable .............................................................................................................................266
5-17 Tableau for trisyllables with penultimate accent (declensions 2/4 and 3) .......................267
5-18 Comparison of heavy/light syllables in initial and accent bearing syllables in output of third declension (HC)('HC) ..................................................................................268
5-19 Prosodic outcomes of HC('HV) in declension 2/4 nouns ........................................274
5-20. Distribution of prosodic templates in output of third declension tetrasyllables with penultimate accent. ..........................................................................................................300
5-21 MAX/IO constraints related to the output of suffix -ione.................................................312
12
5-22 Distribution of heavy/light syllables in pentasyllables with -ione(m) suffix...................313
6-1 Distribution of initial heavy syllables in disyllables and trisyllables from first declension nouns..............................................................................................................327
6-2 Syllable count in outputs of first declension trisyllables with antepenultimate accent ...331
6-3 Distribution of three and four-syllable outputs from first declension tetrasyllables with antepenultimate accent.............................................................................................333
6-4 Hierarchy of preferred vowels in non-prominent positions.............................................342
6-5 Distinctive features in a seven vowel system ..................................................................343
6-6 OT tableau for final vowel deletion in Catalan................................................................344
6-7 OT tableau for vowel reduction in seven vowel systems ................................................345
6-8 Distribution and source of one and two-syllable nouns...................................................352
6-9 Distribution and source of three-syllable nouns with penultimate accent .......................356
6-10 Distribution and source of three-syllable nouns with antepenultimate accent ................358
6-11 Distribution and source of three-syllable nouns with ultimate accent.............................360
6-12 Distribution and source of four-syllable nouns with penultimate accent ........................361
6-13 Distribution and origin of four-syllable words with antepenultimate accent. .................363
B-1 Treatment of Ω on the metrical grid ................................................................................372
C-1 Duple rhythm in the Iberian ballad tradition....................................................................380
13
LIST OF FIGURES
Figure page 1-1 Branching trees with s/w nodes.. .......................................................................................25
1-2 Illicit configuration of a strong branch. .............................................................................25
1-3 Bracketing of contiguous elements (Liberman).................................................................26
1-4 Metrical tree with labeling and assignment of the feature [±stress]. .................................26
1-5 Insertion of foot boundary. ................................................................................................27
1-6 Application of LCPR .........................................................................................................28
1-7 Grid alignment showing stress deletion through application of RPPR..............................29
1-8 Selkirk’s model of syllable constituents. ...........................................................................31
1-9 Stress foot configurations for English (Selkirk). ...............................................................31
1-10 Prosodic hierarchy of the word. .........................................................................................32
1-11 Application of prosodic category labels to differentiate word accent... ............................32
1-12 Autosegmental representation of a prosodic word.............................................................34
1-13 Metrical grid.......................................................................................................................36
1-14 The revised prosodic hierarchy..........................................................................................40
1-15 Tree structures for heavy and light syllables based on moraic count. ...............................40
1-16 Selkirk’s superfoot. ............................................................................................................41
1-17 Latin prosodic word.. .........................................................................................................42
1-18 The intrinsic structure of the syllable (Hooper 1976). .......................................................48
1-19 Universal strength hierarchy for consonants (Hooper 1976).............................................49
1-20 Nested prosodic constituents..............................................................................................56
1-21 Overlapping constraints. ....................................................................................................63
2-1 Typology of accent (van Coetsem). ...................................................................................69
2-2 Alignment of scansion and word accent in Latin dactylic hexameter ...............................71
14
2-3 Iambic shortening...............................................................................................................83
2-4 Prosodic repair with REMOVE-μ......................................................................................93
2-5 The prosodic hierarchy ......................................................................................................94
3-1 Parsing of candidates for syncope in Classical Latin and Vulgar Latin. .........................107
3-2 Raddoppiamento sintattico as moraic compensation.......................................................121
3-3 Feature spreading in consonants in contact with /j/. ........................................................132
4-1 Likelihood of apocope by declension class and language. ..............................................145
4-2 Permissible syllable structures in Latin words of 2-syllables..........................................148
4-3 Extrasyllabic /s/ in Latin word initial and word final position. .......................................151
4-4 Outcomes of initial HC syllables from Latin first declension disyllables in Catalan, Castilian, and Portuguese.................................................................................................156
4-5 Coincidence of subsets of L type initial syllable in Catalan, Castilian, and Portuguese resulting from HC first syllable in Latin first declension disyllables. .............................158
4-6 Distribution of heavy/light syllables in tonic syllable of nouns from Latin first declension disyllables. .....................................................................................................164
4-7 Comparison of prosodic patterns in three-syllable Latin nouns. .....................................165
4-8 Comparison of distribution of heavy/light in S3 and S2 from first declension trisyllables with penultimate and antepenultimate heavy syllables. ................................169
4-9 Percentage of heavy syllables in word initial syllable and tonic syllable from first declension ('HC) and (HC)('HC)........................................................................173
4-10 Percentage distribution of heavy/light syllables in outcomes from first declension trisyllable patterns (HV)('HV) and (L)('HV) .....................................................177
4-11 Comparison of 3 syllable and 2 syllable results of first declension ('HV)L input....186
4-12 Percentage distribution of outcomes of first declension ('HV)L and ('L.L). .....189
4-13 Percentage distribution of outcomes of first declension ('HC)L...............................190
4-14 Percentage distribution of outcomes of first declension ('HC)L...............................191
4-15 Percentage of heavy/light syllables in word initial syllable resulting from first declension ('HC)L, (HC)('HC), and ('HC). .................................................193
15
4-16 Hasse diagram of OT constraints related to primary accent in Fijian loanwords. ...........193
4-17 Output of first declension HC.HC.HV1.L and HC.L.L1.L by syllable type. ..................205
4-18 Distribution of syllable types in the output of all first declension tetrasyllables with penultimate accent. ..........................................................................................................213
4-19 Heavy/light syllable configurations for first declension tetrasyllables with antepenultimate accent.....................................................................................................214
4-20 Distribution of syllable types in outcomes of first declension pentasyllables with penultimate accent. . ........................................................................................................226
4-21 Distribution of syllable types in outcomes of first declension pentasyllables with antepenultimate accent.....................................................................................................230
5-1 Vowel quadrilateral for 7 vowel systems with [ə] in nonprominent positions. ...............236
5-2 Nature of stressed syllable in Catalan, Castilian, and Portuguese nouns from Latin 2nd and 4th declension disyllables with L or HV initial syllable. ...................................248
5-3 Prosodic outcomes of all declension 1 and declensions 2/4 disyllables. .........................249
5-4 Comparison of percent values of prosodic outcomes of 2nd/4th and 3rd declension disyllables with HC initial syllable. .................................................................................252
5-5 Comparison of percentage values of prosodic outcomes of declensions 2/4...................257
5-6 Percentage distribution of prosodic outcomes of all declension 2/4 and declension 3 disyllables. .......................................................................................................................258
5-7 Prosodic templates for three-syllable nouns. ...................................................................259
5-8 Comparison of syllabic count and nature of the syllable with primary accent in outcomes of second/fourth declension (HC)('HC)....................................................263
5-9 Comparison of syllable count and nature of the syllable with primary accent in outcomes of third declension (HC)('HC). .................................................................269
5-10 Outcomes of HV('HC) and L('HC) in declension 2/4 nouns: Syllable count and nature of accented syllable. .......................................................................................270
5-11 Outcomes of HV('HC) and L('HC) in declension 3 nouns relative to syllable count and nature of accented syllable. .............................................................................272
5-12 Outcomes of HC('HV) in declension 2/4 nouns: Syllable count and nature of accented syllable. .............................................................................................................273
16
5-13 Relative frequency of 3-syllable and 2-syllable templates resulting from declension 2/4 input HC('HV).....................................................................................................275
5-14 Outcomes of HC('HV) in declension 3 nouns: Syllable count and nature of accented syllable. .............................................................................................................276
5-15 Outcomes of HV('HV) and L('HV) in declension 2/4 nouns: Syllable count and nature of accented syllable. .......................................................................................278
5-16 Outcomes of HV('HV) and L('HV) in declension 3 nouns: Syllable count and nature of accented syllable. .......................................................................................279
5-17 Outcomes of 'HC.L.L declension 2/4 nouns: Syllable count and nature of accented syllable (n=68). ................................................................................................................281
5-18 Outcomes of 'HC.L.L declension 3 nouns: Syllable count and nature of accented syllable. ............................................................................................................................282
5-19 Prosodic templates from second and fourth declension 'HV.L.L and 'L.L.L. .................283
5-20 Prosodic templates from third declension 'HV.L.L and 'L.L.L. ......................................285
5-21 Prosodic outcomes of second declension tetrasyllables with HC type initial and penultimate syllables........................................................................................................287
5-22 Prosodic outcomes of third declension tetrasyllables with penultimate accent and HC type initial and penultimate syllables...............................................................................287
5-23 Prosodic outcomes of second/fourth declension tetrasyllables of the type HV.X.'HC.X and L.X.'HC.X with penultimate accent. ...................................................289
5-24 Prosodic outcomes of third declension tetrasyllables of the type HV.X.'HC.L and L.X.'HC.L (n=13) with penultimate accent. ....................................................................291
5-25 Prosodic outcomes of second/fourth declension tetrasyllables of the type HC.X.'HV.L with penultimate accent. .............................................................................293
5-26 Prosodic outcomes of third declension tetrasyllables of the type HC.X.'HV.L with penultimate accent. ..........................................................................................................293
5-27 Prosodic outcomes of declension 2/4 tetrasyllables of the type HV/L.HV/L.'HV.L with penultimate accent. ..................................................................................................296
5-28 Prosodic outcomes of third declension tetrasyllables of the type HV/L.HC.'HV.L with penultimate accent. ..................................................................................................298
5-29 Prosodic outcomes of third declension tetrasyllables of the type HV/L.HV/L.'HV.L with penultimate accent. ..................................................................................................299
17
5-30 Prosodic outcomes of declension 2/4 tetrasyllables with antepenultimate accent...........302
5-31 Prosodic outcomes of declension 3 tetrasyllables with antepenultimate accent..............303
5-32 Outputs of declension 2/4 pentasyllables with penultimate accent. ................................307
5-33 Portuguese outputs of third declension pentasyllabic nouns with suffixes -tāte(m) and -tūdō .................................................................................................................................308
5-34 Catalan and Castilian outputs of third declension pentasyllabic nouns with suffixes -tāte(m) and -tūdō ............................................................................................................309
5-35 Realignment of syllables, moras, and segments in outputs of suffix -ione(m) in Catalan, Castilian, and Portuguese...................................................................................311
5-36 Outcomes of third declension pentasyllables with suffix -iōne(m). ................................313
5-37 Outcomes of third declension pentasyllables with suffix -ōre(m). ..................................314
5-38 Outcomes of second declension pentasyllables with antepenultimate accent and heavy initial syllable. .......................................................................................................317
5-39 Outcomes of second declension pentasyllables with antepenultimate accent and LH initial syllables. ................................................................................................................319
5-40 Outcomes of second declension pentasyllables with antepenultimate accent and all light syllables. ..................................................................................................................321
6-1 Prosodic outcomes of all 2-syllable and 3-syllable first declension nouns with penultimate accent ...........................................................................................................323
6-2 Outcomes of all first declension tetrasyllables with penultimate accent. ........................327
6-3 Outcomes of all first declension pentasyllables with penultimate accent........................329
6-4 Distribution of heavy and light syllables in tonic and word initial position in nouns derived from the first declension. ....................................................................................330
6-5 Prosodic outcomes of all 3-syllable first declension nouns with antepenultimate accent . .............................................................................................................................332
6-6 Prosodic outcomes of all 4-syllable first declension nouns with antepenultimate accent . .............................................................................................................................334
6-7 Distribution of heavy/light syllables in initial and tonic positions of all nouns with antepenultimate accent from the first declension.............................................................337
6-8 Distribution of heavy/light syllables in initial and tonic positions of all nouns with penultimate accent from the first declension. ..................................................................338
18
6-9 Outcomes of disyllablic nouns with penultimate acent in declensions 2, 3, 4.................339
6-10 Prosodic templates from declension 2/4 trisyllables with penultimate accent. ...............347
6-11 Prosodic templates from declension 3 trisyllables with penultimate accent....................348
6-12 Prosodic templates from declension 2/4 trisyllables with antepenultimate accent..........349
6-13 Prosodic templates from declension 3 trisyllables with antepenultimate accent. ............350
6-14 Preferred templates for nouns of one and two syllables. .................................................351
6-15 Preferred templates for nouns of three syllables with penultimate accent.......................355
6-16 Preferred templates for nouns of three syllables with antepenultimate accent. ...............357
6-17 Preferred templates for nouns of three syllables with ultimate accent. ...........................359
6-18 Preferred template for nouns of four syllables with penultimate accent. ........................360
6-19 Preferred template for nouns of four syllables with antepenultimate accent...................362
B-1 Skeleton to grid association of a superheavy syllable. ....................................................370
B-2 Perfect Grid L → R (Cairene Arabic)..............................................................................371
B-3 QS and stress assignment (Latin).....................................................................................371
C-1 Metrical analysis of a line of Saturnian verse..................................................................375
19
Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy
PERSISTENCE OF THE LATIN ACCENT IN THE NOMINAL SYSTEM OF CASTILIAN,
CATALAN AND PORTUGUESE
By
Sonia Ramírez Wohlmuth
December 2008
Chair: David A. Pharies Cochair: Caroline R. Wiltshire Major: Romance Languages
The Latin Stress Rule is well known and is the object of centuries-long study through
various theoretical prisms, including, most recently, generative phonology, autosegmental
phonology, metrical theory, and optimality theory. The basic facts are that primary accent of
stressable words is never word final; it falls on the penultimate syllable if and only if that
syllable is heavy. However, the loss of quantitative differences in the vowel system in the
transition from Latin to Romance necessitates a new basis for assignment of the primary accent
of a word. In generative phonology terms the Latin Stress Rule is opaque because the required
environment for application of the rule may or may not be present.
Optimality theory (OT) provides a mechanism for the study of diachronic phenomena
that is not based on rules or the existence of a particular environment to trigger change. Rather,
OT establishes ranked constraints to account for the relationship between input and output forms.
The input/output forms selected for this study are Latin nouns that have correspondents in the
three major Romance languages of the Iberian Peninsula, Castilian, Catalan, and Portuguese.
The tension between faithfulness to the input form and conformity to preferred constraints, often
universal in scope, is reflected in the set of active constraints and their ranking. The link
20
between primary accent and the segment that displays the effect of that accent (through duration,
quality, and intensity) is rarely broken. However, major differences among the languages of this
study are evident in the treatment of final unstressed vowels of second and third declension
nouns. Deletion of the unstressed final vowel in such cases, giving rise to patterns of ultimate
accent, is an important innovation in Ibero-Romance and corresponds to a constraint that rewards
right alignment of word edge and head foot. Frequency of occurrence of this new accentual
pattern follows an East to West gradient with the highest rate of frequency in Catalan and the
lowest in Portuguese. This study also shows that universal retention of the final unstressed
vowel of first declension nouns does not correspond to a facile morphological explanation;
rather, it manifests the desirability of a trochaic foot at right word edge with the familiar pattern
of duple rhythm. Such a pattern is obtained when the rightmost syllable contains an optimal
peak, not subject to elision, as is the case of /a/.
Limitation of this study to Castilian, Catalan, and Portuguese nouns with common etymon
provides an opportunity to view the effect of language specific constraints on common input
forms through comparison of resulting outputs. Within this reduced linguistic microcosm it is
possible to examine the role of positional prominence, optimal syllable architecture, alignment
with word edges, and rhythmic preferences as constraints that influence outcomes. The
divergence of Catalan with regard to preference for a monosyllabic head foot in nouns from the
second and third declensions is expected. However, the areas of coincidence are of greater
import and correspond to linguistic universals: light syllables are generally preferred; alignment
of the head foot is at the right word edge; the optimal trochee consists of two light syllables. The
single heavy syllable that emerges as variant of the head foot may be viewed as a potential
disyllable with an empty nucleus, a concept reinforced by paragoge in poetry and music.
21
CHAPTER 1 INTRODUCTION: THE PROSODIC WORD
Approaches to Word Level Accent
Although it is difficult to find a single, universal definition of accent, there is cross-
linguistic recognition of the prominence of a syllable within the word or lexical unit. The basis
of that prominence, however, is rarely unidimensional. Rather, various factors including pitch,
duration, and intensity coalesce to place a syllable in relief (Beckman 1986; Von Coetsem 1996;
Crosswhite 2001; Kager 1995). Studies of word level accent, cross-linguistic as well as language
specific, have offered insights into the typology of accent, its assignment at word level, including
predictability, iterability of metric feet, and restrictions imposed by morphology and phonology
(Beckman 1986; Von Coetsem 1996; Delgado Martins 1982; Fox 2000; Harris 1992; Hayes
1995; Liberman and Prince 1977). This study examines the perseverance of the original locus
of accent in nouns of Latin origin as they are reflected in Castilian, Catalan, and Portuguese and
the changing nature of that accent from a system of moraic trochees to one which derives its
binarity from the presence of two-syllables defaulting to one heavy syllable when it is not
possible otherwise to build a prosodic foot at the right word edge.
The selection of nouns for this study presents several advantages: they constitute a corpus
of sufficient size (approximately 3,600 items that were found to have correspondences in all
three Ibero-Romance languages) and are by and large unencumbered by morphological variants
in the declension cases that pass into Romance, primarily accusative and occasionally
nominative, that would alter computation of the primary accent. Analysis of the Latin input
forms and the three-way output forms in Castilian, Catalan, and Portuguese is effected using core
principles of metrical theory in the framework of optimality theory (hereafter OT) in order to
describe the nature of the Latin accent as well as project differences in constraints and priorities
22
of those constraints in the three major Romance languages of the Iberian peninsula. Treatment
of word accent prior to the rise of metrical theory is briefly examined followed by discussion of
metrical and optimality theoretic approaches to word level assignment of accent. Since neither
classical Latin nor the Romance languages formally depend on tone for prominence of the
primary accent the terms stress and accent are at times used interchangeably in the literature that
discusses the prosody of these languages.
Description of word level prosodic phenomena begins as a footnote or short chapter in the
works that form the foundation of modern phonological theory. While the earliest attempts to
describe the processes that determine primary and secondary stress assignment are often
incomplete and at times misleading they evince the realization that treatment of accent cannot
follow the precepts of segmental phonology. Assignment of accent is obligatory for prosodic
words and its placement corresponds to language-specific criteria. These include among others,
recognition of positions of prominence within the word, avoidance of sequential and equal
accents, and the role of syllabic structure in attracting or repelling accent. A brief review of the
literature provides insights into the different approaches that form the historical backdrop for
optimality theoretic approaches to word prosody.
Metrical Theory and Its Precursors
Major advances in phonological theory, including the nature of word level accent, were
stimulated by the publication of The Sound Pattern of English (henceforth, SPE; Chomsky and
Halle 1968). The rules governing assignment of word level accent in English, as seen in the SPE
Main Stress Rule and its corollaries (Chomsky and Halle 1968, 240-245), are extraordinarily
complex. The context portion of such rules at times conflates derivational and phonological
processes in order to produce the desired output. Despite their unwieldiness the SPE stress rules
demonstrate sensitivity to weight, the role of lexical categories, an awareness of differing stress
23
patterns in words of Latin and Germanic origin, the effect of morpheme boundaries, constraints
regarding stress clash, and stress subordination. There is also a recognition that stress is unlike
other distinctive features in that “its domain extends over sequences that are longer than a word”
(Chomsky and Halle 1968, 68). However, absent in the SPE approach to word accent are
formulations capable of capturing broad patterns of regularity such as which elements may bear
stress, the direction of computation of stress, and the required distance between stressed
elements. The several elements that work in tandem to determine stress are dispersed through
various rules such as the Main Stress Rule (Chomsky and Halle 1968, 240-245), the Alternating
Stress Rule (1968, 77-79), and the Stress Adjustment Rule (1968, 84) as well as related rules that
affect segments such as the Vowel Reduction Rule (1968, 100-126).
Early Metrical Theory
The work of Liberman (1979, based on his 1975 dissertation) and Liberman and Prince
(1977) represented an important departure from SPE in that stress was considered to be a “matter
of RELATIVE prominence among syllables, rather than as a degree of absolute prominence
attached to each vowel” (Hayes 1981, 1). Liberman (1979, 207-208) assigned metrical values, s
(strong) and w (weak), to pairs of syllables in binary branching tree structures. The binary
feature [±stress] is then assigned to each vowel dominated by s or w. According to Liberman
(1979, 207-208) the distinction strong/weak is assigned to nodes whereas the feature [±stress]
applies to segments. In its simplest form, the metric tree, labeled R for the rhythmic unit, results
in undesirable opacity. Although one would expect [ə] for both weak syllables in Figure 1-1, the
weak syllable of example A (‘bombast’) retains its full value of [æ] while in B (‘ballast’) it has
been reduced to [ə] (pronunciations from OED Online s. v. bombast, ballast). Liberman and
Prince (1977:264-265) note that while many two-syllable words follow the pattern of Figure1-1
24
B “this perfect correlation cannot be maintained in general, since a [+stress] vowel may well be
metrically weak, as words like gymnast, raccoon show.”
R R
s w s w [ˈbɒm. bæst] [ˈbæl. ləst] + + + - A B Figure 1-1. Branching trees with s/w nodes. A) Weak node with [+] segment. B) Weak node
with [-] segment.
The more important metrical consideration can be captured formally through a rule of
this type (Liberman and Prince 1977, 265):
(1.1) Stress Rule: If a vowel is s, then it is [+stress]. (Liberman and Prince 1977)
Since there are abundant counterexamples to the expected ‘if a vowel is w, then it is [-stress],’
Liberman and Prince suggest that this rule be considered rather a principle of well-formedness
for metrical structures whose main purpose is to disallow the output seen in the tree structure in
Figure 1-2.
*s | V
[-stress]
Figure 1-2. Illicit configuration of a strong branch.
In order to process the segments that have now been designated as strong and weak
Liberman (1979, 209-217) introduces two further concepts to account for the output of metrical
rules: bracketing and labeling. Bracketing of segments proceeds from the beginning of the word
from left to right creating a series of binary groups in which the leftmost member is strong. The
25
configuration of Figure 1-3 is representative of languages where words, unencumbered by
derivational or morphological constraints, routinely have initial stress.
s s s s
x x x x x
Figure 1-3. Bracketing of contiguous elements (Liberman).
Bracketing allows the construction of a tree to which labeling correctly assigns s and w
labels, with the strong element always appearing as the left daughter of a constituent. Finally,
the segmental feature [±stress], whose primary role is “determining vowel-reduction,” is
assigned (Liberman 1979:210). This segmental feature is a carryover from the SPE approach to
analysis of accent. For English, assignment of [-stress] to a vowel indicates that it will most
likely be realized as [ə]. In the example shown in Figure 1-4, the word ‘testament’ is correctly
rendered through bracketing, labeling, and assignment of the feature [±stress]. The value [-] here
generates [ə] for the two unstressed vowels, or [ˈtɛs.tə.mənt].
R
s w
s w tes tə mənt + - -
Figure 1-4. Metrical tree with labeling and assignment of the feature [±stress].
26
The vowel labeled [+stress] in Figure 1-4 is also in a closed or heavy syllable, an environment
that Halle and Vergnaud assert attracts stress (Selkirk 1980, 575). Bracketing is also subject to
conditions imposed by foot boundaries. The foot corresponds to a metrical algorithm
superimposed on disyllabic constituents, one of which must be stressable. For example, in a
three-syllable word in English there are two allowable configurations of the metrical tree, seen in
Figure 1-5, A and B. However, C violates the constraint on insertion of a foot boundary because
it divides a metrical unit (Liberman 1979:211-213).
| X X | X | | X | X X | *| X | X | X |
A B C Figure 1-5. Insertion of foot boundary. A) Left-branching tree. B) Right-branching tree. C)
Violation of well-formedness.
It can be readily seen that words like testament correspond to pattern A, a common pattern
for English nouns of Latin origin (cf. argument, firmament, instrument). On the other hand,
pattern B is representative of deverbal nouns from a later English derivational process that
preserves the original stress of the verb as in agreement. Because argue is also a verb and
argument follows pattern A it can be readily seen that neither morphological analysis nor stress
assignment are transparent for many such polymorphemic words in English. Patterns A and B
reflect broad patterns of regularity but cannot account for deviations from the expected word
accent which may be the result of factors such as relative chronology, analogy with other lexical
items, or faithfulness to original stress for borrowed words.
Liberman (1979, 212) is concerned not only with a description of word accent but also
with the broader metrical properties of language that extend beyond word level. He recognizes
27
that “stress rules are of a fundamentally different type from segmental rules, and thus have
inherently different properties, which include the misleading appearance of cyclical application”
(1979, 232). Liberman proposes two foot-boundary insertion rules and assumes that the first is
universal while the second, although specific to English, is probably valid for many other
languages as well. The first rule (1.2) is precursor to the alignment concept employed in
optimality theory. The second rule (1.3) indicates that when there are two-syllables, the first of
which is strong or stressed, a foot boundary must be inserted to the left. The third rule (1.4)
provides a basis for determination of relative strength among metrical units.
(1.2) All # boundaries are foot boundaries. (Liberman 1979)
(1.3) A foot boundary is inserted in the environment __ / [+] [-]. (Liberman 1979)
(1.4) In any lexical metrical constituent [MN], N is strong if and only if it is complex. (Liberman 1979)
Complex is defined as “dominating non-terminal material” (Liberman 1979, 213). The final
formulation of the principles seen in the three rules above is the Lexical Category Prominence
Rule (LCPR).
(1.5) LCPR: In the configuration [N1 N2], N2 is strong if and only if it branches. (Liberman and Prince 1977, 270).
Application of the LCPR is illustrated in Figure 1-6 where N2 is complex because it dominates
elements that are lower in the hierarchy.
Wd
N1 N2 s w s w [ə 'ɡriː mənt]
Figure 1-6. Application of LCPR
28
The word agreement in Figure 1-6 Liberman and Prince (1977, 270) corresponds to a
familiar pattern of deverbal nouns formed with the Latin instrumental suffix -mentum (Miller
2006, 78-84). However, there are also verbs such as augment, cement, complement, document
with word final accent. Liberman recognizes that there are many exceptions to accentual
patterns based on grammatical function as is the case of words formed with –mentum, as well as
those that reflect French versus Germanic origin (1979, 304).
Visualization of word level accent assignment as an iterative process (Liberman and Prince
1977, 298-304) is enhanced by use of the metrical grid. The baseline of the metrical grid is built
on a one-to-one correspondence to syllables (Liberman and Prince 1977, 315), that is, the grid’s
terminals reflect the relative strength of the corresponding syllables, numbered from left to right
beginning with the lowest level. This is formalized as the Relative Prominence Projection Rule
(Liberman and Prince 1977, 316).
(1.6) Relative Prominence Projection Rule (RPPR) (Liberman and Prince 1977) In any constituent on which the strong-weak relation is defined, the designated terminal element of its strong subconstituent is metrically stronger than the designated terminal element of its weak subconstituent.
Utilizing the same example as in Figure 1-6, application of RPPR is illustrated in Figure 1-7.
6 *4 *5 1 2 3 a.gree.ment | | | w s s s R A
4 1 2 3 a.gree.ment | | | w s w s R B
Figure 1-7. Grid alignment showing stress deletion through application of RPPR. A) Input
scansion. B) Output scansion.
29
The initial input, seen in Figure 1-7A contains violations such as clashing (stress bearing
adjacent syllables or terminal elements) in contrast to the preferred pattern of alternating stress.
Avoidance of non-alternating stresses licenses repair strategies such as stress retraction or stress
deletion (Prince 1983, 21). Kiparsky (1979, 424) ascribes these adjustments to the Rhythm Rule
which formalizes the universal principle of eurhythmicity, the alternation of stressed and
unstressed syllables in a word or phrase.
The bidimensional representation in Figure 1-7 captures the linear arrangement of the SPE
approach in the horizontal display of segments in the word, labeled weak or strong, while the
vertical dimension, the metrical grid, shows stress assigned to those elements that can bear stress.
The Liberman/Prince (1979, 1977) approach correctly predicts the locus of primary and
secondary stresses in words (and phrases) but does not develop a full-fledged metrical theory.
The Prosodic Hierarchy
The word level Lexical Category Prominence Rule (LCPR) of Liberman and Prince
leaves some unresolved questions which are addressed in Selkirk (1980). Beginning with a
discussion of the English syllable and generation of a tree-like structure of the syllabic template
for the word flounce, Selkirk (1980, 568-569) also incorporates the use of s/w markers.
However, in this case the s/w distinction reflects the relative sonority of the components of the
onset and rhyme (nucleus and coda). At the first level, the s/w distinction designates the onset as
weak in terms of sonority, or [+cons], while the rhyme is [+son]; all other s/w labels also reflect
a sonority hierarchy. The syllable can be without onset and without coda but the strongest
element (s terminal dominated by s terminal), the nucleus, must be present and constitutes the
stress bearing element. In Figure 1-8 the nucleus is the only [+vocalic] element in the complex
30
monosyllable. Therefore, it is marked as strong within a higher level strong constituent. The
structure here offers an interesting approach to the problem of the superheavy syllable.
σ s w s w w s s w s w | | | | | | f l a w n s Figure 1-8. Selkirk’s model of syllable constituents.
Role of the Stress Foot
Selkirk’s (1980, 570) prosodic hierarchy posits a stress foot (Σ) as the next level above
the syllable, which as seen above, has its own sonority hierarchy. For English there are two
basic configurations for the stress foot, monosyllabic and disyllabic, as well as a third type, the
stress superfoot (stray foot adjunction in later metrical theory). The constituents labeled as weak
in B and C in Figure 1-9 are stressless.
Σ Σ Σ´
σ σs σw Σs σw A B C Figure 1-9. Stress foot configurations for English (Selkirk). A) Monosyllabic. B) Disyllabic.
C) Superfoot.
Selkirk has effectively shown that below the level of the prosodic word there are constituents
that have their own hierarchical principles of organization. This is a significant enhancement to
the Lexical Category Prominence Rule of Liberman and Prince which relates only to the
prosodic word level. The hierarchy proposed by Selkirk (Figure 1-10) culminates at word level
31
but builds on the lower level prominence principles that operate at the level of the stress foot and
the syllable.
ω (Word) |
Σ (Stress Foot) |
σ (Syllable)
Figure 1-10. Prosodic hierarchy of the word.
The three tier system presented here is able to effectively solve the dilemma of the word pair
considered above, bombast and ballast. The Liberman/Prince analysis is given in Figure 1-11
alongside Selkirk’s (1980, 565) prosodic hierarchy. The addition of the stress foot layer provides
an explanation for the phonetic difference between the unstressed final syllables in the two
words. Since the final syllable in bombast forms a stress foot it receives a secondary accent
while the strong member of the binary branch carries the primary accent.
R
s w [ˈbɒm. bæst]
+ +
ω | Σ
Σs Σw | | σs σw
[ˈbɒm. bæst]
R
s w [ˈbæl. ləst]
+ -
ω |
Σ
σs σw
[ˈbæl. ləst] A A1 B B1
Figure 1-11. Application of prosodic category labels to differentiate word accent. A, B
(Liberman/Prince analysis). A1, B1 (Selkirk analysis).
Selkirk (1980, 574-575) cites the work of Halle and Vergnaud as further support for
dispensing with the segmental feature [±stress] in favor of a hierarchical metrical system that
computes word level accent on the basis of universal principles of prosody as well as language
32
specific rules. In the Halle/Vergnaud analysis a potential stress bearing syllable in English is one
with a branching rhyme consisting of a tense or long vowel, or a vowel followed by a coda. The
final syllable of ballast does branch yet it would be labeled weak in the Halle/Vergnaud metrical
tree (cf. Figure 1-11B). The failure to project on the metrical plane is not easily explained
without reliance on the stress foot (cf. Figure 1-11, A1).
Autosegmental Theory
Autosegmental theory represents a significant departure from the SPE approach to accent
and early metrical theory accounts. In terms of visual or spatial representation of the prosodic
word autosegmental theory utilizes a multidimensional representation in opposition to the linear
sequence of SPE and the bidimensional, hierarchical representation of metrical trees and grids.
The foundation of the prosodic word Goldsmith (1990, 48-50) construes as the skeletal tier. The
skeletal tier is a string of units or slots that could be represented as C (consonant), V (vowel), or
simply as X without specifying the nature of the segment. The phonological features associate to
the CV elements on the skeletal tier. However, other segments may also associate with the CV
tier. These constitute autosegments. The autosegmental tier was initially devised as a way of
representing tone; tone often affects several segments including discontinuous segments. In
similar fashion, an autosegmental tier can be used to visualize the interaction between
phonological segments, syllable structure, and assignment of word accent. Following the model
given by Goldsmith (1990, 195), an autosegmental representation of the Latin word bŏnĭtātem,
fem. sg. acc. (‘goodness’), is provided in Figure 1-12. Proceeding from the syllable tier the solid
lines represent the relationship between the phonological segments and the syllable in the
familiar hierarchy: syllable > onset, rhyme (> nucleus, coda). The syllable requires a vocalic
segment in order to be well formed. The number of segments associated with the syllable appear
in the skeletal tier. Thus, the last two-syllables are seen to branch. Goldsmith’s autosegmental
33
planes or tiers greatly facilitate the treatment of long vowels, diphthongs, geminate consonants, d
tautosyllabic clusters. The interaction between planes is clear in the three dimensional
representation.
The metrical plane in Figure 1-12 marks all of the potential stress bearing elements with x
on the syllable tier while indicating the metrical feet with parentheses on the foot tier; the final
syllable, enclosed in angled brackets, is considered to be extrametrical and does not project a
foot. On this level only the head of each foot is marked with x. The word level tier indicates the
locus of primary stress. The one other foot on the foot tier, a left-head trochee, may be said to
carry a secondary stress corresponding to the initial syllable of the word.
σ σ σ σ
X X X X X X X X X X
X X X < X > ( X ) ( X ) < X > ( 'X )
Figure 1-12. Autosegmental representation of a prosodic word, Latin bŏnĭtātem, f., acc. sg.
Independence of the syllabic structure from the metrical grid is of critical importance as
Halle (1998, 543) notes in a retrospective on the accent of English words: “It was originally
thought that feet are made up of syllables, but subsequent work has shown this to be incorrect ...
It was therefore proposed in Halle and Vergnaud (1987) that feet are composed not of syllables,
bŏ nĭ tā tem
Metrical Plane Syllable Foot Word
Syllable Plane Skeletal Plane
34
but of those pieces of a syllable that may bear stress.” Nevertheless, for Latin it is important to
consider the syllable in its entirety because the accented syllable, if it is penultimate, must also
be bimoraic. Because autosegmental representations are not linear but multidimensional it is
possible to see the constituents of a syllable on the skeletal plane. The lines connecting these to
the syllable have a branching rhyme for the last two-syllables whereas the first two do not.
Metrical Grid
The grid notation developed by Liberman (1979), Liberman and Prince (1977), Selkirk
(1980), Hayes (1981), and others is formalized in Halle and Vergnaud (1987) and Hayes (1995).
For Hayes (1995, 27) the metrical grid is a representation of a sequence of beats equally spaced
(horizontal dimension) within prescribed temporal limits. The beats vary in intensity and this is
shown by the vertical dimension. These general observations on construction of the grid
establish that the metrical grid is (1) hierarchical in nature; (2) has evenly spaced intervals; and
(3) observes downward implication, that is, a grid mark on a higher level has a corresponding
grid mark on all lower levels. The hierarchical nature of the metrical grid can be seen in its
graphic representation. Using Selkirk’s (1980, 565) hierarchy of word, stress foot, syllable as a
point of departure, the syllable level corresponds to line 0. Setting aside extrametricality for the
moment it can be said that every member of line 1 (foot level) has a corresponding member in
line 0, that is, line 1 is a subset of line 0; the same is true at word level and phrase level. In other
words, construction of the vertical dimension must respect the principle known as Continuous
Column Constraint (Hayes 1995, 34):
(1.7) Continuous Column Constraint (CCC) (Hayes 1995) A grid containing a column with a mark on layer n + 1 and no mark on layer n is ill-formed. Phonological rules are blocked when they would create such a configuration.
35
Because the principle of eurhythmicity rejects adjacent stresses the result is that the subsets
decrease as higher level lines are constructed. This can be seen in Figure 1-13, an abstract
representation of a four-syllable word consisting of a string of CV syllables with penultimate
stress (foot type is trochaic). In this schematic only the head of the foot is marked with ‘x’ on
lines 1 and 2. The boxes on the first and third V columns in Figure 1-13 demonstrate CCC in
determination of primary and secondary stress for a hypothetical four-syllable word. The
construction of lines 2 and 3 will also depend on language specific criteria, for example, some
languages require exhaustive parsing of the word whereas others construct only one word-level
accent, typically at left or right word edge. Construction of the foot row ensures that only parsed
segments will be ultimately designated as having a primary or other word level accent.
x
x x
x x x x
C V1 C V2 C V3 C V4
Line 2: Word level
Line 1: Foot level
Line 0: Syllable level
Figure 1-13. Metrical grid.
The construction of Row 1 (foot row) entails recognition of the relationship between head
and constituent. Halle and Vergnaud (1987, 8-28) assert that a relatively small number of
parameters are able to describe this relationship: [±HT] and [±BND]. The first parameter
designates whether or not a constituent is head-terminal, that is, whether or not the head is
aligned with one of the constituent boundaries; the second whether or not the head of the
constituent is separated from the constituent boundary by more than one element. Additionally,
[+HT] constituents can be either left headed or right headed. In Table 1-2 the conventions [+LH]
and [+RH] will be used to indicate location of foot head; --- corresponds to unattested patterns
36
and a period indicates non-head elements of the constituent. The resulting possibilities, then, are
the following:
Table 1-1. Parameters for the construction of a metric foot. [-BND] [+BND] [+HT] [+HT] / [+LH] *…. *. [+HT] / [+RH] ….* .* [-HT] --- .*. Because bounded feet have beginning and end it is now possible to construct bracketed grids.
Hayes (1995, 55) departs from the Halle/Vergnaud parametric approach on several issues. He
rejects non binary feet and asserts that the inventory can be limited to [+HT] and [±BND]. This
excludes ternary feet from any possible foot inventory leaving as possible foot types the four
shaded cells in Table 1-1.
Metrical Theory and Language Typology
Hayes’s inventory of possible foot types and case studies builds on a long tradition.
Liberman and Prince (1977) foresee that direction of scansion and type of foot provide important
data for discussions of typology and in the same year Hyman’s (1977) survey of 444 languages,
summarized in Fox (2000, 170) recognizes the following patterns:
(a) languages with dominant initial stress1 114 (b) languages with dominant second-syllable stress 12 (c) languages with dominant penultimate stress 77 (d) languages with dominant final stress 97 (e) languages with non-dominant stress 144 In the list above the preference for stress at word edges is well attested and follows a long
tradition in which stress serves a demarcative function. The third most frequent pattern,
penultimate stress, on the other hand, is viewed by as a function of preferred intonational pattern,
1 Dominant stress gives prominence, through devices other than pitch, to a particular syllable while reducing the prominence of others (van Coetsem 1996, 43). A frequent byproduct of dominant stress is vowel reduction, present in both Catalan and Portuguese.
37
that is, a contrast of high + low (Fox 2000, 170-171). Furthermore, the penultimate stress pattern
of Latin may well have morphological grounding. In the case of a noun consisting only of stem
+ declensional suffixes it can be assumed that accent on the stem might be a more desirable
outcome than accent on the inflection. The stem contains the ‘new information’ and would,
therefore, be treated in the same manner as inflectional focus in the phrase. While this may well
result in a high + low intonation pattern the melody is a byproduct of the word accent rather than
a determinant.
Hayes’s typology of accent (1995) is based on the metrical foot. He proposes three basic
bounded foot types (1995, 71): syllabic trochee, moraic trochee, and iamb. The syllabic trochee
(Hayes 1995, 63) is one which depends only on syllable count. The moraic trochee (Hayes 1995,
69) takes into account the weight of the syllable; accordingly, a moraic foot may consist of two-
syllables of one mora each or a single bimoraic syllable. The addition of the moraic trochee is
crucial for analysis of the Latin stress system. Kager (1995, 370-373) uses the operative
parameters that emerge from metric bracketing to describe foot typology and construction:
(a) Foot typology a. Extension: Bounded/unbounded b. Dominance: Left headed/right headed c. Quantity-sensitivity: Quantity sensitive/quantity insensitive/quantity-determined
(obligatorily branching) (b) Foot construction
a. Directionality i. Right to left/left to right
ii. Bidirectionality: Noniterative foot assignment beginning at one edge and iterative foot assignment beginning at the opposite edge
b. Iterativity: Iterative/noniterative (c) Word tree parameters
a. Dominance: Left-dominant and right-dominant b. Labeling: Strong (branching)/weak (nonbranching)
(d) Word tree dominance a. Left dominant/right dominant
38
With regard to extension, bounded systems are well known. These are the ones
corresponding to the three basic foot types: moraic trochee, syllabic