Date: Fri, 29 Oct 2004 00:54:18 -0400 From: Jeffrey Steele Subject: The Syllable in Optimality Theory
EDITORS: Féry, Caroline; van de Vijver, Ruben TITLE: The Syllable in Optimality Theory YEAR: 2003 PUBLISHER: Cambridge University Press
Jeffrey Steele, University of Toronto
The chapters of this edited volume each present an analysis of a syllable- related processes in Optimality Theory (OT). The fourteen papers that follow the introduction are grouped into four main sections -- 'Syllable Structure and Prosodic Structure', 'Nonmoraic Syllables and Syllable Edges', 'Segments and Syllables' and 'How Concrete is Phonotactics' -- based on their main focus. Given the volume's length and the breadth of topics covered, this review will be restricted to presenting a summary of the core arguments of each article, followed by an overall evaluation of the volume.
Chapter 1: Overview (Caroline Féry & Ruben van de Vijver). This first chapter consists of an introduction, as well as a summary of each of the following chapters. Féry and van de Vijver propose that the volume offers two important insights into the syllable and OT, namely the ways in which OT can provide solutions to previously problematic syllable phenomena and how the syllable can both reveal and solve problems within OT.
Chapter 2: Sympathy, cumulativity, and the Duke-of-York gambit (John J. McCarthy). McCarthy examines Duke-of-York (DY) gambits, that is serial derivations of the form A --> B --> A. He proposes that they are of two types: vacuous DYs, for which the postulation of stage B follows from independently motivated rules and strict serialism; and feeding DYs, in which the presence of B is necessary for the feeding of some intermediate rule. McCarthy argues that vacuous DYs are unnecessary. Using data from Nootka, Catalan and Slovak, he demonstrates that they can be accounted for through blocking under constraint domination. As concerns feeding DYs, using data principally from syllable-stress interactions in Bedouin Arabic, he argues that such derivations do not occur in general. Rather, the surface facts can be explained using Sympathy theory. Sympathy theory proposes that apparently opaque interactions involve a sympathetic candidate, which is faithful to the input vis-à-vis some faithfulness constraint. The optimal candidate must not only incur the fewest violations of the most highly ranked constraints, it must also be the most similar to the sympathetic candidate in some respect. The version of Sympathy theory proposed differs from previous versions which evaluated sympathy using intercandidate faithfulness constraints. The author argues that such an evaluation metric allows for an overly rich theory. To correct for this, McCarthy proposes that sympathy be defined in terms of cumulativity, i.e., in terms of subsets of unfaithful mappings. Only candidates whose set of unfaithful mappings is cumulative vis-à-vis the sympathetic candidate will satisfy the highly ranked sympathy constraint. DY derivations are inherently non-cumulative and thus always fail to be optimal, a typologically-desirable outcome.
Chapter 3: The controversy over geminates and syllable weight (Stuart Davis). Davis investigates the representation of geminate consonants, his principal claim being that they are best represented as underlying moraic consonants, and not as consonants linked to two root nodes or X-slots. The primary data involve two groups of inanimate singular-plural noun alternations in Sinhala that differ in output form (e.g. [mal@]- [mal] 'flower-flowers' versus [mull@]-[mulu] 'corner-corners', where [@] =schwa). Davis argues that the differences in gemination in the singular and in the absence versus presence of word-final vowels in the plural are related to underlying representation. Specifically, in the case of '[mull@]-[mulu]' type pairs, the final consonant of the root is underlyingly moraic. In Sinhala, which enforces moraic faithfulness while prohibiting final geminates, the geminate are realized in the singular, where a vocalic grammatical suffix follows. In contrast, in the plural, a final epenthetic vowel is inserted with the underlying geminate's mora being associated to this position. Davis then demonstrates that the constraint ranking proposed accounts for a similar alternation with roots involving final prenasalized stops. As final evidence for the moraic representation of geminates, the author presents data from genitive allomorphy, where both [e] and [ee] occur. The latter form occurs with monomoraic roots; in all other cases, the short form is used. Importantly, the geminate-final roots involved in the singular-plural alternations that Davis argues to be bimoraic do pattern with multimoraic forms. The author concludes this chapter by presenting reanalyses of previous data argued to favour against a moraic representation of geminates (viz. Hume et al. 1997, Baker 1997).
Chapter 4: The syllable as a unit of prosodic organization in Japanese (Haruo Kubozono). Through a discussion of five independent phonological phenomena, namely loanword truncation, zuzya-go (jazz musician's secret language), motherese, emphatic mimetics, and compound formation, Kubozono demonstrates the dominance of HL over LH structures in Japanese. In the data discussed, HL generally results from quantity alterations (e.g. HH-- >HL shortening; LL-->HL, L-->HL, H-->HL, LH-->HL lengthenings), which the author notes as being crucially different from iambic lengthening or trochaic shortening. It is argued that the HL-LH asymmetry cannot be explained with reference to moras alone. Rather, it is demonstrated that the syllable is an equally important prosodic unit as the mora in Japanese.
Chapter 5: Prosodic weight (Draga Zec). This chapter focuses on positional asymmetries involving light and heavy syllable nuclei in English, Mordwin, Old Church Slavonic and Asheninca. Rejecting the idea of a hybrid weight hierarchy based on mora count and the sonority of the nucleus, Zec instead proposes a prosodic hierarchy-based set of sonority constraints, with asymmetries falling out principally from sonority interactions between syllables and feet. Specifically, the Foot and the Prosodic Word (PWd) impose further sonority restrictions with the result that certain classes of segments may be excluded from headship of these constituents. Asymmetries, including the asymmetry in the ability of /l/ and /r/ to constitute the head of the foot in English, fall out from such Foot and PWd-associated sonority constraints.
Chapter 6: Syllables and moras in Arabic (Paul Kiparsky). The focus of this chapter is the typology of syllable structure differences among three Arabic dialect groups. The differences in question principally involve the wellformedness of consonant clusters and opaque epenthesis/stress interactions. Kiparsky argues that the dialects differ in whether they license semisyllables, i.e., moras unaffiliated with syllables and adjoined to higher prosodic constituents. Such a difference is represented formally as the relative ranking of LICENSE-MORA and markedness constraints governing the wellformedness of syllables and feet. The author argues that the analysis proposed provides evidence for a non-parallel, stratal model of OT (i.e. a constraint-based version of Lexical Phonology and Morphology). Under such a model, word and sentence-level phonology is shaped by different constraint systems that act in serial fashion. Indeed, under Kiparsky's proposal, syllable typology is a consequence of differences in the level at which semisyllables are licensed: lexically, lexically and postlexically, or at neither stage.
Chapter 7: Semisyllables and universal syllabification (Young-mee Yu Cho & Tracy Holloway King). In this chapter, Cho & Holloway King seek to account for apparent consonant cluster sonority violations in languages such as Georgian, Polish and Bella Coola, all the while maintaining the universality of the Sonority Sequencing Principle (SSP) and Exhaustive Syllabification (ES). The essence of their proposal is that such apparent violations involve semisyllables, i.e., syllables containing no moras which permit the syllabification of such clusters as a sequence of onsets. Once the complex morpheme structure of the three languages is taken into account, the principal difference between them is the range of segments that may be moraic. The languages in question differ from languages that disallow semisyllables in that the faithfulness constraints DEP and MAX dominate SYLLABLE-MORA, a constraint requiring all syllables contain a mora.
Chapter 8: Onsets and nonmoraic syllables in German (Caroline Féry). Féry investigates the representation of word-final non-appendical consonants in German. She argues that such segments are best analysed as onsets of syllables lacking nuclei. Such an analysis allows for representations in which syllables are maximally bimoraic, thus doing away with the positing of superheavy syllables. It also explains the variable patterning of the laryngeals [h] and [?] as well as [g] following angma. The first section of the paper convincingly motivates the presence of word-final semisyllables, primarily based on asymmetries in the patterning of final consonants following tense versus lax vowels. The author then looks at asymmetries in the realization of the largyneals [h] and [?] as well as [g] following angma. It is argued that these segments fail to surface when syllabified as onsets to nonmoraic syllables.
Chapter 9: Extrasyllabic consonants and onset well-formedness (Antony Dubach Green). Focusing on position-based asymmetries in word-initial versus medial clusters in Icelandic, Attic Greek and Munster Irish, Green argues that clusters that are illicit syllable onsets may nonetheless be syllabified as onsets to the Ft and PWd. He proposes that the asymmetries in question are related to the interaction of a fixed hierarchy of Onset Wellformedness (OWF) constraints, which may take the syllable, Ft or PWd as arguments, and require a steep sonority rise between the two members of the cluster. The set of onsets possible in a given language results from the interaction of these constraints with Faithfulness constraints (MAX, DEP), as well as NoCODA, constraints on exhaustive syllabification, and a Syllable-contact-law-based constrained mitigating against sonority rises across syllable boundaries.
Chapter 10: Beyond codas: word and phrase-final alignment (Caroline R. Wiltshire). In this chapter, Wiltshire demonstrates how a unified alignment theory, in which the PWd and Prosodic Phrase (PPhr) can also serve as constraint arguments, allows for clearer generalizations concerning syllable typology, particularly in languages in which differences exist between the sets of syllable-final segments and the larger range of segments observed word and phrase-finally. The author demonstrates that languages that require phrases or words to end in consonants or vowels (Atampaya Uradhi; Leti; Yapese; and Pitjantjatjara respectively), all the while not imposing such constraints on syllables, involve grammars in which constraints requiring the alignment of the PPhr and PWd with consonant and vowels outrank correspondence constraints. Such an analysis also allows an interesting account of word and phrase-final consonants that violate otherwise well-respected coda constraints, often analysed as extraprosodic or extrametrical in derivational analyses. Wiltshire argues that both types of analyses are superior to previous, non- constraint-based analyses, in that they do not necessitate cyclical syllabification.
Chapter 11: On the sources of opacity in OT: Coda processes in German (Junko Itô and Armin Mester). This chapter addresses the question of how best to express opacity effects in an output-oriented framework like OT. Itô and Mester's answer is that opacity results from the existence of separate modules for word and phrasal phonology interacting serially, as well as the possibility of constraint conjunction; it is the latter aspect that constitute the main focus of the chapter. The phenomena of focus are coda devoicing, spirantization and cluster simplification in German. The authors argue that all three processes result from the conjunction of *COD, which bans any coda segment, and some segmental markedness constraint (ex. *VoiObs in the case of devoicing), when ranked above the relevant Faithfulness constraints. The interaction of the three processes involve opaque, bleeding or counterbleeding effects. In the remainder of the chapter, Itô and Mester demonstrate that neither transparent OT (i.e. using only standard Markedness and Faithfulness constraints) nor Sympathy theory can adequately account for such effects. Rather, they show that the same type of constraint conjunction necessary for the three coda-related processes is at the source of the opacity observed.
Chapter 12: Ambisyllabicity and fricative voicing in West Germanic dialects (Marc van Oostendorp). The primary focus of this chapter are position-sensitive fricative voicing asymmetries in Frisian, with subsequent comparison with patterns in Thurgovian German, Roermond Dutch, and Standard Dutch. Oostendorp proposes that voicing is not phonemic in fricatives in Frisian and that the asymmetries observed ~V voiceless in word-initial onsets and codas; intervocalically, voiceless following short vowels versus voiced after long vowels -- can be explained by the relative ranking of markedness constraints on voicing in fricatives and the syllabic affiliation of the segments. When syllabified as word-initial onsets or codas, the ranking of *FRICATIVE/voice and FINALDEVOICING over FAITH(Voice) results in voiceless fricatives in outputs. Intervocalic fricatives syllabified as codas, i.e. long geminate fricatives, too will be devoiced. In contrast, short intervocalic onsets syllabified solely as onsets are voiced in outputs given the ranking of VIF, a constraint requiring intervocalic fricative voicing, over *FRICATIVE/voice. As noted by the author, the paper makes two important contributions to phonological theory. First, as concerns OT in particular, Oostendorp argues that syllabification may be minimally specified underlyingly. Second, as concerns the representation of geminates in general, the analyses proposed argue for a bisegmental and not moraic representation of geminates.
Chapter 13: The CiV-generalization in Dutch: What petunia, mafia, and sovjet tell us about Dutch syllable structure (Ruben van de Vijver). In this paper, van de Vijver examines the CiV-generalization in Dutch, which states that, when preceding CiV sequences, lax vowels occur in closed syllables while tense vowels occur in open ones. Along with the canonical examples, two exceptional patterns are also analysed, exemplified by words such as mafia and sovjet, in which a lax vowel occurs in an open syllable. The general analysis proposed is that tense vowels are unmarked, with their presence being favoured by *LAX. Lax vowels surface only when specified as such in inputs, given the domination of MAX-LAX over *LAX, or when syllabified in closed syllables. This latter condition results from the presence of high-ranking CONNECT(LAX,BN), which requires "a lax vowel to appear in a branching nucleus, and when a nucleus branches, its vowel is lax". Van de Vijver argues that exceptional forms like mafia involve long consonants (here the /f/) in inputs, which surface as such in order to satisfy FAITH-CONSONANTAL LENTH. Such long consonants are syllabified ambisyllabically, thus closing the preceding syllable and necessitating a preceding lax vowel given the presence of CONNEXT(LAX,BN). Sovjet-type examples do not involve long consonants in their inputs, but rather vowels specified for [LAX]. The difference between mafia and sovjet-type forms as concerns the realization of the high front vocoid results from syllabification: in mafia, the segment is part of a complex onset headed by the second half of the geminate whereas the /v/ of sovjet is syllabified as a coda. Van de Vijver finishes the chapter by comparing his analysis with ones based on vowel-length contrasts or Morpheme Structure constraints.
Chapter 14: The relative harmony of /s+stop/ onsets: Obstruent clusters and the sonority sequencing principle (Frida Morelli). In this chapter, Morelli seeks to account for the cross-linguistic generalization that, when expanding upon core stop+sonorant and fricative+sonorant onset clusters, languages typically allow fricative+stop and not stop+fricative clusters. Given the general role proposed for sonority in determining onset wellformedness, one should expect the opposite. The author proposes that the relative unmarkedness of fricative+stop onsets with respect to other possible obstruent+obstruent clusters is related to their being the least marked vis-à-vis the feature [continuant]. Under the assumption that the constraint set contains two OCP constraints, OCP[-cont] and OCP [+cont], as well as the constraint SO (which disallows any tautosyllabic stop+obstruent sequence), FS are the least marked onsets given that the two members differ in terms of continuancy and that the stop is not initial. Within fricative+stop clusters, she argues coronal /s/+stop clusters are the least marked because place is marked in obstruents that are not released into sonorants, as release cues -- the primary perceptual cues to obstruent place -- are impoverished in this position. This is formalized using constraints based on Padgett (1995).
Chapter 15: The independent nature of phonotactic constraints: An alternative to syllable-based approaches (Juliette Blevins). This final chapter seeks to demonstrate that phonotactic constraints, which are most often defined with respect to syllables, are better defined in terms of phonetic strings. Blevins cites three types of evidence to support this claim: (i) divergences observed in many languages between syllable structure as determined with respect to phonotactics and syllabification necessary to account for metrical structure and native speaker intuitions; (ii) cross-linguistic phonotactic similarities in languages with different syllable structure; (iii) inviolable phonotactic constraints that defy syllable-based characterization. The author proposes that phonotactics be expressed with respect to constraints of the form "the set of feature values X is/is not licensed in string K", where K corresponds to a string of features, segments and word/morpheme boundaries. Blevins continues by investigating specific instances of such constraints as necessary for the explanation of laryngeal and place neutralization in a variety of languages. Such investigation includes the way in which such constraints -- if inviolable -- can explain emergent cross-linguistic implicational hierarchies. Discussion then moves to explanatory adequacy. Building on proposals in Steriade (1994,1997,1998,1999), the author argues that such constraints are phonetically motivated. Specifically, neutralization occurs in positions in which the acoustic/perceptual cues for a contrast are least salient. The chapter concludes with a discussion of two aspects of the general proposal for OT. As concerns the possibility of inviolable constraints, Blevins argues that the validity of the approach begs further investigation of whether subsystems of inviolable constraints should be introduced in OT. As concerns markedness, she proposes that, given positions of neutralization have been demonstrated to be directly related to phonetic conditioning environments in previous research, a first step towards building a theory of phonotactic markedness in OT will be to redefine syllable-based Markedness and Faithfulness constraints in string- based terms.
The fourteen papers included in this volume constitute a very important contribution both to research on the syllable and the development of OT. As concerns the syllable, a number of issues concerning representation (e.g. geminates, semisyllables) and processes (e.g. cyclicity, opacity) are addressed. The papers also raise many interesting questions re OT, including the nature of inputs and the possibility of inviolable constraints. The analyses proposed cover a large range of syllable phenomena and an impressive variety of languages, all the while not requiring more than a basic background in OT to follow the majority of analyses. This makes this publication of interest to a wide public.
Another strength of the research presented is the clear effort on the part of many authors to discuss previous analyses, including their weaknesses, and to demonstrate the ways in which OT analyses may contribute to solving such problems. As such, the articles provide not only a solid grounding in OT, but also a good training in the basics of solid hypothesis evaluation and phonological analysis; this increases the volume's pedagogical value.
A recurring and interesting aspect of many analyses is the demonstration that what have been proposed to be complex interactions of rules and constraints on surface forms in previous analyses result from the satisfaction of one or a few high-ranking constraints. For example, in Chapter 2, McCarthy demonstrates that trisyllabic deletion in Bedouin Arabic is directly motivated by the enforcement of quantity requirements on heads in an iambic situation. As the author states, such an analysis avoids positing complex, nonlocal, and highly arbitrary environments for rule application.
There are, however, two criticisms that can be levelled at the volume as a whole, both of which run counter to Féry and van de Vijver's claim in the introductory chapter that "OT is capable of providing answers to old issues that have been problematic in procedural analyses". The first criticism regards the level of explanatory adequacy of many of the constraints proposed. While some authors seek to motivate their constraints (e.g. Blevin's use of acoustic/perceptual cues), other constraints seem stipulated purely for the needs of the analysis in question. Indeed, many constraints are not validated with data from other varieties/languages or with broader typological or phonetic criteria. While the analyses proposed provide excellent descriptive coverage of the phenomena in question, if they are based on constraints lacking explanatory power, one must question the extent to which the OT analyses provided are superior to previous derivational ones. Note that the presence of such constraints does not invalidate the analyses. Rather, it necessitates further research into their motivation.
Second, it is not obvious that all proposals, including those involving representation (e.g. the possibility of semisyllables) require OT for their implementation. Indeed, some authors go as far as to state that their proposals are compatible with other frameworks (Wiltshire p.255; Blevins, p.376).
However, all in all, the volume provides stimulating discussion of the syllable in OT and should become a frequently-cited reference.
ABOUT THE REVIEWER:
ABOUT THE REVIEWER
Jeffrey Steele holds a PhD in Linguistics from McGill University and is currently Assistant Professor of French Linguistics at the University of Toronto. His research focuses on second language acquisition and linguistic theory, particularly the acquisition of prosodic structure, as well as Romance phonetics and phonology.