<<

[Re v i e w ] The in Optimality Theory

Ed. by Caroline Féry and Ruben van de Vijver, Cambridge University Press, Cambridge, 2008, ix+415pp.

Ma f u y u Ki ta h a r a Waseda University/MIT*

Keywords: Optimality Theory (OT), opacity, syllable, cumulativity, Evolutionary (EP)

1. Introduction According to the preface, this book is based on a conference held in 1998, more than a decade ago, on the same topic as its title. The book was published first in 2003 (reissued in 2008 as paperback) including papers by contributers other than the participants of the conference. Although many updates and more extended discussion following the conference are given, some of the theoretical devices and viewpoints are already outdated in the rapidly changing world of theoretical phonology. For example, Sympathy Theory is an incomplete and flawed framework that nobody looks back on (see McCarthy (2007) for details); comparative tableaux have become popu- lar to see the relative harmony of candidates at a glance (Prince (2002)); stochastic approaches to OT helped by simulation software are flourishing (Boersma and Pater (2008)); and so on. Some of the theoretical insights about the nature of OT presented in this collection of papers are still viable, though. Therefore, the purpose of this review is not to give a thorough introduction of each paper, but to try to shed light on fundamental issues and problems in theoretical phonology through the keywords and topics raised in this book. In my view, the fore- most and outstanding issue in OT since the late 1990’s is opacity: a hidden intermediate stage of derivation plays a role in determining the output. Two articles in this book directly address this topic: McCarthy’s “Sympathy, Cu-

* I would like to thank Haruka Fukazawa, John McCarthy, two anonymous reviewers, and the editors of EL for helpful comments, support, and suggestions. All errors are my own.

English 28: 1 (2011) 120–130 -120- © 2011 by the English Linguistic Society of Japan REVIEWS 121 mulativity, and the Duke-of-York Gambit” and Ito and Mester’s “On the Source of Opacity in OT: Coda Processes in German.” We will first con- centrate on these two articles to grasp the properties of the opacity problem in OT. Through the discussion of opacity, however, we will see that many theoretical assumptions about phonological representations, the architecture of the , and the nature of phonological processes are shared by other articles in the volume. Those will be briefly covered in this review when it is necessary. One of the contributers to this book, Juliette Blevins published a stimulat- ing and controversial book in 2004: Evolutionary Phonology, which includes an in-depth criticism of the Optimality-Theoretic framework. Her article in this volume entitled “The Independent Nature of Phonotactic Constraints: An Alternative to Syllable-Based Approaches” shares the core argument of EP. This is an interesting contrast within a single volume: OT vs. EP. We will cover this issue in a separate section in order to see the development of major theories in a larger perspective. There are 15 papers, each comprising a chapter, in this volume. Those other than the three mentioned above are: “Overview” by the editors; “The Controversy over Geminates and Syllable Weight” by Stuart Davis; “The Syllable as a Unit of Prosodic Organization in Japanese” by Haruo Kubozono; “Prosodic Weight” by Draga Zec; “ and Moras in Arabic” by Paul Kiparsky; “Semisyllables and Universal Syllabification” by Young-mee Yu Cho and Tracy Holloway King; “Onsets and Nonmoraic Syl- lables in German” by Caroline Féry; “Extrasyllabic Consonants and Onset Well-Formedness” by Antony Dubach Green; “Beyond Codas: Word and Phrase-Final Alignment” by Caroline R. Wiltshire; “Ambisyllabicity and Fricative Voicing in West Germanic Dialects” by Marc van Oostendorp; “The CiV-Generalization in Dutch” by Ruben van de Vijver; and “The Rela- tive Harmony of /s+stop/ Onsets” by Frida Morelli. As is obvious from the titles, recurrent topics up to Féry’s paper are syllable weight and semisyl- lables. From Dubach’s paper on, the topic gradually shifts toward syllable edges, syllable contact, and clusters. Those are common representational devices inherited from the derivational era. Thus, some of the papers are still concerned with OT-translation or OT-reanalysis of derivational rem- nants. However, the Al i g n family of constraints is so powerful that al- most all aspects of prosodic organization including syllables and moras are reorganized with some OT flavor. Morelli’s paper is worth-mentioning in this respect. One of the strong points of OT is its intrinsic orientation to typology. The typology of syllables involving /s+stop/ clusters successfully 122 ENGLISH LINGUISTICS, VOL. 28, NO. 1 (2011) shows the merits of the OT analysis.

2. Opacity 2.1. Revised Sympathy Theory Opacity in its original form in the derivational era was defined as in (1). (1) Opacity: Kiparsky (1973) A phonological rule ρ of the form A → B/C D is opaque if there are surface structures with any of the following characteris- tics: a. Instances of A in the environment C D b. Instances of B derived by ρ that occur in environments other than C D (1a) says that A appears even though it should have been changed to B, and (1b) says that B appears even though it should not. McCarthy (page 36 of the book under review) lays out why this is problematic for OT by using examples from Bedouin Arabic. (2) Bedouin Arabic derivation UR /katab/ /katab-at/ /kitib-at/ Gloss: ‘he wrote’ ‘she wrote’ ‘she was written’ Initial syllabifi- ka.tab ka.ta.bat ki.ti.bat cation: : I → φ/C σ kit.bat Trisyllabic V → φ/C Lσ kta.bat Deletion: Raising: a → i/C σ ki.tab kti.bat In this derivation, kti.bat is opaque with respect to Syncope, because it con- tains i in an open syllable. An interaction of stress and the above deriva- tion provides a challenge for Classic OT. The stress of the UR /katab-at/ is first assigned to the initial syllable as it is the head of the trochaic foot: (káta)bat. However, the Trisyllabic Deletion process in (2) deletes the stressed vowel resulting in a stress shift: ktíbat. Before getting into an actual tableau analysis, let us see the relevant con- straints. McCarthy proposes that, from an LL sequence, an iambic foot is achieved either by lengthening the second syllable (LH) or by deleting a mora of the first syllable (ΔL: Δ stands for a moraless semisyllable). The following Gr p Ha r m constraint formally handles this quantitative difference, where the weight of the second syllable y is greater than the first one x. REVIEWS 123

(3) Formulation of Grouping Harmony (Gr p Ha r m ) In an iambic foot (x ’y), |y| > |x|. (|α| ≡ weight of α in moras) The next relevant constraint is a relative faithfulness constraint which has a wide applicability over any phonological scale. The two versions in (4) provide that the input and the output are exactly the same or adjacent to each other. (4) Relative faithfulness on phonological scales a. St a y (S): Input and output have the same position on the scale S. b. St a y -Ad j (S): Input and output have the same or adjacent po- sitions on the scale S. These relative constraints are for handling Raising and Syncope pro- cesses in (2) by incorporating the duration scale: a > i > φ. This is a straightforward representation of the intrinsic duration of vowels. Thus, St a y (Dur(ation)) assigns one violation for an a → i change, and two viola- tions for an a → φ change, while St a y -Ad j (Dur) assigns one violation for an a → φ change. An antagonistic constraint against these relative faithfulness constraints is called Re d u c e , originally proposed by Kirchner (1996). (5) Re d u c e A short vowel in an open syllable has zero duration. Assign one violation mark for each increment of duration above φ on the scale Dur. Now, the tableau in (6) illustrates the problem. The opaque candidate (6a) is the desired output, but less harmonic than the transparent candidate (6c), indicated by a leftward pointing hand.

(6) Transparent candidate wins over opaque candidate

/katab-at/ Gr p Ha r m St a y -Ad j (Dur) Re d u c e St a y (Dur) Opaque a. (k.tí).bat * *! ** b. (ki.tí).bat *! ** ** Transparent c. (kát).bat * *

An early version of Sympathy Theory (McCarthy (1999)) can handle this situation by invoking intercandidate faithfulness relations. A sympathetic constraint Sy m (pathy) measures the faithfulness between a particular sym- pathetic candidate and all the other candidates. This sympathetic candidate is chosen by a “selector” constraint, notated by the symbol . As shown 124 ENGLISH LINGUISTICS, VOL. 28, NO. 1 (2011) in (7), the opaque, but correctly winning candidate (7a) is more similar to the sympathetic candidate (7b) than (7c) to (7b).

(7) Opaque candidate wins over transparent candidate via sympathy

/katab-at/ Gr p Ha r m Sy m St a y - Re d u c e St a y Ad j (Dur) (Dur) Opaque a. (k.tí).bat * * * ** Sympathetic b. (ki.tí).bat *! ** ** Transparent c. (kát).bat **! * *

A critical revision of the theory in McCarthy’s paper has arisen from the recognition that intercandidate faithfulness is too powerful a device. The Sy m constraint can directly measure the similarity on any dimension, which predicts the unattested type of opacity called feeding Duke-of-York (DY) derivations. DY derivations have the general form A → B → A, which looks like they are going forward and then backward without leaving any trace of change at the surface. McCarthy then proposes the notion of cumulativity, which indirectly mea- sures the similarity by the shared faithfulness violations. Let us take a look at the faithfulness part of the tableau in (7), recast as (8). The difference between the opaque (8a) and the transparent (8c) candidate is that the viola- tion marks for St a y (Dur) are shared by (8b) and (8a), but not by (8c) and (8a). Thus, the set of unfaithful marks for (8b) is a subset of those of (8a), which is an informal definition of Cumulativity.

(8) Shared faithfulness violations

/katab-at/ St a y -Ad j (Dur) St a y (Dur) Opaque a. (k.tí).bat * ** Sympathetic b. (ki.tí).bat ** Transparent c. (kát).bat * *

Cu m u l which watches the cumulativity relations to the sympathetic candidate now supersedes the Sy m constraint in (7). There is only a tiny difference between (9) and (7): the number of violation marks for the flower constraint. Cu m u l assigns one violation mark for (9c) due to its noncu- mulative unfaithful mapping to (9b). However, the restrictiveness of the theory is greatly improved, and moreover, the idea of cumulativity is an im- REVIEWS 125 portant step toward a later development of the theory, the notion Candidate Chain.

(9) Opaque candidate wins over transparent candidate via cumulativity

/katab-at/ Gr p Ha r m Cu- St a y - Re d u c e St a y m u l Ad j (Dur) (Dur) Opaque a. (k.tí).bat * * ** Sympathetic b. (ki.tí).bat *! ** ** Transparent c. (kát).bat *! * *

Essentially, a candidate chain is a series of forms with cumulative faith- fulness violations. A chain starts from the input and moves step-by-step upward on a harmonic ascent. OT with such chains is called OT-CC (McCarthy (2007)), which is one of the most advanced solutions to the problem of opacity. At this point, it is necessary to mention why Sympathy Theory is not pur- sued anymore. One of the serious objections is that there are cases where multiple sympathetic relations are necessary for the analysis, which requires a very unrestricted theory as well as the loss of generalization (Kiparsky (2000)). Although Kiparsky’s paper in this volume does not bring the sym- pathy issue to the foreground, his argument for the Lexical Phonology and combined with OT (LPM-OT) raises a fundamental question about the fully parallel architecture of Classic OT.

2.2. Local Conjunction Ito and Mester’s paper has a radically different approach to opacity from McCarthy’s. They first cast doubt on the validity of the rule-based defini- tion given in (1) in constraint-based approaches. (10) Quote from p. 272 It is a priori quite unlikely that two radically different theoreti- cal paradigms like rule-based sequentialism and constraint-based parallelism would have mechanisms corresponding to each other in such a direct way, with ordered rules applying in a multistage derivation directly matched by sympathetic faithfulness to a spe- cially selected candidate that fulfills the role of the abstract deri- vational stage. The problem lies not in the abstractness but in the very direct one-to-one matching of mechanisms. In other words, they redefine the problem of opacity in OT as seeking 126 ENGLISH LINGUISTICS, VOL. 28, NO. 1 (2011) the source, not the solution of the cases identified in the traditional sense of opacity. This is in striking contrast to McCarthy’s paper reviewed in the previous section. His analyses have consistently focused on a better and more general solution of the opacity problem. Ito and Mester’s focus is on “Which elements of current OT architecture can in principle give rise to opaque output patterns? (p. 272)” The core data of their analysis come from two varieties of German: Stan- dard German (SG) and Colloquial Northern German (CNG). Cluster sim- plification and devoicing take place in CNG and the correct outcome forSG is also easily obtained by reversing the order of the two processes, as shown in (11) and (12). (11) CNG /dɪŋɡ/ /dɪŋɡ+ə/ Gloss: ‘thing’ ‘things’ Devoicing: [−son] → [−voi] / ]σ dɪŋk — Cluster simplification: g → φ/ [+nas] ]σ — dɪŋə (12) SG /dɪŋɡ/ /dɪŋɡ+ə/ Cluster simplification: g → φ/ [+nas] ]σ dɪŋ dɪŋə Devoicing: [−son] → [−voi] / ]σ — — However, an integrated OT analysis of these two varieties cannot be as simple as rule reordering. Based on independent grounds (Ito and Mester (1998)), they advocate decomposing complex constraints into local conjunc- tions of simplex constraints. In order to militate against voiced coda, a conjunction of two markedness constraints is necessary. In (13), *Vo i Ob s is a simplex constraint that bans a voiced obstruent and *Co d is the familiar NoCo d a constraint. As for cluster simplification in this case, a conjunction of four markedness constraints is proposed, where *Do r s Pl o s i v e prohibits a dorsal plosive and *Co m p l e x disfavors a cluster. (13) Conjoined markedness constraint for devoicing *VC: [*Vo i Ob s & *Co d ] (14) Conjoined markedness constraint for cluster simplification *VCDC: [*Vo i Ob s & *Co d & *Do r s Pl o s i v e & *Co m p l e x ] With these constraints, a successful analysis of CNG is readily available as in (15). A reranking of Ma x and Id e n t (voi) seems to work for the de- sired output in SG as well. However, it is clear from an independent set of data that reranking is impossible: the devoicing of /ta:g/ ‘day’ → [ta:k], for example, requires that the Id e n t (voi) is ranked lower than the Ma x of the coda consonant in both SG and CNG. REVIEWS 127

(15) An OT analysis of CNG and SG

/dɪŋɡ/ *VCDC *VC Ma x Id e n t (voi) a. /dɪŋɡ/ *! * b. CNG /dɪŋk/ * c. SG /dɪŋ/ *!

Ito and Mester claim that this output pattern is a case of opacity caused by constraint conjunction, which is independent from the traditional defini- tion of opacity.1 It is shown that not only Classic OT but also Sympathy Theory fails to account for this pattern. Their solution also lies in the con- junction of not just markedness constraints but faithfulness constraints. Ito and Mester’s main point is that opacity is not a monolithic entity in phonology. Some opaque patterns arise from traditional counter-bleeding relations, but others arise from constraint conjunction. Therefore, Sympathy Theory cannot be the one and only solution for opacity.

3. String-based Phonotactics Blevins has been one of the major researchers of the syllable (Blevins (1995)). As mentioned in Introduction, however, her recent work Evo- lutionary Phonology covers a much wider scope than syllable-related is- sues. Evolution, in her terms, is not just an analogy taken from biological evolution, but a direct pointer to diachronic changes of . Her article in this volume focuses on non-syllable-based phonotactics, which is quite contrastive to the other articles in the first place. Most au-

1 An anonymous reviewer pointed out that Ito and Mester should have provided an al- ternative definition of opacity which can subsume the problem raised by their data in(11) and (12). I agree that their use of the term “opacity” deviates from the traditional defi- nition in (1). However, they also use the term “opaque interactions” elsewhere, which implies that two (or potentially more) phonological processes interact in a non-transparent way: cluster simplification and devoicing in the case of (15). In OT terms, such phono- logical processes are modeled as partial rankings: *VC » Ma x » Id e n t (voi) and *VCDC » Ma x , respectively. When these partial rankings are integrated into a single overall ranking, the two processes interact to result in a problematic output (15c) for SG. Their use of “opacity” thus implies such opaque interactions. They, then, attribute the source (and the solution) of this problem to the notion of constraint conjunction. Recent theo- retical developments of constraint conjunctions can be found in Farris-Trimble (2008). 128 ENGLISH LINGUISTICS, VOL. 28, NO. 1 (2011) thors do not even argue for the usefulness of syllable in phonotactics: it is already a basic assumption in the dominant paradigm. Her argument goes as follows. (i) syllable-based phonotactic statements are not valid in many cases, (ii) languages with different syllabification patterns show a conver- gence of phonotactic constraints, and (iii) emergent phonotactic universals become visible only when a non-syllble-based view is put forth. These three domains serve as evidence for her proposal of string-based phonotac- tics. A handy example of string-based phonotactics is the constraint *CCC, banning a triconsonantal cluster anywhere. According to Blevins, *Co m - p l e x at the onset and/or the coda is a common syllable-based phonotactic restriction aiming at the same effect. However, there are languages where initial and/or final CC clusters are found but no medial CCC cluster is allowed (Lenakel and Leti). Along with other evidence from typologi- cally and chronologically diverse languages, Blevins goes one-step further: “In what way do string-based phonotactics themselves have explanatory value?” Her answer is that there are phonetic origins for string-based pho- notactics which are sometimes hidden in the process of diachronic chang- es. This is a very clear statement of the inception of EP. She further argues that these phonotactic constraints are inviolable, which conflicts with a fundamental tenet of OT. Criticizing the whole OT framework appears to be outside the scope of her paper, but we can clearly see that a departure has begun at this point.

4. Concluding Remarks As concluding remarks, let me give a critical evaluation of the book in a Question and Answer format. Q1. Is it worth spending time/energy on this book from the current theoretical perspective? A1. Yes, but somewhat ironically so. This book corresponds to an intermediate stage of derivation in a typical opacity case: we need that stage, though it is hidden, in order for later stages to be connected to the initial stage. The analogy here is that the initial stage corresponds to Classic OT, while the later stage is comparable to the current OT-CC. We do not see Sympathy Theory any more today, but the idea of cumulativity has survived and connects Classic OT with OT-CC. So, this book surely helps us to understand the origin and the growth of a theoretical REVIEWS 129

device in a larger perspective. Q2. Is this a handy reference for the syllable literature? A2. Again, yes, but with some reservations. Most of the authors are currently active in the field, and some of their ideas, core observations, and analyses date back to the articles in this volume. Readers who want to trace the development of OT analyses of the syllable will get sufficient information from this book. However, it is not a comprehensive volume with a planned structure and coverage of major topics about the syl- lable. Q3. What is left to be done and what is the current focus in syllable related research? A3. The recent trend in phonetics is to view the syllable as a pho- netic domain of coordination of articulatory gestures (Krakow (1999)). The basic idea is that consonants are coordinated with respect to the vocalic syllable nucleus, and vowels provide the basis for the articulation of consonants. At the most re- cent conference on Laboratory Phonology 12 (LabPhon12, July, 2010: University of New Mexico, Albuquerque), this phonetic notion of the syllable was confirmed on multiple independent grounds. Sophisticated laboratory techniques, such as electro- magnetic articulography (EMA), ultrasound images, and optical tracking show a fine-tuned organization of the phonetic syllable in human languages. In my opinion, the integration of the pho- netic syllable and the phonological syllable is still a task for future research. At least some versions of the OT framework, those with stochastic components may readily accomodate the integration.

REFERENCES Blevins, Juliette (1995) “The Syllable in Phonological Theory,” The Handbook of Phonological Theory, ed. by John A. Goldsmith, 206–244, Blackwell, Cam- bridge, MA. Blevins, Juliette (2004) Evolutionary Phonology: The Emergence of Sound Patterns, Cambridge University Press, Cambridge. Boersma, Paul and Joe Pater (2008) “Convergence Properties of a Gradual Learning Algorithm for Harmonic Grammar,” ms., University of Amsterdam and Univer- sity of Massachusettes, Amherst [ROA < http://roa.rutgers.edu/ > 970]. 130 ENGLISH LINGUISTICS, VOL. 28, NO. 1 (2011)

Farris-Trimble, Ashley W. (2008) “Cumulative Faithfulness Effects in Phonology,” Doctoral dissertation, Indiana University [ROA 991]. Itô, Junko and Armin Mester (1998) “Markedness and Word Structure: OCP Effects in Japanese,” ms., University of California, Santa Cruz [ROA 255]. Kiparsky, Paul (1973) “Abstractness, Opacity and Global Rules,” Three Dimensions of Linguistics Theory, ed. by Osamu Fujimura, 57–86, TEC, Tokyo. Kiparsky, Paul (2000) “Opacity and Cyclicity,” The Linguistic Review 17, 351–367. Kirchner, Robert (1996) “Synchronic Chain Shifts in Optimality Theory,” Linguistic Inquiry 27, 341–351. Krakow, Rena A. (1999) “Physiological Organization of Syllables: A Review,” Jour- nal of Phonetics 27, 23–54. McCarthy, John J. (1999) “Sympathy and ,” Phonology 16, 331–399. McCarthy, John J. (2007) Hidden Generalizations: Phonological Opacity in Optimal- ity Theory, Equinox, London. Prince, Alan (2002) “Arguing Optimality,” University of Massachusetts Occasional Papers in Linguistics 26, ed. by Angela Carpenter, Andries Coetzee and Paul de Lacy, 269–304, GLSA, Amherst, MA.

[received July 21 2010, revised and accepted November 25 2010]

School of Law Waseda University 1–6–1 Nishiwaseda, Shinjuku-Ku Tokyo 169–8050 e-mail: [email protected]