McCarthy, John J., ed. (2003) Optimality Theory in Phonology: A
Reader, Blackwell Publishing.
Announced at http://linguistlist.org/issues/14/14-2539.html
Andrew Carstairs-McCarthy, University of Canterbury, New Zealand
DESCRIPTION
This book contains 33 chapters on applications of Optimality Theory
(OT) to phonology, arranged in five parts: I 'The Basics', II 'Formal
Analysis', III 'Prosody', IV 'Segmental Phonology', and V
'Interfaces'. The chapters are excerpts from material much of which
has never appeared in print but merely been stored at the Rutgers
Optimality Archive. The first section contains 70 pages from the
foundation document of OT, 'Optimality Theory: Constraint Interaction
in Generative Grammar' by Prince and Smolensky (1993). Most other
chapters, however, have been pared down to 10 or 20 pages. For
example, McCarthy and Prince's article 'The emergence of the unmarked'
(1994), 46 pages long in the original, is represented here by a
thirteen- page excerpt, and their 74-page article 'Generalized
alignment' (1993) is represented by three excerpts in parts I, III and
V, totalling 29 pages.
Each chapter begins with an 'Editor's Note' by McCarthy, putting it in
context, and ends with a set of 'Study and Research Questions',
ranging from straightforward applications of ideas in the chapter to
cutting-edge questions that could be the topic of a PhD thesis. In
the bibliography for each chapter, items from which an excerpt is
included in this book are helpfully identified. There is also a
bibliography of items cited in the editor's notes and research
questions.
Part I contains, apart from the Prince-Smolensky and 'Generalized
alignment' excerpts, another chapter by McCarthy and Prince on
faithfulness and identity. Part II contains three somewhat technical
papers by Smolensky, Bruce Tesar, and Elliot Moreton. In Part III, as
well as another 'Generalized alignment' excerpt, are papers on prosody
by Nine Elenbaas, René Kager, Michael Kenstowicz, Eric Bakovic, John
Alderete, Moira Yip and Scott Myers. Part IV contains chapters on
assimilation, dissimilation and the notion 'positional faithfulness'
by Joe Pater,Bruce Hayes, Jill Beckman, Linda Lombardi, Cheryl Zoll,
Jay Padgett, John Alderete, and Robert Kirchner. Part V covers a
variety of issues relating to morphology, lexical exceptions, clitics,
and redundancy and underspecification, with contributions by Laura
Benua, Jerzy Rubach, Elisabeth Selkirk, Paul de Lacy, Joan Mascaró,
Anna Lubowicz, Junko Itô, Armin Mester, Jaye Padgett, Sharon
Inkelas, Orhan Orgun, Cheryl Zoll, Arto Anttila, Young-mee Yu Cho, and
(again) McCarthy and Prince.
EVALUATION
This is the book we have all been waiting for. By 'we' I mean
everybody who is not a cutting-edge researcher in OT phonology but who
hopes to become one, or who needs to know about OT in order to teach
phonology in an up-to-date fashion, or who is interested in what has
been perhaps the most vigorous and fruitful trend in generative
linguistic theory in the last decade. I have already mentioned the
pruning that most of the material reproduced here has undergone. This
has made it possible to cover a wide range of topics with minimal
repetition. I will discuss first the choice of topics, secondly (from
a morphologist's point of view) the state of play in phonological OT
as revealed by this book.
The emphasis on phonology rather than morphology means that the
chapter taken from Prince and Smolensky (1993) does not include what
is perhaps the most famous of all OT analyses: the interplay between
NoCoda and Edgemost that determines whether the Tagalog affix 'um-'
gets prefixed or infixed. But plenty of space is devoted to Prince
and Smolensky's other seminal analysis, of fascinating data from
Imdlawn Tashlhiyt Berber: any sound whatever is potentially a syllable
nucleus, but suitably ranked OT constraints provide accurately for
nearly every word a unique syllabification. Part II may be heavy
going for beginners reading the book from cover to cover; but McCarthy
evidently feels that something should be said early on about a prima
facie flaw in OT, namely the open-endedness of candidate sets and its
implications for learnability.
Students are usually introduced to segmental phonology first and
suprasegmental (or metrical or prosodic) phonology second. So why
does McCarthy place part III 'Prosody' before and part IV 'Segmental
Phonology'? As soon as one reads them a reason emerges. Within
'pure' phonology (i.e. ignoring morphophonological phenomena such as
reduplication and infix-prefix alternations), it is in the domain of
prosody that OT has scored its most striking successes. In metrical
phonology before OT, it was conventional to say that feet (the units
of metrical structure between the syllable and the phonological word)
can be 'bounded' or 'unbounded', and that bounded feet can be binary
(consisting of two syllables) or, less often, ternary (with three
syllables). In other words, it was conventional to think in terms of
parameter-setting. It hardly seemed likely that ternary rhythms could
be somehow a byproduct of a requirement that all feet should be
binary. But in chapter 8, Elenbaas and Kager show precisely this.
They derive ternary rhythms through combining foot binarity with three
independently motivated constraints, suitably ranked: *Lapse ('Every
weak beat must be adjacent to a strong beat or the word edge'),
All-Ft- Left ('The left edge of every foot coincides with the left
edge of some prosodic word') and Parse-Syllable ('All syllables must
be parsed by [i.e. contained in] feet'). If a satisfying analysis is
one that derives unexpected but accurate conclusions from simple
assumptions, this is one of the most satisfying analyses that I have
ever come across in linguistics, in the same league as Karl Verner's
article on the law that bears his name.
In parts IV and V, if readers are forced to pick winners, they will
come up with a wide range of choices, because there is so much good
material. My own choice is for chapters 20 by Alderete and 27 by de
Lacy. Alderete discusses dissimilation, or the avoidance of similar
items in close proximity, and shows how it can be handled by 'local
conjunction': even if constraint B is ranked below constraint A, a
double violation of constraint B within some local domain may be
ranked above A. Thus, if constraint B is 'avoid labials' and
constraint A is 'be faithful', a single labial will surface
unmolested, but where faithfulness would yield two labials in close
proximity, one will be replaced by another consonant, such as a
coronal. De Lacy discusses the old chestnut of Maori passive
formation, where it seems that the neat phonological solution
(underlying stem-final consonants and a uniform suffix '-ia') does not
reflect how Maori speakers' competence actually handles these forms.
De Lacy exploits a hitherto unnoticed problem with the 'neat'
solution: it does not explain why stems with an underlying final
consonant are so limited in their prosodic characteristics. His
solution relates passive formation in an intriguing way to new
observations about maximal prosodic words in Maori. There are some
rough edges here, but de Lacy nevertheless demonstrates what can
emerge from bringing a wider sampling of data to bear on a hackneyed
problem.
What are the weaknesses of OT phonology, then, as revealed in this
collection? To my mind, as a morphologist, the main weakness is one
that has beset generative phonology from the outset: an assumption
that, if two alternants of a morpheme are phonologically similar, they
must have a single shared underlying phonological representation (or a
single input, in OT terms). Everyone now recognises that no common
underlying representation can handle ablaut sets such as 'sing, sang,
sung, song', unless one invokes rules or constraints that are
intolerably ad hoc. But this recognition is grudging. Therefore, in
discussing morphological or morphophonological alternations in
languages less well studied than English, OT phonologists still
struggle for a single-input analysis, no matter what the cost in terms
of weird constraints or constraint conjunctions. To have more than
one phonological input for a single 'morpheme' is seen as necessarily
involving lexical idiosyncrasy (as in Mascaró's chapter 28 on
'external allomorphy'). That is not correct, however. Many
''morphemes'' (scare quotes!) can exhibit two or more phonologically
similar but distinct inputs with the same morpholexical relationship
between them. This was argued years ago by Jackendoff (1975) and
Lieber (1981), and has been illustrated more recently by
Cameron-Faulkner and Carstairs-McCarthy (2000) in relation to the
inflection class behavior of Polish nouns.
Oddly enough, one of OT's most promising notions, namely 'positional
faithfulness', ought to encourage OT phonologists to question their
single-input bias. According to positional faithfulness, markedness
constraints are less likely to enforce deviations from the input in
'strong' positions (such as roots, initial syllables, stressed
syllables and onsets) than in 'weak' positions (such as affixes,
noninitial syllables, unstressed syllables and codas). If that is
correct, then positions that are strong by several criteria, such as
word-initial stressed syllables belonging to roots, should participate
in morphophonological alternations only rarely. But this seems hardly
correct. Any linguist will have no difficulty in finding examples of
morphophonological alternations in strong positions, such as ablaut in
Indo- European generally, umlaut in German in particular, and onset
mutations in Celtic and Fula. Does this mean, then, that the
positional faithfulness idea (neatly exploited in chapter 16 by
Beckman, for example) is flawed? Not necessarily. What it may mean,
however, is that phonologists should be more ready than they are to
entertain the possibility that some morphophonological alternations
(particularly ones with nonphonological triggers) involve more than
one input.
This could help in Guugu Yimidhirr, discussed by Zoll (chapter 18).
Here, a certain class of suffixes requires a long vowel in the
preceding syllable, even when that syllable is in a strong position
(in a 'head prosodic word'). One might think that this is because
positional faithfulness is outranked here by a parochial lengthening
constraint associated with these suffixes in particular; but Zoll
shows that this will not work. However, her own solution involves
abandoning the restrictiveness of positional faithfulness while still
relying on this odd-looking and hardly universal vowel-length
constraint. Yet the clearly morphological basis of this phenomenon
should alert us: maybe more than one input representation is involved.
Let us suppose that every word in Guugu Yimidhirr has at least one
input alternant with a long vowel in the final syllable (and this
alternant may be the only one). That is hardly a difficult
generalisation for a child to learn, involving no lexical
idiosyncrasies. One need merely say then that the affixes in question
are subcategorised to appear with this alternant. Precisely how one
will handle that in OT morphology does not matter for present
purposes. What matters is that exploiting the independently
demonstrated possibility of multiple inputs in contexts where
morphology interferes with phonology means that Guugu Yimidhirr is no
longer problematic for the purely phonological notion of positional
faithfulness.
I have focussed here on an issue that intrigues me as a morphologist.
But this is an extraordinarily rich collection of material, sure to
stimulate all readers, whatever angle they approach it from. We
should be grateful to John J. McCarthy for compiling it.
REFERENCES
Cameron-Faulkner, Thea and Carstairs-McCarthy, Andrew. 2000. Stem
alternants as morphological signata: evidence from blur avoidance in
Polish nouns. Natural Language and Linguistic Theory 18:813-35
Jackendoff, Ray. 1975. Morphological and semantic regularities in
the lexicon. Language 51: 639-71.
Lieber, Rochelle. 1981. On the Organization of the Lexicon. (MIT
PhD dissertation.) Bloomington: Indiana University Linguistics Club.
McCarthy, John J. and Prince, Alan. 1993. Generalized alignment. In
Yearbook of Morphology 1993, ed. Geert Booij and Jaap van Marle,
79-153. Dordrecht: Kluwer.
McCarthy, John J. and Prince, Alan. 1994. The emergence of the
unmarked: optimality in prosodic morphology. In Proceedings of the
North East Linguistic Society 24, ed. Mercè Gonzàlez, 33-79.
Amherst, MA: GLSA Publications.
Prince, Alan and Smolensky, Paul. 1993. Optimality Theory:
Constraint Interaction in Generative Grammar. New Brunswick, NJ:
Rutgers University Center for Cognitive Science, RuCCS TR-2.
|
