LINGUIST List 10.1933
Tue Dec 14 1999
Review: Kager et al.: Prosody-Morphology Interface
Editor for this issue: Andrew Carnie <carnielinguistlist.org>
What follows is another discussion note contributed to our Book Discussion
Forum. We expect these discussions to be informal and interactive; and
the author of the book discussed is cordially invited to join in.
If you are interested in leading a book discussion, look for books
announced on LINGUIST as "available for discussion." (This means that
the publisher has sent us a review copy.) Then contact Andrew Carnie at
carnie
linguistlist.org
Directory
mike_maxwell, review of "The Prosody-Morphology Interface"
Message 1: review of "The Prosody-Morphology Interface"
Date: Mon, 29 Nov 1999 15:42:43 -0500
From: mike_maxwell <mike_maxwellsil.org>
Subject: review of "The Prosody-Morphology Interface"
Kager, Rene; van der Hulst, Harry; and Zonneveld, Wim (eds.). 1999.
The Prosody-Morphology Interface. Cambridge: Cambridge University
Press. 442 Pages. $69.95 (hardback).
Reviewed by Mike Maxwell, Summer Institute of Linguistics.
This book contains ten of the sixteen papers presented at a workshop
on prosodic morphology in 1994 at Utrecht University. The theory of
prosodic morphology was originally developed in papers by McCarthy
and Prince, particularly McCarthy and Prince (1990). As it happened,
the authors soon rendered their own theory obsolescent by their new
approach to phonology, Optimality Theory (henceforth OT). This book
looks at the impact of OT on the theory of prosodic morphology.
The work begins with a brief introduction by Rene Kager and Wim
Zonneveld to the history of phonology (starting with SPE, Chomsky
and Halle 1968), the (short) history of prosodic morphology, an
introduction to OT, and finally a synopsis of the other papers.
I suspect that most readers not already familiar with the theories
will find Kager and Zonneveld's necessarily brief overview of
phonology and morphology confusing. For example, it will be unclear
why an SPE-type analysis requires the use of a diacritic feature for
vowel harmony in Igbo (page 3). (Answer: it doesn't, although this
is indeed how Chomsky and Halle analyzed it. But this particular
use of diacritic features was abandoned shortly thereafter.)
Another confusion is introduced by the statement that in Arabic,
"consonants play the role of stems (or binyanim)" (page 5). In
fact, a binyan is not a set of consonants, but more like a slot in a
paradigm, combining a particular meaning with a particular templatic
form.
The remaining chapters consist of the papers themselves, which I
will summarize and comment on in the following paragraphs. I have
written "ROA" after the title where the paper is available for
download from the Rutgers Optimality Archive
(http://ruccs.rutgers.edu/ROA/), although the version there might be
different.
- -------------------------------------
Stuart Davis "On the moraic representation of underlying geminates:
evidence from prosodic morphology": Under one version of moraic
phonology, geminates are represented as moraic consonants. In a
language without a rule of "weight-by-position" (which would also
assign a mora to a consonant in a syllable coda), syllables closed
by a geminate consonant should then pattern with syllables
containing a long vowel (because both of these syllable types would
have two moras), and unlike open syllables or syllables closed by a
non-geminate consonant. The existence of such languages has been
doubted. Davis uses evidence from phonology and prosodic morphology
to argue that such languages do exist. However, the morphological
evidence seems questionable, given the non-derivational assumptions
of OT. Davis' analysis of CVC syllables in Hausa as monomoraic, for
example, depends on a derivational analysis with ordered rules,
incompatible with a pure OT analysis. (See Davis' footnote six.)
There is a typo in example (4e) (page 40): the stressed vowel should
be long (a short vowel here would be unstressed).
Laura Downing "Verbal reduplication in three Bantu languages":
Downing contrasts prefixing (and even infixing) reduplication in
three Bantu languages. As usual in the OT literature, the
differences are attributed largely to the relative ranking of the
constraints governing the form of the reduplicant. But unlike
earlier analyses, the author argues that (for two of the languages)
the template for reduplication must be defined in terms of the
"canonical verb stem", a definition which is not purely prosodic
(prosodic constituents include the mora, syllable, foot, and
prosodic word). I might also add that while proponents of OT have
often claimed that the constraints which the theory manipulates are
universal, actual analyses often contain constraints that are
clearly non-universal. Downing's analysis is no exception, for
example (her (37) page 83, for Kinande):
At least one stem vowel (excluding the final vowel) of the
morphological stem must have a correspondent in the prosodic filter.
Larry Hyman and Al Mtenje "Prosodic morphology and tone: the case of
Chichewa": This is not an article for the faint-hearted (or the
reviewer on a deadline!), for the depth of detail, dialectal
variation, and alternative analyses which the authors examine (or
which the reader may imagine) make for a hard read. Hyman and
Mtenje argue that a level-ordered (derivational) analysis fails to
explain tone in reduplicated forms in Chichewa (Bantu), and that an
OT analysis is superior. But we have not heard the last word, for
the OT analysis has difficulties, too: they admit (in footnote 26)
that in at least one case, a derivational analysis achieves
observational adequacy where the OT analysis does not, and elsewhere
(page 117 and footnote 29) allude to the need for "repair
strategies" a la Paradis (1988), something which is outside the
scope of 'true' OT.
Sharon Inkelas "Exceptional stress-attracting suffixes in Turkish:
representations versus the grammar" (ROA): A common assumption is
that metrification is not stored lexically. Morphemes which are
exceptions to the usual stress patterns of the language must then be
treated by morpheme-specific constraints (MSC) in the grammar (i.e.
phonology). Inkelas argues that marking exceptional metrical
structure in the lexicon is explanatorily superior to the MSC
solution, which in turn has implications for the use of prosodic
templates in reduplication (see McCarthy and Prince article below).
I might add that the MSC solution is vague, in that it is not clear
how a constraint specifies a morpheme: does it point to the lexical
entry, or to an underlying form? The answer to this question has
important implications for A-Morphous morphology (Anderson 1992),
the Separation Hypothesis (Beard 1995), and suppletive allomorphy.
Junko Ito and Armin Mester "Realignment": In a paper devoted largely
to phonology, Ito and Mester propose that syllabification should be
seen as the result of constraints on alignment. As in much work in
OT, this approach calls for a family of very general constraints,
e.g. "Align a consonant with the left boundary of a syllable".
Individual languages rank more specific forms of these constraints,
e.g. "Align an obstruent...".
One of the claims of OT is that constraints are universal, and that
languages differ only in the ranking of those universal constraints.
But I fail to see a substantial difference in universality between a
family of constraints represented universally by a general
constraint (as in Ito and Mester's proposal), and a rule-schema such
as "spread a single feature node" in the later versions of
derivational autosegmental phonology.
Another issue Ito and Mester bring up in passing is the function of
constraints: "We hypothesize that segment-alignment constraints are
related to a more fundamental requirement: segments should be
prominent" (page 199). But it is unclear how this functionalist
approach can be reconciled with the idea that the constraints are
innate.
John McCarthy and Alan Prince "Faithfulness and identity in prosodic
morphology" (ROA): This is by far the longest paper in this
collection, and best read first, since many of the other papers are
reactions to it. McCarthy and Prince begin by revising their
earlier theory of prosodic morphology, in which the content of e.g.
reduplicative morphemes was characterized in prosodic terms (as a
syllable, foot, minimal word etc.), in favor of their
characterization as affix, root, stem or word; the prosodic form
then follows (indirectly) from constraints on the shape of those
morphemes or words. (See Inkelas' article above for one reaction to
this.)
The authors then turn to languages with reduplicative morphology in
which a phonological process appears to apply to both base and the
reduplicant, even though from a purely phonological viewpoint it
would be expected to apply only to one or the other. Their
analysis, based on a constraint calling for identity between Base
and Reduplicant (B/R identity, parallel to the I/O faithfulness
constraint holding between Input and Output forms), seems strong
evidence in favor of OT and against a derivational account. (An
earlier derivational analysis, Wilbur 1973, relied on an otherwise
unattested notion of identity between a constituent and its copy.
It might however be interesting to seek a connection between Wibur's
approach and later work on geminate inalterability in derivational
phonology.)
An interesting implication of the OT analysis in which B/R identity
may be ranked below (less important than) I/O faithfulness, is that
other phonological constraints may be ranked in between these. In
particular, constraints on prosodic shape of affixes, roots etc. may
be ranked lower than I/O faithfulness. With "ordinary" affixation,
such constraints would not be evident, being masked by the I/O
identity. But since reduplicants do not have any input, these
constraints manifest themselves in reduplication. McCarthy and
Prince refer to the unmasking of such constraints in reduplicated
forms as "the emergence of the unmarked." (There is an un-addressed
question here: if the ranking of these constraints is visible only
in reduplicative constructions, is that ranking more difficult to
learn?)
Joe Pater "Austronesian nasal substitution and other NC effects"
(the "C" has the IPA symbol for voicelessness under it; for
typological reasons, I use "C" here to mean "voiceless consonant")
(ROA): In another paper largely about phonology, Pater argues that
while languages get rid of sequences of Nasal+C sequences in a
variety of ways (fusing the N+C into a nasal with point of
articulation of the C, or voicing the C, or deleting the N, etc.),
all these effects are the result of a single universal constraint
against a nasal followed by a voiceless consonant, a constraint he
characterizes as "hugging the phonetic ground." (A non-OT analysis
might be that phonological processes doing away with N+C sequences
arise diachronically for phonetic reasons, and that what can be
explained functionally does not need to be explained by a universal
constraint.)
In some languages, there is a complication: N+C fuse into N only at
morpheme boundaries (a derived environment effect). Pater accounts
for this by splitting one of the constraints into a root-internal
constraint and a general constraint, with the former more highly
ranked. (Is this a stipulation, or an explanation? Do all
constraints come in two such versions?)
Sam Rosenthall "The prosodic base of the Hausa plural": This paper
explores the formation of sound and broken plurals in several Hausa
noun classes. (In Hausa "broken plurals", part of the stem appears
inside the plural suffix, or vice versa.) The analysis refers
explicitly to various iambic templates (H, HH, LH, where L = Light
syllable, and H = Heavy), as in the version of prosodic morphology
developed in McCarthy and Prince (1990). It might prove challenging
to reformulate Rosenthall's insights in terms of the version of
prosodic morphology in McCarthy and Prince's article in this book
(see above). Unfortunately, this point is not expanded on.
Grazyna Rowicka "Prosodic optimality and prefixation in Polish":
Rowicka gives a clever solution to the much-discussed problem of
"yers": vowels (in Slavic languages) whose appearance is governed by
the presence of a following (non-appearing) yer. The solution
relies on the possibility that in a single language (and even a
single word), there may be both syllable-based and mora-based feet,
a controversial claim. The connection with prosodic morphology lies
in the idea that in Polish, prefixes are not normally included in
the prosodic word, but may be included to satisfy certain
constraints.
Suzanne Urbanczyk: "Double reduplications in parallel" (ROA):
Lushootseed (Salishan) has several reduplicative affixes; two such
affixes can attach to a single word, and Broselow (1983) had used
the properties of such "double reduplications" to support cyclic
rule application. Urbanczyk argues that in the context of OT, the
data provide no evidence for cyclicity. Her analysis relies on
morpheme-specific constraints (one morpheme obeys a faithfulness
constraint, while the other violates it in the relevant
construction). An apparent alternative within OT would be two sets
of constraint rankings, plus cyclic application of the constraints.
One affix would belong to one of these co-phonologies, and the other
affix to the other co-phonology, similar to the rule strata of
Lexical Phonology. But Urbanczyk shows that this alternative does
not account for the dependence of the shape of the inner affix on
that of the outer affix. The downside of her analysis is the need
for morpheme-specific constraints, but the putative universal status
of constraints in other OT work seems doubtful in any case (see my
earlier comments on Downing's and Ito and Mester's papers).
---------------------------------------
The editing has been done well, save that typesetting has slightly
mangled some constraint names. There are separate indices for
subjects, constraints, languages, and linguists' names.
I close with a general comment. Perhaps the greatest shortcoming of
this book is the delay between the workshop in 1994 and publication
of (a subset of) the papers in 1999, particularly unfortunate given
the rapid changes in OT during those years. Possibly some of the
papers have been revised in the interim. (An occasional footnote
reflects input from the conference, and there are some references
with publication dates up through 1999, albeit mostly by the authors
themselves.) As noted earlier, many of the papers have been made
available in the Rutgers Optimality Archive in the interim, and
other authors (e.g. Downing) have published revised analyses
elsewhere, which further reduces the interest of this printed book.
The delay would have been more understandable if there had been more
added value to the book, such as inclusion of the discussion from
the floor.
References
Anderson, Stephen. 1992. A-Morphous Morphology. Cambridge.
Cambridge Studies in Linguistics 62. Cambridge: Cambridge
University Press.
Beard, Robert. 1995. Lexeme-Morpheme Base Morphology. Albany, NY:
SUNY Press.
Broselow, Ellen I. 1983. "Salish double reduplications: Subjacency
in morphology." NLLT 1: 317-346.
Chomsky, Noam; and Morris Halle. 1968. The Sound Pattern of English.
New York: Harper and Row.
McCarthy, John J.; and Alan S. Prince. 1990. "Foot and Word in
Prosodic Morphology: the Arabic broken plural." NLLT 8: 209-283.
Paradis, Carole. 1988. "On constraints and repair strategies."
Linguistic Review 6: 71-97.
Wilbur, Ronnie. 1973. "The phonology of reduplication." Ph.D.
dissertation, University of Illinois; published (1973) by the IULC.
Reviewer: Mike Maxwell has a Ph.D. in linguistics from the
University of Washington. He has been a consultant for field
linguists working in indigenous languages of Ecuador and Colombia,
and now works in computational morphology and phonology.