16.1597, Review: Phonology: Scheer (2004)
LINGUIST List
linguist at linguistlist.org
Thu May 19 02:05:08 UTC 2005
LINGUIST List: Vol-16-1597. Wed May 18 2005. ISSN: 1068 - 4875.
Subject: 16.1597, Review: Phonology: Scheer (2004)
Moderators: Anthony Aristar, Wayne State U <aristar at linguistlist.org>
Helen Aristar-Dry, Eastern Michigan U <hdry at linguistlist.org>
Reviews (reviews at linguistlist.org)
Sheila Dooley, U of Arizona
Terry Langendoen, U of Arizona
Homepage: http://linguistlist.org/
The LINGUIST List is funded by Eastern Michigan University, Wayne
State University, and donations from subscribers and publishers.
Editor for this issue: Naomi Ogasawara <naomi at linguistlist.org>
================================================================
What follows is a review or discussion note contributed to our
Book Discussion Forum. We expect 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 review." Then contact Sheila Dooley at collberg at linguistlist.org.
===========================Directory==============================
1)
Date: 17-May-2005
From: Katalin Balogné Bérces < bbkati at yahoo.com >
Subject: A Lateral Theory of Phonology
-------------------------Message 1 ----------------------------------
Date: Wed, 18 May 2005 22:01:24
From: Katalin Balogné Bérces < bbkati at yahoo.com >
Subject: A Lateral Theory of Phonology
AUTHOR: Scheer, Tobias
TITLE: A Lateral Theory of Phonology
SUBTITLE: What is CVCV, and why should it be?
SERIES: Studies in Generative Grammar 68.1
PUBLISHER: Mouton de Gruyter
YEAR: 2004
Announced at http://linguistlist.org/issues/15/15-3348.html
Katalin Balogné Bérces, Department of English Language and Literature,
Pázmány Péter Catholic University (PPKE), Piliscsaba, Hungary
This is the first volume of a gigantic project aiming to introduce a
radical offspring of Government Phonology (henceforth GP) to the general
public: Strict CV Phonology, or, as Scheer refers to it, CVCV. This
framework, as its name suggests, claims that suprasegmental structure
universally reduces to strictly alternating consonantal ("C") and vocalic
("V") positions; surface sequences of identical segments (i.e., consonant
clusters, geminates, long vowels, diphthongs, hiatuses) enclose empty
skeletal slots. This general introduction is planned to be soon followed
by a second volume, entitled "A lateral theory of phonology. Vol 2: On
Locality, Morphology and Phonology in Phonology."
On the one hand, this first volume is a comprehensive summary of the
fundamentals of both Standard GP and mainstream CVCV, on the other, it
complements the discussion with Scheer's own innovations and amendments,
cleverly and wittily presented and amply supported by arguments and
evidence from a wide array of languages, from French and German through
Czech to Arabic, into all of which Scheer exhibits amazingly reliable
insights. At the same time, the book confronts Standard GP with CVCV as
well as with its own failures. It is meant as a reference rather than a
textbook, as explained in the "How to use this book" section (pp. li-lii);
its intended "look-up function" is supported by the detailed indices at
the end as well as by extensive referencing and cross-referencing to
chapter numbers and the running numbers in the page margins indicating
sections and subsections.
Part One (subtitled "What is CVCV?") is made up of 10 chapters providing a
detailed account of the basics of CVCV.
Chapter 1 ("Introduction") introduces CVCV briefly along with the major
tenets of Standard GP (separately summarized in Appendix 4). Throughout
the chapter Scheer attempts to show how the principles and representations
of Standard GP (KLV 1985, KLV 1990, Charette 1992, Harris 1994, etc.), if
carried to their natural consequences, necessarily result in CVCV
(Lowenstamm 1996, 1999, Scheer 1996, among others). First, Standard GP's
claim that lateral relations (namely, government and licensing -- see
below) exist among segments, which derive syllabic structure (roughly,
head-initial government circumscribes branching constituents; head-final
government characterizes cross-constituent situations) makes syllabic
arborescence redundant, says Scheer. A maximally constrained phonological
theory only needs reference to the lateral relations, and suprasegmental
structure has but two functions left: timing, and the encoding
of "syllabicity" or consonantalness/vocalicness. That is, the
lateralization of prosodic structure inevitably leads to a skeleton-only
model with strictly alternating C and V positions.
Second, Standard GP introduces the phonological equivalents of empty
categories, and with them that of the Empty Category Principle (henceforth
ECP), familiar from syntax. GP claims, based on Kaye's (1990) Coda
Licensing Principle (roughly, a coda must be licensed by a following
onset, that is, codas only exist morpheme-internally), that all words end
in nuclei, and domain-final consonants are onsets licensed by a following
empty nucleus (Final Empty Nucleus -- henceforth FEN). Standard GP also
posits empty nuclei at the left edge of words starting with those
mysterious sC-clusters ("Magic" Licensing -- Kaye 1992) and at vowel-zero
alternation sites. Therefore, CVCV only concludes what was started by
Standard GP when it assumes the presence of an empty V position
(henceforth indicated by a lower-case v) sandwiched between any two
surface consonants, and an empty C position (henceforth c) between any two
surface vowels. All consonant clusters, including geminates, then, are CvC
sequences, while long vowels and hiatuses are VcV strings. There are two
types of consonant sequences: one whose members exhibit selectional
restrictions since they enter into a dependency relationship (called
government), and fake or "bogus" clusters, which do not show such
characteristics. Similarly, the V's in long vowels and diphthongs contract
some kind of relation, while the two terms of a hiatus do not communicate
with each other.
Chapter 2 ("Open versus closed syllables in CVCV") introduces the
Projection Principle, which states that governing relations are
established at the level of lexical representation and remain unchanged
throughout the derivation. In light of this, vowel-zero alternations
cannot be analysed as instances of either deletion or insertion, both of
which would modify the governing relations constituting syllable
structure. Instead, Standard GP resorts to underlying empty nuclei, which
vocalize if not affected by any external force, but remain silent if they
enter into an internuclear relation, called Proper Government (PG), with a
following phonetically expressed vowel. Crucially, phonetically
unexpressed nuclei are unable to properly govern. Another major condition
on PG in Standard GP is that no governing domains may intervene between
the governor and the governee, that is, alternating vowels are pronounced
when immediately followed by a consonant cluster, blocking the application
of PG. Scheer, however, claims that CVCV does not need this latter
subclause in the formulation of PG, as the blocking effect of clusters
falls out naturally from the structure of the skeleton: since all
consonants in clusters are separated by an empty nucleus, which is unable
to govern, a governor is never available to pre-cluster (potential)
alternation sites.
Next, Scheer turns to the distinction between open and closed syllables in
CVCV, more specifically, between rising- and falling-sonority clusters.
The former, traditionally analysed as branching onsets, leave preceding
syllables open, while the latter, traditional coda-onset sequences, make
the syllable to the left closed. Apparently, CVCV represents both alike:
as CvC sequences. Scheer uses the example of Czech prefixes exhibiting
vowel-zero alternation (e.g. "pod-" 'under' surfacing as "pode-" before
syncope sites, i.e., before unpronounced empty nuclei, which indicates
that the prefixal alternation is driven by PG) to point out that
unvocalized prefixes occur only before (tautomorphemic) branching onsets
(e.g. "pod-bradek" 'double chin'). He concludes that governing domains do
not always block PG, and introduces a new governing relation, which he
calls Infrasegmental Government (IG), along with the melodic and
phonotactic conditions on consonants entering into IG. At this point, he
needs to discuss the internal structure of consonants, which is modelled
in GP with unary elements, i.e., with primes which are either present or
absent in melodic expressions. As a consequence, the segmental
representation of sounds differs in complexity: certain classes (e.g.
obstruents in Standard GP) contain more elements and are thus more complex
than others (e.g. sonorants in Standard GP). (Notice that sonority in GP
is a derived category, following from complexity.) As expressed in Harris
(1990), complexity plays a pivotal role in government since only segments
of higher or equal complexity can govern others (i.e., obstruents are good
governors, sonorants are typical governees). As opposed to this, Scheer
here claims that the sonorant is the head of IG, and supports this with an
investigation of segmental alternations (whose details will now be omitted
for reasons of space). Finally, he concludes that (i) sonorants are
considerably more complex than obstruents, in fact, they are "too complex
to be governees" (p.58, title of Section 4.3.6.3), and (ii) complexity
counts place elements only. Then he formalizes the final version of the
definition of IG (p.64): (i) a consonant A may contract a governing
relation with its neighbour B iff there is a place-defining autosegmental
line where A possesses a prime, while the corresponding slot in the
internal structure of B is empty. In this situation, the prime belonging
to A governs the empty position of B; (ii) the empty nucleus that is
enclosed within such a domain of IG is circumscribed. Its ECP is satisfied.
The melodic condition on IG is, then, based on complexity. The other main
condition is phonotactic: Charette's (1990, 1991) Government Licensing is
adapted to constrain IG, namely, a domain of IG may only be established if
the head of the domain is licensed to act as a governor by its own
nucleus. (p.65) As a result, the directionality of IG (right-to-left) is
not stipulated but derived: in a CvC sequence, only the C on the right can
ever be followed by a potentially licensing nonempty nucleus. In addition,
the second subclause of the definition reveals that IG is one of the
factors in the phonological ECP, whose final version runs as follows
(p.67). A nucleus may remain phonetically unexpressed iff it is: (a)
properly governed; or (b) enclosed within a domain of IG; or (c) domain-
final. Already in Standard GP, domain-final empty nuclei (FENs) are
parametrically licensed to remain empty: if the parameter is set ON, the
language will exhibit consonant-final words (e.g. English), but in the
case of its negative setting, all words in the language will end in a
vowel (e.g. Italian). Scheer also points out that in fact, CVCV predicts
that the right edge of domains is special and irregular. Since all lateral
relations (viz., PG and IG) are right-headed, an element lacking a
following neighbour is necessarily problematic and requires special
treatment. The chapter closes with a preliminary definition of open vs.
closed syllables in CVCV (p.78) (the final definition is to come in
Chapter 7): a vowel stands in an open syllable iff it is the target of a
lateral relation; a vowel stands in a closed syllable iff it is the target
of no lateral relation. In addition, the traditional concept of coda is
reinterpreted in CVCV (p.80, repeated on p.164) as a consonant that occurs
before a governed empty nucleus.
Chapter 3 ("A unified theory of vowel-zero alternations") takes vowel-zero
alternations under scrutiny again. Recall that in Standard GP alternation
sites are vocalized if they are followed by (i) an empty nucleus unable to
govern, or (ii) a governing domain, which blocks PG. As it has been
pointed out above, this description involves disjunctivity. In CVCV,
however, all vocalizations have the same cause, namely, the absence of PG,
therefore it offers a unified analysis. PG is only restricted by two
language-specific parameters: whether the alternation is obligatory (e.g.
Slavic) or optional (e.g. French), and what vowels take part in the
alternation.
Chapter 4 ("Alternating vowels are present in the lexicon") discusses the
status of empty nuclei in Standard GP, where it is generally accepted that
they are indeed devoid of any melody, and the language-specific
differences in the quality of alternating vowels are due to the fact that
in the course of phonetic interpretation, "ambient" elements are inserted
through language-specific epenthesis. Scheer argues against this analysis
by bringing examples from Slavic again. He points out that in Slavic
languages the quality of the alternating vowels is not always predictable
since quite often there is more than one vowel participating in the
alternation. He proposes to assume the presence of some floating melody as
well as a mechanism deleting unassociated segmental material in these
cases, and then he goes on to generalize the deletion analysis to all
cases of vowel-zero alternation. He claims that this choice is also
enforced by the need to distinguish alternating empty nuclei from non-
alternating ones. Ultimately, then, the system derives three types of
nuclei: (i) non-alternating nonempty ones with associated melody, (ii)
alternating ones with floating melody, and (iii) non-alternating empty
ones with no melody at all.
Chapter 5 ("The beginning of the word: "#"=CV") addresses the issue of the
phonology-morphosyntax interface. Scheer assumes that phonology can only
handle phonological objects and not morphological or syntactic ones,
therefore any morphosyntactic information which affects phonological rule
application must be first translated into "the language of phonology".
Following Lowenstamm (1999), he identifies the beginning of the word as an
empty cv-span that is treated by the phonology in the same way as any
empty category; more specifically, the empty v position in this boundary-
marker (which is then the phonological embodiment of #) needs to be
silenced by PG to avoid the violation of the ECP. Scheer also looks into
why # triggers the same effects cross-linguistically (i.e., "Why do pink
panthers always do the same things?" -- title of Section 2.1, p.97),
namely: (i) in a number of languages, there exist heavy restrictions on
what sequences can constitute initial consonant clusters; (ii) the first
vowel of words is cross-linguistically stable, that is, it resists
syncope; (iii) word-initial consonants occupy a strong phonological
position. He repeats the Czech data dealt with above (recall "pod-
bradek" 'double chin'), where an empty nucleus was shown to exist to the
left of a rising-sonority cluster, and based on the observation that word-
initial clusters behave in the same fashion, he concludes that they are
also preceded by an empty nucleus. Since V positions in CVCV always follow
C positions, this means that all words (in the relevant languages) start
with an empty cv boundary-marker. In a cvCvCV sequence, there are two
empty nuclei which require special treatment. If the two melodically
filled consonants enter into IG, which satisfies the ECP of the sandwiched
v, then the PG of the filled nucleus can hit the v of the boundary marker.
However, no cases of falling-sonority clusters can arise in that position,
since the obstruent cannot govern the sonorant (due to its smaller
complexity), the sonorant is not licensed to govern (due to the absence of
a following nonempty nucleus), so no IG is possible, and as a result there
remain two empty nuclei waiting to be silenced by a single filled one. The
ECP of one or the other empty v will be left unsatisfied. The presence of
the boundary-marker, says Scheer, is determined by the setting of a
parameter. If it is not sent down to phonology, initial consonant clusters
are unrestricted (as in Moroccan Arabic, Slavic, or Greek); if it is
present, only rising-sonority clusters are possible (as in English).
Scheer also points out the circularity of the sonority approach to initial
clusters, and stresses that his analysis is based on restrictions which
follow from principles and mechanisms independently justified (the ECP,
Government-Licensing, IG). Finally, he contrasts IG with Standard GP's
Interonset Government, and concludes that in CVCV it is not Interonset
Government that is responsible for word-initial clusters.
Chapter 6 ("The Coda Mirror") investigates into phonologically strong
positions, that is, the word-initial and post-coda environments (i.e.,
AFTER the notorious disjunction {#, C}), which systematically resist
lenition or even trigger fortition. Scheer brings empirical evidence from
the evolution of French, allophonic variation in Somali, syncope in
various languages, and Sievers' Law in Gothic, and dubs the position in
question The Coda Mirror. In CVCV, the two subcases have something in
common: the consonant is after a governed empty v. He argues that the
consonant in a vCV sequence is in a strong position because the empty
nucleus distracts the PG of the filled V from it, and consequently it is
licensed by the following vowel but not governed by it. He gives proper
definitions of government and licensing ("government inhibits, licensing
enhances the segmental expression of the target" -- title of Section 6 on
p.134; "a good guy and a bad guy" -- title of Section 6.4 on p.138), and
highlights the fact that this theory achieves explanatory adequacy: it is
able to account for why precisely those two positions are strong. Other
consonants, e.g. the ones appearing intervocalically (i.e., VCV) are both
governed and licensed, while "coda" consonants (i.e., VCv) are ungoverned
and unlicensed, therefore these two are weak phonological positions, which
support lenition. At the end of the discussion, Scheer remarks that OT's
Positional Faithfulness is inadequate since it fails to notice the
disjunction -- it is not only the beginning of the word that is strong,
but the post-coda position, too.
Chapter 7 ("Consequences of the Coda Mirror: no confusion between
Government and Licensing anymore") surveys Standard GP's confusing (and
confused) usage of the two terms. As it has already been demonstrated in
the previous chapter, no such confusion is apparent in CVCV, where
government and licensing are defined as two antagonistic forces (although,
as Scheer himself realizes, IG is peculiar since it has no segmental
effects, it can satisfy the ECP of empty nuclei without targeting them,
and it is directly determined by the melodic make-up of the participants).
Here Scheer turns to Internuclear Licensing, one possibility not
considered up to this point. He examines vowel shortening phenomena before
an empty nucleus ("closed-syllable shortening") along with vowel length
alternations, e.g. Italian Tonic Lengthening, and concludes that long
vowels are underlying Vcv sequences, and the melody of the first term
spreads onto the empty v during the derivation. As the conditions on the
occurrence of long vowels indicate, the target of spreading must be
licensed, that is, the empty v must be followed by a nonempty one in the
next CV-unit, which licenses it. Thus we arrive at the final typology of
lateral relations. All of V-to-V government (effecting vowel-zero
alternations), V-to-V licensing (alternations in vowel length), V-to-C
government (intervocalic lenition), V-to-C licensing (strong position,
*#RT), C-to-C government (or licensing? -- the reader is referred to the
second volume) (IG) are attested. C-to-V relations are impossible since it
is a phonological commonplace that vowels (nuclei) dominate consonants.
And the final definition of open vs. closed syllables in CVCV is as
follows (p.177): (a) a vowel stands in an open syllable iff it is the
target of either government or licensing; (b) a vowel stands in a closed
syllable iff it is the target of neither government nor licensing.
Chapter 8 ("A syntax of phonology") surveys the lateralization of
structure and causality in Standard GP, where some vertical structure and
causality are still present (e.g. syllable structure, coda effects, the
Binary Theorem, arboreal structure in Harris' Licensing Inheritance). In
CVCV, however, the total lateralization of both structure and causality is
in effect. Scheer does not only claim that syllable structure is redundant
but he also argues that it is redundant to distinguish coda-onset clusters
(falling-sonority clusters exhibiting cooccurrence restrictions) and bogus
clusters (syncope-created clusters and consonant sequences only attested
word-medially, e.g. /tl/, /tk/). He introduces the distinction between
what he calls HIGH and LOW: the territories above and below the skeleton,
respectively. In the HIGH sphere, no melodic conditioning on processes is
possible (this is where government and licensing are contracted, and
position-triggered processes like lenition/fortition take place), while
the LOW area is heavily affected by the melodic make-up of segments (as in
e.g. IG, and adjacency-triggered processes like assimilation). Scheer
points up a side-effect of CVCV, too: all lateral relations are head-final.
Chapter 9 ("Lateral relations are head-final: length in phonology")
discusses issues related to segment length. First, the difference between
systems with alternating vs. inalterable vowel length is under scrutiny.
Three traditional labels (closed-syllable shortening, open syllable
lengthening, tonic lengthening) are shown to actually stand for one single
phenomenon: as argued in Chapter 7, alternating vowels are only partially
associated lexically to a skeletal position, and they are long when the
complement is licensed, but short when it is not. In the case of
Compensatory Lengthening, spreading to both directions is on record, but
the loss of a consonant universally triggers the left-to-right spreading
of a preceding vowel, whereas right-to-left spreading only seems possible
in cases when the first member of a vocalic sequence is elided. Notice
that in both cases the target of spreading is licensed by a following
nonempty nucleus, therefore Scheer concludes that all long vowels obey the
same requirement, and he also proposes that non-alternating vowels are
head-final. In CVCV, geminate consonants can also be either head-initial
or head-final, depending on the direction of spreading, but in both cases
the sandwiched empty v needs to be properly governed by a following
nonempty V, which correctly predicts that geminates cannot be preceded or
followed by consonants.
In Chapter 10 ("Syllabic and trapped consonants in CVCV"), Scheer develops
his view that glides are underlying vowels spreading into c positions,
while syllabic consonants are underlying consonants spreading into v
positions. Based on the synchronic and diachronic distribution of syllabic
consonants, he concludes that they are left-branching, while Polish
trapped consonants are right-branching. The emergence of both is due to
the ECP, since a consonant spreads into an empty v to prevent it from
remaining empty. Later, he modifies this representation and claims that
syllabic consonants are ternary branching (VCV) sequences, since they
exhibit the properties of both CVRC and CRVC strings, although he himself
notices the dubious identity of such a structure. Finally, he makes the
surprising (and yet unexplained) observation that trapped sonorants take
part in voicing alternations as if they were obstruents, and then he turns
to the evolution of syllabic and trapped consonants in the Slavic
languages.
Part Two ("Why CVCV?") lists ten good reasons why CVCV is worth
considering seriously. After a brief introduction and tabular overview
(Chapter 1 - "Introduction"), Chapters 2-5 (pp.369-457) introduce the
basic principles of argumentation and reasoning used in the discussion of
the ten arguments in favour of CVCV: (i) disjunctive contexts should be
eliminated from phonological theory; (ii) the function of (autosegmental)
representations, serving as "overgeneration-killers" (p.371), is stressed,
and their explanatory power underlined, in fact, Scheer here attacks OT
for failing to put emphasis on representations (it "has turned back the
wheel to the times of wild overgeneration" - p.371); (iii) the generality
of phonological processes is crucial for the analysis, and a theory that
can account for a wider range of phenomena using the same mechanism is
superior; (iv) extrasyllabicity as the consequence of syllabification
algorithms and serialism is only real as an observation but not as an
analysis or explanation of observations. He surveys the reasons for
positing extrasyllabicity in generative phonological theory and expresses
the need for a theory of margins. He pinpoints that in CVCV
extrasyllabicity is not available, instead, the special status of the
beginning and end of the word is expressed by the special edge-marking
elements introduced earlier: the boundary-marker and the FEN.
And here come the arguments.
In Chapter 6 ("Argument One: Languages without initial restrictions") the
sonority sequencing generalization is examined, and Scheer finds that it
is based on data from IE languages only. It leaves others like Moroccan
Arabic unexplained, since they impose no restrictions on initial clusters
(and consequently Scheer dubs them "anything-goes languages"), therefore
even falling-sonority sequences occur. In CVCV, however, all consonant
sequences are separated by an empty nucleus, that is, all languages are
analysed as fitting the Moroccan Arabic pattern.
Chapter 7 ("Argument Two: What you get is NOT what you see: Tina Turner
was wrong") argues for the existence of empty nuclei. Empty onsets are
quite generally accepted both word-initially and in hiatus, and floating
consonantal melody is frequently employed in analyses, e.g. to account for
liaison. That is, for consonants it is commonplace that what you see is
not necessarily what you get -- there are latent consonantal positions and
melodies. The same is not true, however, for vowels: neither empty
skeletal positions nor floating melodies are traditionally used for
nuclei. In contrast, CVCV makes the treatment of empty categories uniform.
Chapter 8 ("Argument Three: Description vs. explanation of restrictions on
word-initial consonant clusters") takes another look at initial clusters,
and accuses the sonority approach of circularity. It is claimed that an
initial cluster is well-formed if it contains a rise in sonority, because
the observation is that in initial clusters sonority usually rises. In
fact, constraints are inherently circular, says Scheer. In CVCV, however,
complexity and lateral relations derive the attested pattern. There are no
initial falling-sonority clusters in typical IE languages, since the
potential governor (the sonorant) is followed by an empty nucleus and
therefore remains unlicensed. Consequently, no IG is contracted, and the
ECP of the said empty nucleus is not satisfied. As the insertion of the
boundary-marker is parameterized, it is not sent down by morphology in
anything-goes languages. As a result, the ECP of the empty v between the
initial consonants can always be satisfied by the government emanating
from the first pronounced vowel, irrespective of the quality of the
flanking consonants. CVCV is also able to make correct predictions: (i)
there are no languages in which only falling-sonority onsets are legal
since the boundary-marker is either present (in which case only sequences
exhibiting a rising sonority profile are well-formed initially), or it is
absent (in which case anything goes); (ii) there can only be at most one
extrasyllabic consonant. These facts remain unaccounted for in traditional
frameworks.
Chapter 9 ("Argument Four: Lower: empty Nuclei and regressive internuclear
relations have been used for over 30 years in the analysis of Slavic vowel-
zero alternations") presents traditional analyses of the Slavic yers, and
discovers that "empty Nuclei and lateral relations are an established tool
in the phonological analysis of Slavic for over 30 years; for some reason,
nobody has equated yers with empty Nuclei, and Lower with Government"
(p.495).
Chapter 10 ("Argument Five: The life of "yers" outside of Slavic and in
locations where vowels do not alternate with zero") surveys occurrences of
the familiar yer context other than vowel-zero alternations. Czech vowel
length alternations, vowel quality alternations in Czech and Polish, and
the ATR value of French mid vowels and alternations involving schwa are
examined, and several previous (pre-GP) analyses of schwa and yers (incl.
Anderson (1982) and Hall (1992) for schwa in French and German, resp., and
Spencer (1986) for Polish yers) are described. Scheer points out that a
unified analysis of processes in the yer context is only available if the
assumption is made that all coda-onset sequences are broken up by an empty
nucleus. He sets up new parameters governing language-specific variation
in the lateral ability of schwa and the FEN, for government and licensing
separately.
Chapter 11 ("Argument Six: Unified representations for the syllable and
stress") presents joint work with Péter Szigetvári. It claims that stress
is a property of nuclei rather than syllables, and in weight-sensitive
stress systems, like Latin, stress rules count nuclei other than
ungoverned empty ones. It naturally follows that onsets never count
separately because their nuclei are always filled, whereas codas do add
to "syllable weight" since they are followed by governed empty nuclei.
Onset clusters do not add to weight either since their enclosed nucleus is
ungoverned and therefore disregarded by the algorithm. An explanation is
also available for why ungoverned empty nuclei are ignored: they are
invisible from ABOVE because they are silenced by a LOWER mechanism, viz.,
IG. Therefore, in CVCV syllabic and prosodic generalizations are accounted
for by the same structure; the weight-by-position parameter is in fact a
parameter determining the visibility of governed empty nuclei. One of the
most significant features of the analysis is that it offers an explanation
of why onsets do not count -- a mere stipulation in traditional approaches.
Chapter 12 ("Argument Seven: Licensing power of final empty Nuclei
parameterised: paired vs. impaired behaviour of internal and final Codas")
surveys cases where internal codas have different effects from final
codas, and cases where the two behave in the same way. Scheer concludes
that the facts call for a parametric explanation. Standard GP's Coda
Licensing, however, cannot be parameterized, but the ability to contract
lateral relations can. He then claims that word-final consonant clusters
only exist in languages where the FEN can govern. If the FEN cannot
license, word-final consonants are in weak position. Final consonants and
internal codas behave alike if the FEN can neither govern nor license.
CVCV explains, while the extrasyllabicity approach doesn't explain, why
there are no extrasyllabic vowels, and why extrasyllabicity is not
selective (no cases are on record in which a consonant is extrasyllabic
for one rule but not for another, although this is theoretically possible
in traditional frameworks). Finally, the lateral abilities of schwa are
also parameterized.
Chapter 13 ("Argument Eight: The Coda Mirror") shows that the Coda Mirror
context introduced in Part One, Chapter 6 above, i.e., the word-initial
and post-coda positions, can only be grasped as a natural class only in
CVCV, along with the account of why it witnesses fortition and the absence
of lenition.
Chapter 14 ("Argument Nine: News from the yer context: what happens in
Codas and before an unpronounced alternating vowel") describes processes
affecting a sonorant in the same way in tautomorphemic clusters and before
unvocalized vowel-zero alternation sites, and states that in CVCV these
environments can be subsumed under a single structural configuration:
before governed empty nuclei.
Chapter 15 ("Argument Ten: What sonorants do in Codas: a unified theory of
melodic reaction on positional plight") aims to shed light on "the world
of sonorants through the prism of CVCV" (p.707) by comparing apparently
unrelated processes. First, in the well-known cases of nasal place
assimilation, Scheer claims that the nasal is active rather than the
following obstruent, and traditional analyses are misled by "the mirage of
assimilation" (p.708). Since the nasal is followed by an empty v, it is
ungoverned unlicensed, which results in its instability -- it is "in
positional plight", as Scheer puts it. Nasals then assimilate to
obstruents in search of support. Second, nasals in final codas are under
scrutiny: nasal place is neutralized in alveolar in Somali and in velar in
Southern French (Midi French), and in a nasalized velar glide in Polish.
Scheer seeks for a uniform interpretation of internal homorganicity and
final loss of place, which is available in CVCV: in both cases, the nasal
is in positional plight. The same explanation is given for the emergence
of nasal vowels in French, Portuguese, and Slavic. Also, Scheer looks into
German homorganic CN-clusters arising after schwa-deletion and syllabic
sonorant formation in final syllables (as in "haben"), and finds that the
nasal is again in positional plight and prevents the schwa from being
pronounced by occupying its position. Although the direction of spreading
in nasal place assimilation is usually predicted in traditional theories,
it isn't in CVCV: the nasal either docks to the left (in the "haben"
pattern) or to the right (in NC sequences). Notice that the same
explanation is offered for syllabic consonant formation in Germanic and
Polish trapped consonants. At the end of the chapter, the analysis is
extended to all sonorants, to cover all instances of sonorant
syllabification and lenition.
In the "General conclusion", Scheer repeats the two major claims of the
book: on the one hand, CVCV takes the GP research programme to its logical
end (that is, it carries out the ultimate lateralization of structure and
causality); on the other hand, its flat phonological structure correctly
predicts the absence of recursion in phonology. He makes a final remark
concerning the scarcity of evidence provided in the book for the empty
position straddled by the members of onset clusters and branching nuclei.
He explains it by admitting that the collection of arguments presented in
the book strongly reflects his personal choice, and by claiming that the
ample evidence for the empty v within coda-onset sequences constitutes a
fatal attack on branching structure altogether anyway. Scheer repeats the
major goals of the book, too: to introduce CVCV and argue for its
superiority to either its predecessor, Standard GP, and its most serious
competitor, OT.
The book contains useful appendices in the back: a list of the parameters
introduced above, a detailed description of closed syllable shortening vs.
diminutive lengthening in Czech referred to at various points in the book,
an exhaustive list of Polish tautomorphemic two-member word-initial
clusters, and finally, a short guide to Standard GP (incl. melodic and
prosodic representations, basic principles, and its conception of
derivation).
Despite its "rather scary size" (p. lii) (854 pages in all), the book is
surprisingly reader-friendly. Given Scheer's long section titles, the
overview of the table of contents at the beginning is very practical. In
addition, there are several "roadmaps" intertwined with the running text,
e.g. at the beginning of Chapter 2 (pp.366-368). The book also applies
section numbering to help orientation. This is supplemented by Scheer's
witty style (mentioning Tina Turner, pink panthers, etc.), which lightens
the otherwise strict argumentative atmosphere. Throughout the book, an
extensive survey of the relevant literature is offered, which is proven by
the references section, which is more than 40 pages long (pp.779-823).
Another general remark is in order here, although it concerns the
publisher rather than the author. The book contains an unfortunate number
of typos and other errors, most probably owing to the lack of proper proof-
reading.
As to the contents and mode of presentation, the book is comprehensive,
self-contained, and the argumentations are clear and presented in an easy-
to-follow fashion. A possible objection might be that the book is perhaps
too bulky to popularize a relatively new, radical theory of the minority.
In this OT-dominated world, informing the open-minded about alternative
theories may be one of the main goals of publication -- and discouraging
your audience by the profusion of information is not necessarily the best
strategy.
Also, although the alternative CVCV treatments of the phenomena discussed
in the book are mentioned and referenced, they are not described or
introduced in more detail, which creates the false impression that what is
included in the book constitutes the mainstream of Strict CV Phonology
(rather than Scheer's own views, some of which are not shared by other
proponents of CVCV). Of course this is not a major defect, all the less so
since Scheer does warn the reader at the very beginning (p. xlviii).
Still, it diminishes the intended "look-up function" of the book --
a "dictionary-like" reference book is expected perhaps to be a bit less
biassed. Therefore, Scheer's suggestion as to its usage: "you want to know
what CVCV says about X, so you look it up" (p.li), is in fact best
interpreted as "you want to know what Scheer says about X, so you look it
up". (Take it from me, the book serves that function brilliantly!)
It has been mentioned above that the General conclusion admits the book's
failure to bring arguments for the existence of empty slots within
branching onsets and nuclei, and Scheer explains this by referring to the
personalness of the choice of data and topics covered. Again, if the book
is not meant to be exhaustive, its look-up function may not be totally
fulfilled.
In spite of these minor weaknesses, the book is a comprehensive summary of
major work done in CVCV, and even if in time it turns out to best serve
its intended look-up function for CV-ers and GP phonologists, and not be
as instructive for non-GP-ers as the author planned when he set out to
write it, it is such a useful collection of linguistic data from a wide
range of languages and phenomena (faithfully reflected by the impressive
size of the indices at the end -- the subject index (pp.825-839) is 14
pages long, the language index (pp.842-854) lists more than 80 languages
and their (diachronic or synchronic-regional) varieties, along with their
nearly 200 processes and phenomena) that any analyst of any theoretical
taste can resort to it for raw material and inspiration.
REFERENCES
Anderson, Stephen R. (1982) The analysis of French shwa: or, how to get
something for nothing. Language 58.3: 534-573.
Charette, Monik (1990) Licence to govern. Phonology 7: 233-253.
Charette, Monik (1991) Conditions on Phonological Government. Cambridge:
CUP.
Charette, Monik (1992) Mongolian and Polish meet Government Licensing.
SOAS Working Papers in Linguistics and Phonetics 2: 275-292.
Hall, Tracy (1992) Syllable structure and syllable-related processes in
German. Tübingen: Niemeyer.
Harris, John (1990) Segmental complexity and phonological government.
Phonology 7: 255-300.
Harris, John (1994) English Sound Structure. Cambridge, Mass. and Oxford:
Blackwell.
Kaye, Jonathan (1990) 'Coda' licensing. Phonology 7: 301-330.
Kaye, Jonathan (1992) Do you believe in magic? The story of s+C sequences.
SOAS Working Papers in Linguistics and Phonetics 2: 293-313.
KLV (1985) = Kaye, Jonathan, Jean Lowenstamm and Jean-Roger Vergnaud
(1985) The internal structure of phonological representations: a theory of
charm and government. Phonology Yearbook 2: 305-328.
KLV (1990) = Kaye, Jonathan, Jean Lowenstamm and Jean-Roger Vergnaud
(1990) Constituent structure and government in phonology. Phonology 7: 193-
231.
Lowenstamm, Jean (1996) CV as the only syllable type. In Jacques Durand
and Bernard Laks (eds.) Current Trends in Phonology: Models and Methods.
European Studies Research Institute, University of Salford Publications:
419-442.
Lowenstamm, Jean (1999) The beginning of the word. In John Rennison and
Klaus Kühnhammer (eds.) Phonologica 1996. Syllables!? The Hague: Holland
Academic Graphics: 153-166.
Scheer, Tobias (1996) Une théorie de l'interaction directe entre
consonnes. PhD dissertation, Université Paris 7.
Spencer, Andrew (1986) A non-linear analysis of vowel-zero alternations in
Polish. Journal of Linguistics 22: 249-280.
ABOUT THE REVIEWER
Katalin Balogné Bérces took her M.A. in English Language and Literature
from the Faculty of Humanities, Eötvös Loránd University (ELTE), Budapest
in 1998, and started her research as a doctoral student in the English
Linguistics PhD Programme of ELTE in the same year. Her field of research
is the phonology, more specifically the syllable structure, of English.
She completed her PhD dissertation (entitled "Strict CV Phonology and the
English Cross-word Puzzle") in February 2005, and is expecting to defend
it in September 2005. She works as a full-time assistant lecturer in the
Department of English Language and Literature, PPKE, and teaches various
courses on English linguistics, phonology, syntax, and dialectology.
-----------------------------------------------------------
LINGUIST List: Vol-16-1597
More information about the LINGUIST
mailing list