Raimy, Eric (2000) The Phonology and Morphology of
Reduplication. Berlin: Mouton de Gruyter, 200pp.
Reviewed by Jason Haugen, University of Arizona
OVERVIEW
Reduplication has been the central topic of much fruitful
research within recent work in Prosodic Morphology and
Optimality Theory (OT) (McCarthy and Prince 1986, 1995,
among many others). In this book Raimy presents a novel
approach to reduplication which abandons both the notions of
prosody and optimal parallel processing in favor of a unique
rule-based derivational account. The primary innovation of
Raimy's model is the introduction of an articulated
linearization process and the utilization of a looping
mechanism prior to linearization to derive the effects of
reduplication. Along the way, Raimy tackles most of the
classic data sets used by McCarthy and Prince to argue for
the Correspondence Theory model, and provides new analyses
based on his own modular and derivational model.
SYNOPSIS
This book is an extension of the author's 1999 PhD
dissertation, completed at the University of Delaware under
William Idsardi. The model developed herein is modular in
that it assumes a specific level of Morphological Structure
(MS), which is post-syntactic and pre-phonological. This is
an assumption that he shares with proponents of Distributed
Morphology (DM) (e.g. Halle and Marantz 1993, Harley and
Noyer 1999, etc.), the framework in which Raimy is working.
In this model, reduplication is treated as a phonological
rewrite rule inserted in the morphology, triggered by a 0-
morpheme which provides secondary exponence for some
morphosyntactic feature. This contrasts with Optimality
Theory, where reduplication is generally assumed to be an
independent morpheme (RED), which can be manipulated by
phonological constraints. The second aspect of the Raimy
model is that it is derivational, in that crucial rule
orderings derive over- and under-application effects, etc.,
as we will see below.
At 8 pages, Chapter 1 is very brief and lays out the
bare outline of what is to come. The notion of precedence
in phonology is made explicit here, and Raimy introduces the
notation that he will use in the rest of the book: '#' to
mark the beginning of a string, '%' to mark the end, and
arrows to mark the precedence relations among the segments
of a string. In this model, reduplication is derived by
introducing a "loop" into the structure, represented in this
work by an arrow from one segment to another. For instance,
in the case of full reduplication of the string #->b->u->k->
u->%, an arrow would be drawn from the final [u] back to
the [b], and after linearization the string would emerge as:
b-u-k-u-b-u-k-u. A syllabic reduplicant would be derived by
drawing the arrow from the first [u] back to the [b] (b-u-b-
u-k-u), a suffixal syllabic reduplicant by drawing the arrow
from the second [u] to the [k] (b-u-k-u-k-u), etc.
(This notation had to prove nightmarish in typesetting, and
I will not even attempt to replicate it here in text-only
format).
Two things are particularly important to note. First,
reduplication occurs with the addition of the precedence
link in morphology. Linear ordering is a phonological rule
which can be ordered with respect to other rules. Second,
an "economy condition" rules out infinite loops: i.e. once a
loop is followed, it is satisfied.
Chapter 2, "The Phonology of Reduplication", introduces the first case
studies which put Raimy's model to work. He starts off by covering over-
and under-application. In the over-application case he presents the
familiar case of Malay nasal spreading, where the first
vowel of the phonological word is nasalized, even though it
is not preceded by a nasal segment to trigger such
nasalization. In OT, this is accounted for by (and is given
as motivation for) constraints necessitating correspondence
between base and reduplicant: IDENT-BR[+nasal]. In Raimy's
model, however, there is no such notion of correspondence.
Instead, the precedence link from the final segment of the
word back to the first segment in fact puts the first
segment in the relevant environment: it is preceded by a
nasal segment, triggering the application of the rule. That
is, once the link is added, the first segment actually does
follow a nasal segment: it is literally in two places at once.
In under-application, the same thing holds, except
that certain conditions are not met, such as the "Uniformity
Parameter", which requires that each element of a segment
meet the structural requirements of the rule for that rule
to apply. For example, in the case of Akan palatalization,
non-low front vowels trigger palatalization of dorsal
segments, except in reduplication: yielding e.g. [ki-ka?]
rather than [tci-tca?] (where c = palatal affricate). Under
Raimy's account, this reduplication adds a precedence link
with a pre-specified [I] segment, and the initial [k] loops
to this [I] and then back to itself, leading to two
different precedence relations: one where [k] precedes [I],
the other where [k] precedes [a]. The latter case does not
trigger the palatalization rule, and since the Uniformity
Parameter is set to "on" this rule does not apply in this
language. If this parameter had been set to 'off', then we
would expect to see forms such as [tci-ka?].
Once again, where OT derives these effects through the
notion of base-reduplicant correspondence, Raimy suggests
instead that what is occurring is precedence relationships
interacting with rules. If linearization occurs before the
application of any given rule (such as nasal spread or
palatalization), then we see normal application of the rule:
the segments involved are no longer identical, but are
instead separated through the linearization process. As
Raimy notes, his model leads to a nice typology of rule and
reduplicant interaction (Raimy's figure (43)).
In Chapter 3, "Precedence in Morphology", Raimy further
develops his idea that reduplication is a readjustment rule
triggered by a 0-morph inserted at morphological structure.
In so doing, Raimy points out what he sees as problems for
treating a reduplicant (un-phonologically-specified-RED) as
an independent morpheme. His primary motivation for the 0-
morpheme claim is internal to the strictures of DM:
readjustment rules can only be triggered by morphemes
providing secondary exponence of a morphosyntactic feature.
Raimy asserts that "reduplication in itself is the secondary
exponent of a morphosyntactic feature in many cases" (p.61),
and even alludes elsewhere that it might be in all cases (p.
66), but the evidence that he gives to support this point is
weak. For example, Raimy has a problem with the high
frequency of polysemy for reduplicants. He cites the
language Nakanai as providing a "particularly clear" (p.63)
case of reduplicative polysemy, and quotes Spaelti (1997)'s
observation that in Nakanai "the shape of the reduplicated
form is independent of the usage. All usages occur with any
of the patterns" (p.63). Unfortunately, he only gives four
examples:
(1) raga-raga 'jumping'/Cont. Habituative Verbs
muluga-luga 'to be first'/Concrete Nouns
bolo-bolo 'many pigs'/Collective Plural Nouns
ilima-lima 'five'/Distributive Numerals (p.63)
There are two things to note. First, there is only one
pattern of reduplication: a foot, so this example set does
not show independence of shape and usage, although it does
show multiple uses for one shape (i.e. polysemy). Other
reduplicative shapes might exist in Nakanai, but Raimy does
not provide us with the data.
One place where Raimy might show more of a problem with
making a claim that RED is a specific morpheme realizing a
specific morphosyntactic feature is in cases where different
reduplicant shapes realize the same morphosyntactic feature.
A case in point is the Uto-Aztecan language Yaqui (also
referred to as Hiaki and Yoeme), where each of the
reduplicated verbs forms in (2) signal habitual action:
(2) a. hina 'hoe, chop weeds' -> hin.na
b. chepta 'jump over' -> chep.chep.ta
c. kinakte 'squint, grimace' -> ki.na.ki.nak.te
The habitual reduplicant (RED) in each of these cases is
realized as a bare mora (in a), a syllable (b, where the
coda consonant is non-moraic), or a foot (c), none of which
is governed by purely phonological conditions (Haugen 2000),
which emphasizes Raimy's point elsewhere that a given
pattern of reduplication is often specific to a particular
lexical item. However, since the morphological gemination
and full foot reduplication are limited to a small set of
verbs in each case, the syllabic reduplicant seems to be the
default, and we can capture this by marking these two cases
as exceptional to the more usual pattern, rather than
marking every verb for one of the three, as Raimy would have
us do.
Second, polysemy is attested elsewhere in morphology,
so I do not see why it should be especially problematic for
reduplication. For example, the [-z] suffix is the primary
exponence of the feature [+plural] in English, and it is
also polysemous: reflecting also the genitive and third
singular verb agreement. I wouldn't want to make any
universal claims about anything based on this pattern, or
any other similar case, as Raimy attempts to do with
polysemy in reduplication.
Perhaps the most interesting section in this book, from
a post-Prosodic Morphology perspective, is the section on
"Deriving Reduplicative Templates", where Raimy abandons all
notions of prosody in favor of segmental X slots. He uses
data from Tohono O'odham for light syllable reduplication,
where he uses "jump links", which promote skipping of
segments, to derive the effects of vowel reduction in the
base, and data from Agta, Ilokano, and Mokilese to argue for
X-slot insertion to derive heavy syllable reduplicants (or
not, in the case of Kusaiean). A-templatic reduplication is
covered through the use of jump links, such as "from the
first segment of the string to the last segment, then to the
first segment preceding a vowel" in Temiar (slog -> s-g-
log), and "first segment of the string to the last segment,
then back to the first segment" in Semai (c?e:t -> ct-c?e:t)
(these are my paraphrases of his rules).
While I grant that such rules can be invoked to derive
many reduplicative templates, Raimy does not discuss any
cases where reduplicative "bases" (to use the OT
terminology; Raimy does not even have "bases", which is also
a problem for him, as we will see presently) are themselves comprised of
prosodic units. McCarthy and Prince (1986) cite the example of
Yidiny, wherein the first foot of the base is fully copied,
and no more. For example, /kintalpa/ surfaces as
[kintal.kintalpa], but /mulari/ surfaces as [mula.mulari]
and not [*mular.mulari]. Similarly, Yaqui verbs (in the
default syllabic case) only copy the first syllable of the
root: /chepta/ yields [chep.chepta], but /vusa/ yields
[vu.vusa], not [*vus.vusa]. In these cases (among others),
the most perspicuous explanation for what is occurring is
not that what is being copied is a series of C's and V's
(e.g. "copy up to the first V in CVCV forms, and up to the
second C in CVCCV forms"), but that what is being copied are
specific prosodic constituents of the root.
Chapter 4 provides a summary to the book, and presents
what Raimy sees as the biggest problems with Optimality
Theory: conspiracies, over-predicting possible
reduplication patterns (e.g. the "Kager-Hamilton problem",
in which a highly-ranked MAX-BR leads to alteration of the
base to conform to the shape of the reduplicant), and
reduplication-specific mechanisms, such as the identity
relation between reduplicant and base.
CRITICAL EVALUATION
One of the most appealing aspects of Raimy's proposal
is his overt articulation of the structure received by the
phonological component, which I will concede is often
neglected in Optimality Theory. Raimy argues specifically
for concatenation of morphemes, and in Chapter 3 he proposes
a series of precedence variables to derive such notions as
prefix, suffix, infix, reduplicant, and free morpheme, and
uses his linearization technique to derive such forms as
English "boyishness" and "table Smable".
Although I share Raimy's theoretical predisposition for
a modular Morphological Structure, as in DM, I do not think
that such a view is incompatible with Optimality Theory.
Here I will address two of Raimy's principal arguments
against OT: the status of RED as a morpheme and the ability
of OT to render phonological explanations.
Raimy has two principle objections to an abstract
morpheme RED. The first has to do with the problem of OT
allowing multiple reduplicants in a specific language.
Raimy reasons thus: "if a reduplication pattern is only and
truly a derivative of language-specific features then there
should be no possibility of multiple patterns of
reduplication in a single language" (pp. 96-7). This might
be true, if we grant the assumption that RED is not a
morpheme. Since OT does not have this assumption, (under the
assumptions of Generalized Alignment) alignment constraints
can freely refer to variable reduplicant shapes. For
example, a hypothetical language might have a single
syllable reduplicant for habitual action in verbs, and a
foot reduplicant for plurals in nouns, and both of these
notions could straightforwardly be accounted for with
alignment constraints: Align (habitual, L, syllable, L) and
Align (plural, L, foot, L).
His second objection to a RED-as-morpheme approach is
that "the Optimality Theory approach that claims that RED
has no phonological content and then derives the surface
phonology of the reduplicative morpheme through the
interaction of constraints denies the morphological nature
of the affix in question" (p.97). I am not quite sure what
he could mean by this, since it is his own theory which
claims that the reduplicant is not an affix, thus is non-
morphological, and it is OT which actively asserts what he
claims that it denies.
Raimy's primary motivation for proposing a new theory
of reduplication is the effort to get rid of reduplication-
specific mechanisms (correspondence), such as Anchor and BR-
Faithfulness. It is true that his model does this.
However, throughout the book Raimy trumpets the fact that
his theory is capable of accounting for the various
phenomena that he discusses. The question that kept raising
in this reviewer's mind is this: What can he not do? Recall
the rule from above: "link from the last consonant to the
segment preceding the first vowel". With such computational
power, in principle anything can be done, and this is not
surprising given the generative power of rule-based systems.
Raimy does appeal to the notion of "analytic simplicity" to
favor relatively few links, but this notion does not seem to
apply to the complexity of the generalization of the links
that he needs to get the results that he achieves.
Finally, I must address one final argument that Raimy lodges
toward OT: the charge that OT is a "conspiracy". Here, I
quote him at length:
It is the entire ranking of all constraints that
evaluate a reduplication structure that
accounts for the reduplication pattern and this makes
it unclear as to what generalization is provided by a
single constraint that affects some dimension of
reduplication. Due to this fact, OT analyses of
reduplication are open to conspiracy arguments since
the generalization accounting for the pattern does not
result from a single statement in the grammar.
Instead, generalizations with respect to reduplication
in OT are provided by the entire grammar of a given
language (p.160).
It seems that what is at issue here is what constitutes a
grammar. Raimy apparently wants a specific statement for
each phonological generalization in the language. On the
other hand, OT provides a single generalization for the
entire language: a list of ranked constraints. It is not
clear to me why the former is superior to the latter, since
the latter captures more information (an entire grammar)
with fewer statements than the former. Ultimately, this
issue, along with other such issues as whether or not there
is utility to such notions as prosodic units, etc., will
have to be left to the reader to decide.
At the conclusion of his book, Raimy suggests possible
extensions of his model into other areas of morphology,
particularly templatic morphology and subtractive
morphology. Both of these morphological phenomena employ
the same mechanism in his system as reduplication: namely,
"jump links", wherein, for example in the latter case, a 0-
morpheme triggers a phonological rewrite rule which adds an
arrow from a particular segment to '%', the end of the
string. This unified account is a nice result for DM, since
phonological material is never actually "erased" (which is a
theoretical impossibility in DM) but instead fails to get
parsed because of the added link. (This is similar to the
Containment model of Prince and Smolensky (1993)). For
theorists who object to 0-morphemes, however, the entire
model will fail to be convincing since it is central to his
theory (and to DM).
Summary
In sum, this brief book is well worth reading, and does
provide an improved derivational model for reduplication.
While I do not think that the arguments laid forth herein
will sway anybody to defect from their current position on
the derivational/optimality debate, this book does serve as
a very useful model for derivationalists and Distributed
Morphologists, and is particularly useful as a problem book
for people working in Optimality Theory. Some of the issues
that Raimy raises about Optimality Theory require rebuttal,
and this kind of critical appraisal of the widely-accepted
(and perhaps currently hegemonic) theory about language is
welcome. After all, we are all looking for the best theory,
and alternative theories, such as the proposal made in this
book by Raimy, usefully serve to keep us working in that
direction.
References
Haugen, J. 2000. Distributing morphology in Yaqui
"secondary" reduplication. Ms, Department of Linguistics,
University of Arizona, Tucson.
McCarthy, J. and A. Prince. 1986. Prosodic morphology. Ms.
Published as Technical Report #32, Center for Cognitive
Sciences, Rutgers University.
McCarthy, J. and A. Prince. 1995. Faithfulness and
reduplicative identity. In University of Massachusetts
occasional papers in linguistics 18, 249-384.GLSA,University
of Massachusetts, Amherst.
Spaelti, P. 1997. Dimensions of variation in multi-pattern
reduplication. Ph.D. dissertation, Department of Linguistics,
University of California, Santa Cruz
About the reviewer:
Jason Haugen is a third-year graduate student in the joint
Ph.D. program in Linguistics and Anthropology at the
University of Arizona. His primary interests are in Native
American linguistics, with particular focus on the Uto-
Aztecan languages.
|
