Review of Studies on Reduplication
|Date: Mon, 09 Jan 2006 22:37:08 -0500
From: Mike Maxwell
Subject: Studies on Reduplication
EDITOR: Hurch, Bernhard
TITLE: Studies on Reduplication
SERIES: Empirical Approaches to Language Typology 28
PUBLISHER: Mouton de Gruyter
Mike Maxwell, Center for Advanced Study of Language, University of
The papers in this book were originally given at the Graz Conference
on Reduplication in 2002. It appears that the authors were given the
opportunity to revise their papers after the conference.
The conference organizers requested that presenters ''strengthen
their theoretical claims with broad substantive evidence'', and the
papers by and large reflect this dual perspective of theory and data,
although naturally the papers range from those making strongly
theoretical points to those mostly concerned with data.
With twenty five papers in this volume, I will not attempt to analyze
each one. The editor does a good job of summarizing the papers in
his introduction (and Carl Rubino gives an overview of reduplication
processes around the world, setting the stage for the further
discussion). Rather, I will discuss the book and its papers by themes.
(The table of contents is available at
1. Repetition vs. reduplication: Diachronically, one plausible path to
reduplication starts from repetition. If reduplication is grammaticalized
repetition, then distinguishing the two requires some criteria. David
Gil proposes several such criteria, e.g. that reduplication is only word-
sized, never phrasal. Gil's criteria give unclear results in some cases
(as he recognizes).
Contrary to Gil, Sharon Inkelas, in discussing her construction-
based ''Morphological Doubling Theory'' (see below), explicitly argues
that phrasal doubling can and should be treated as the same sort of
operation as morphological reduplication.
Also disagreeing with Gil, Utz Maas looks at a number of reduplication
and reduplication-like syntactic constructions in Arabic languages, and
in particular on the use of the 'masdar', a form rather like a gerund in
English. The masdar is morphologically derived from a verb, and is
often used together with an inflected form of the same verb to
strengthen the meaning of the whole, rather like ''Hitting, he hit'' to
mean ''He hit violently.'' (Readers of more literal translations of the
Bible will be familiar with this construction from another Semitic
language, Hebrew.) If this is reduplication, rather than a stylistic form
of repetition, then the two parts of the reduplicant are inflected
As these papers show, the distinction between repetition and
reduplication looks more like a cline than a sharp division. Some
theories are able to handle such gradient distinctions, while others
posit very different mechanisms for repetition and reduplication, so
that a dividing line somewhere is necessary. I anticipate further
disagreement on this issue in the future.
If it is unclear in some cases whether the repetition of a phrase or
differently inflected forms of a word constitutes reduplication, a lack of
clarity is apparent at the opposite end of the continuum, too. Dina El
Zarka argues that consonantal gemination in Arabic should be treated
as reduplication, on the grounds that it is at least partly morphological
(not just prosodic) in nature. While this seems reasonable, El Zarka
further argues against the notion that gemination is an endpoint in
Arab grammaticalization, because some varieties of modern Arabic
utilize both gemination and full reduplication (in biconsonantal roots),
and the geminated forms are demonstrably older. The geminated
forms therefore cannot, El Zarka claims, have arisen from reduplicated
forms. I found this less convincing; surely reduplication can arise
more than once in a language. (El Zarka does not claim that the
forms with full reduplication arose from the geminated forms, only that
the former are replacing the latter in some contexts.)
2. Morphological vs. Phonological Reduplication: Sharon Inkelas
argues for her ''Morphological Doubling Theory'', in which a
reduplicative morpheme is treated not as an empty prosodic structure
that gets filled by copying from an adjacent base, but rather as the
actual (grammaticalized) repetition of a morphological constituent, with
possible subsequent truncation. But several of her pieces of evidence
lend themselves to re-analysis in a phonologically-based theory. For
instance, in Mokilese, most verb stems form progressives through
prefixing CVC reduplication. Monosyllabic verb stems, however, use
triplication. Inkelas claims that a phonological theory would have to
analyze the triplicated forms as being having two instances of a 'RED'
morpheme, which should—but does not—imply a meaning difference.
In contrast, her morphological construction-based theory simply
requires the daughter to appear twice for monosyllabic verbs. I am
less than convinced by this argument; it seems that in either case,
monosyllabic verbs are a special case, and it is not clear why the
morphological theory fares any better than a phonological theory.
Moreover, the phonological theory she uses here is really a straw
man. The right theory to compare, it seems to me, would be a
phonologically-based theory using a prosodic template requiring a
progressive form to contain at least three syllables (or perhaps two
feet). The triplication of monosyllabic stems is then explicable as
being the only way to fill this template.
Inkelas also argues from 'synonym compounds' in Southeast Asian
languages (Khmer and Vietnamese), in which the compound consists
of two more or less synonymous words. (In some cases the glosses of
the two words are identical, while in others they are only similar—
e.g. 'old' and 'mature'. But of course glosses can be misleading, and it
seems likely that all the word pairs in this construction are not
perfectly synonymous.) If these synonym compounds are
reduplications, then it is clear that reduplication does not consist of
phonological copying, since the words in question are phonologically
quite different. Inkelas also brings up an antonym compound
construction in Acehnese (the compounds have glosses like 'old and
young', 'day and night'). While these are interesting constructions, it
is unclear to me that anything like this is attested in clear
(morphological) cases of reduplication. If these compounds are not
reduplication, then they are actually counter-evidence to Inkelas'
Back-copying is the situation where both the base and the reduplicant
undergo a phonological process, despite the fact that only the
reduplicant is in the environment for the rule. This has of course been
cited as evidence for such non-derivational theories of phonology as
Optimality Theory (OT). Fiona McLaughlin gives an interesting
example of this phenomenon, from the Atlantic language Seereer-
Siin. The phonological process involves consonant mutation.
McLaughlin argues that back-copying in this case is not actually
phonological , but rather an instance of an affix consisting of a single
feature which is associated to both the base and the reduplicant.
McLaughlin then uses this analysis to argue for Inkelas' theory of
reduplication. While McLaughlin expresses her theory in terms of OT,
it would be interesting to see if her re-analysis could be recast in a
derivational theory of phonology.
Finally, Rajendra Singh discusses another compound noun
construction, this one in Hindi; the two nouns are constrained to be
synonymous (or antonymous), but non-identical. Elinor Keane
discusses echo reduplication in several south Asian languages,
including cases where the base may be smaller or larger than a word.
3. The Emergence of The Unmarked ('TETU'): Since McCarthy and
Prince (1994), the expectation among Optimality Theory (OT)
practitioners is that where the base and the reduplicant differ, the
reduplicant will be if anything less marked than the base. Laura
Downing discusses cases where the opposite is true—specifically,
cases where the reduplicant, but not the base, has a marked tone.
While her arguments seem sound, I was left wondering why
markedness in reduplication is only attested with marked tone.
Downing discusses this, suggesting that the autosegmental and
prosodic nature of tone explain this asymmetry. I am ready to believe
that these properties of tone account for the facts, but that potentially
removes the issue from the realm of theory. Putting this differently, it
seems to me that TETU can then be explained by the non-
autosegmental, non-prosodic nature of other features, leaving nothing
to be explained by theories such as OT.
4. Complete vs. partial reduplication: Nicole Nelson looks at cases
where a partial reduplicant surfaces on the ''wrong'' side of the base,
i.e. where a reduplicative suffix unexpectedly matches the first part of
the base, or a reduplicative prefix unexpectedly matches the last part
of the base. Nelson concludes that all such cases can be explained
away, for example by a phonological process which (in descriptive, not
theoretical terms) operates after full reduplication to trim off a portion
of the reduplicant.
5. Historical studies: Jason D. Haugen tries to reconstruct
reduplication in proto-Uto-Aztecan. It has often been argued that the
life cycle of reduplication begins with the repetition of complete words,
with the scope of reduplication eroding over time to yield partial
reduplication. Since the time depth of Uto-Aztecan is fairly high,
reconstructing reduplicative processes which are similar to the modern
day reduplicative processes presupposes that reduplication
morphemes in this family have not greatly changed during that time
(although one or another process has been lost in most languages of
the family). That is, while the synchronic similarities are clear, the
nature of the reconstructed processes in the proto-language might, it
seems to me, have been quite different.
Reijirou Shibasaki overviews the history of grammaticalization of
verbal reduplicants in Japanese, insofar as this can be pieced
together from extant corpora.
Leonid Kulikov sketches the issues surrounding reduplication during
the documented history of the Vedic language. The quantity of data
and analysis that has gone into this long extinct but well attested
language is matched by the variety of reduplicative devices. If all
languages could be documented and described so extensively, one
might worry less about language endangerment, at least from a
theoretical standpoint! I can only echo Kulikov's concluding remark
that students of reduplication would do well to look closely at the
evidence from Vedic.
6. Child language: It has been observed that reduplication often
arises in creole languages, despite being virtually unattested in pidgin
languages (see below). One way reduplication might arise ex nihilo is
child language, and several papers explore reduplication among first
language learners. Wofgang Dressler, Katarzyna Dziubalska-
Kolaczyk, Naalia Gagarina and Marianne Kilani-Schoch provide
examples of children producing reduplicants in a variety of languages
which (in their adult form) lack productive reduplication, perhaps as a
simplification of adult polysyllabic words. A theoretical account of this
stage might resemble accounts of reduplication in which an empty
syllable in a bisyllabic foot is filled by spreading of the melody from an
adjacent syllable, although Dressler et al do not put it in those terms.
Marie Leroy and Aliyah Morgenstern observed a child in the very early
stages of acquisition of French. They remark that this child seemed to
be a slow language learner at that point, but that he displayed a great
deal of reduplication. It is not however clear to me that this
reduplication has any bearing on the later use of reduplication in
morphology, since it is not clear at this early stage of language
learning that there is any morphology. That is, the child may simply be
repeating full ''words''.
Finally, Hatice Sofu looks at the acquisition of a partial reduplicant
affix in Turkish. This reduplicant is unusual in several ways: it is
virtually the only prefix in Turkish, and it includes a coda consonant
which is not part of the base. Which coda consonant is used in any
particular word, however, is difficult to pin down. Turkish grammars
give a number of rules for determining the appropriate consonant.
Sofu tested whether children had learned these rules by having them
reduplicate nonce words, and the children are far from consistent. But
as it turns out, adults exhibit even more variation in this task than
children do. One obvious possibility is that the putative rules do not
represent the adults' competence, and that the coda consonant has
simply been lexicalized.
7. Teleology: Suzanne Urbanczyk argues that phonological rules
sometimes function to render two otherwise identical processes of
reduplication non-homophonous. As she admits, the claim is difficult
to argue (as functionalist claims often are); in part, the difficulty lies in
establishing that something requires a functionalist explanation. But
there is also the question of what to ascribe the function or purpose
to. Does language change somehow have as its goal preventing
homophony? This seems unlikely; on the contrary, as has often been
noted, language change seems blind to such goals.
But if language change is blind to preventing homophony, in what
sense can a process be said to fulfill a function? At first glance, a
Darwinian parallel seems plausible, but it is far from clear what is
doing the selecting. Languages do not have more or fewer
descendents because they are more or less iconic or functional, nor is
it at all clear that iconic constructions are preferentially maintained in
daughter languages. Indeed languages seem to be replete with
unnecessary morphological baggage. Declension and conjugation
classes and grammatical gender have been retained through millennia
of Indo-European languages, but are familiar impediments to
generations of second language students of these same languages (a
point made by Carstairs-McCarthy 1994). In sum, if language change
has any goal, it seems unlikely to be that of making languages more
A different issue of teleology is mentioned in passing in the article by
Leroy and Morgenstern discussed above. These authors observe
that reduplication may serve to help in first language learning, by
enabling the child to discover phonological regularities, and to
temporarily avoid some of the complexities of polysyllabic words.
8. Iconicity: One might expect that reduplication would be iconic, i.e.
that its meaning would tend to imply ''more of the same''. For signed
languages, the case for the iconicity of reduplication seems
reasonably clear, as discussed by Ronnie B. Wilbur in her paper ''A
reanalysis of reduplication in American Sign Language.''
But the iconic motivation for reduplication is not always so obvious.
Silvia Kouwenberg and Darlene LaCharite address the fact that the
semantics of reduplication is sometimes anti-iconic, giving meanings
like ''reddish'' or ''biggish (but still growing)''. They begin with the fact
that for some predicates—this is most obvious with those implying
continuous action or states—''more'' means ''customarily'', while for
other predicates—punctual verbs, for instance—it takes on a
discontinuous meaning, such as ''repeatedly''. The distinction can
carry over to predicates which may imply a state that is localized,
e.g. 'yellow' can refer to something that is yellow spotted. But
something that is spotted with yellow is no longer yellow all over, and it
is a slight extension from 'yellow spotted' to 'yellowish'. Kouwenberg
and LaCharite thus claim that ''seemingly opposite interpretations of
reduplication both instantiate the iconic principle.''
Their story certainly seems plausible as an historical explanation, but
it leaves open the question of what the putative iconic principle is. Is it
a deep fact about language, as instantiated in human minds? Or is it
simply a pragmatic fact that some sorts of attributes (color, for
example) can be discontinuous, and therefore lend themselves
pragmatically to a discontinuous meaning, while others (weight, say)
are nearly always properties of continuous objects? Or to put this
differently, does the Iconic Principle have any use in first language
Werner Abraham tackles anti-iconicity from a more theoretic
perspective, although it appears that he reconsidered this position
before the paper was published. He raises the amusing question of
why reduplication is the only morphological process for which linguists
have attempted to find an iconic explanation; why not infixing, or
9. Creoles: Peter Bakker and Mikael Parkvall give an overview of
reduplication in creoles (for a book length discussion, see
Kouwenberg 2003, reviewed here (Linguist List 15.1689). They focus
on the fact that reduplication is virtually absent from pidgins, but
common in creoles. Under the assumption that creoles are
descended from pidgins (and that those pidgins lacked reduplication
in the same way that present-day pidgins do), the obvious question is
where reduplication in creoles comes from. One answer would be that
it is something children do naturally, and that creoles
are ''constructed'' by children (cf. the papers discussed above under '
Child language'). But Bakker and Parkvall lean towards a different
answer, namely that reduplication enters a creole through the
influence of adstratal languages, that is, it is borrowed in from other
languages spoken in the community while the creole is forming.
However, creoles usually have complete reduplication, whereas
reduplication in more ''mature'' languages (including presumably the
adstratal languages) is often partial. While Bakker and Parkvall note
this fact, they do not draw the implication that it might weigh against
the theory of adstratal borrowing. Also, the function of reduplication in
creoles is often—but not always—different from that of the presumed
adstratal (substratal) languages. In the end, of course, both
explanations may be correct; some creoles may have borrowed
reduplicative strategies (and meanings) from adstratal languages,
while other creoles innovated them.
Kouwenberg and LaCharite's paper (mentioned above) also concerns
10. Sign languages: Linguists should be thankful that sign languages
exist, because of the unique light they throw on human language.
Two papers discuss reduplication in signed languages. Roland Pfau
and Markus Steinbach on German Sign Language (GSL), and Ronnie
B. Wilbur on American Sign Language (ASL). Wilbur's article was
discussed above, under the topic of iconicity.
GSL exhibits at least two reduplication morphemes, which Pfau and
Steinbach dub 'backward' and 'sideward' reduplication. In backward
reduplication, the reduplicant sign is performed in the backwards
direction from the base sign; in sideward reduplication, the
reduplicants (of which there are two, i.e. this is a case of triplication)
are signed in a location to the side of where the base was signed.
(Both morphemes have ''phonologically'' conditioned allomorphs which
do not involve backward or sideward production.) Neither kind of
reduplication has been attested in spoken languages. In the case of
backward reduplication, this is presumably because spoken
morphemes consist of a sequence of phonemes, and human
languages do not, so far as we know, reverse such a sequence (nor
do sign languages reverse a sequence of signs). And for spoken
morphemes consisting of a single phoneme, it is unclear what it would
mean to reverse the direction of that phoneme. In contrast, a sign—or
a portion of a sign that involves movement in some direction—can be
easily reversed. As for sideward reduplication, it is not even clear
what it would mean to produce a spoken phoneme to one side;
perhaps the closest analogy would be to attach a different tone to the
reduplicant (something which is indeed attested, as discussed in
11. Other issues: Francoise Rose investigates reduplication in
Emerillon, a Tupi-Guarani language. One of the issues, too often
taken for granted, is how (or perhaps in some cases, whether) one
can distinguish the reduplicant from the base.
In an earlier review on Linguist List (http://linguistlist.org/issues/12/12-
2062.html) I commented on a study of reduplication in which the
corpus was not large enough — it did not contain enough examples of
reduplication—to answer some critical questions. The issue surfaces
in some of the studies here, too. In particular, Keane comments
that ''the sample size…is obviously too small to draw any hard and fast
conclusions.'' Shibasaki makes a similar point: a particular pattern of
reduplication is known from modern Japanese, but not attested in the
corpus which he was working with. This is a critical issue for field
linguists. It is also critical to know whether any particular reduplicative
construction is productive, something which is not always made clear
in descriptions (or even in theoretical analyses).
Finally, I will comment on the price. At 128 euros, I suspect few
linguists will be tempted to buy this book. That is perhaps unfortunate
for the authors, whose contributions will not receive as great
dissemination as they otherwise would. There are, of course,
alternative ways to distribute conference papers. The publisher's
website does contain a list of authors, with their addresses
Kouwenberg, Silvia (ed.). 2003. Twice as Meaningful: Reduplication
in Pidgins, Creoles and other Contact Languages. Westminster
Creolistics Series 8. London: Battlebridge.
Carstairs-McCarthy, Andrew. 1994. ''Inflection Classes, Gender, and
the Principle of Contrast.'' Language 70:737-788.
McCarthy, John J., and Alan Prince. 1994. ''The Emergence of the
Unmarked: Optimality in Prosodic Morphology.'' Mercè Gonzàlez, ed.,
Proceedings of the North East Linguistics Society 24, GLSA, Amherst,
MA. Pp. 333-379. Available from the Rutgers Optimality Archive at
| ABOUT THE REVIEWER:
ABOUT THE REVIEWER
Mike Maxwell works on computational morphology and low density
languages for the Center for the Advanced Study of Language at the
University of Maryland. He has a Ph.D. in linguistics from the
University of Washington.