japanese concessive reduplication

27
Concessive Reduplication in Japanese or, “It’s true it’s reduplication, but…” Sky Onosson 1. Introduction Languages possess, among other devices, certain constructions which, in addition to providing a proposition, allow the speaker to express their own attitude towards the proposition itself. For example, there is an ‘indifference’ construction in Persian (Sadat-Tehrani, 2002) which expresses the speaker’s attitude that the facts of the proposition are unimportant to them, at the same time as expressing the proposition itself. English pejorative shm-reduplication (Nevins & Vaux, 2003, and Grohmann & Nevins, 2004) expresses the attitude that the proposition (or often simply an entity) expressed within the construction is not only unimportant, but contemptible. In order to account for the semantic properties of such constructions, Sadat-Tehrani proposes that, within Jackendoff’s Parallel Architecture framework, a ‘Modality Tier’ exists within the conceptual component of a construction which “encodes the type of propositional attitude that an utterance contains,” (Sadat-Tehrani, 2002: 64). Accepting this framework, we should expect to encounter various other types of propositional attitudes, besides just the Persian and English examples cited, which would be represented by various encodings within this tier of modality. An interesting contrast to explicit expressions of the speaker’s attitude, such as the ‘indifference’ and ‘pejorative’, is a construction which may be taken as actively withholding the speaker’s attitude towards the proposition: the concessive. The essence of the propositional content of the concessive is that the speaker will admit only that the proposition being conceded is true, and will commit to no more than that, either in terms of facts or, more importantly, in terms

Upload: cornelia-badulescu

Post on 11-Apr-2017

219 views

Category:

Education


0 download

TRANSCRIPT

Page 1: Japanese concessive reduplication

Concessive Reduplication in Japanese

or, “It’s true it’s reduplication, but…”

Sky Onosson

1. Introduction

Languages possess, among other devices, certain constructions which, in addition to

providing a proposition, allow the speaker to express their own attitude towards the proposition

itself. For example, there is an ‘indifference’ construction in Persian (Sadat-Tehrani, 2002) which

expresses the speaker’s attitude that the facts of the proposition are unimportant to them, at the

same time as expressing the proposition itself. English pejorative shm-reduplication (Nevins &

Vaux, 2003, and Grohmann & Nevins, 2004) expresses the attitude that the proposition (or often

simply an entity) expressed within the construction is not only unimportant, but contemptible. In

order to account for the semantic properties of such constructions, Sadat-Tehrani proposes that,

within Jackendoff’s Parallel Architecture framework, a ‘Modality Tier’ exists within the conceptual

component of a construction which “encodes the type of propositional attitude that an utterance

contains,” (Sadat-Tehrani, 2002: 64). Accepting this framework, we should expect to encounter

various other types of propositional attitudes, besides just the Persian and English examples cited,

which would be represented by various encodings within this tier of modality.

An interesting contrast to explicit expressions of the speaker’s attitude, such as the

‘indifference’ and ‘pejorative’, is a construction which may be taken as actively withholding the

speaker’s attitude towards the proposition: the concessive. The essence of the propositional

content of the concessive is that the speaker will admit only that the proposition being conceded is

true, and will commit to no more than that, either in terms of facts or, more importantly, in terms

Page 2: Japanese concessive reduplication

126.792 Final Paper - 2 - S. Onosson

of their own attitude towards the matter. “I admit it’s true, but…” conveys the message that the

speaker is unmoved by the truth of the matter under discussion, and will continue to hold whatever

opinion they wish regarding it, however contrary it may seem. Opinions contrary to the conceded

facts are generally expressed in an immediately adjacent clause, the pragmatic usage of the

concessive usually being a means to introduce such opinions in a softer, more agreeable manner.

Hence the term ‘concession’, something offered to one’s opponent in a negotiation. One can see

this quite clearly in the English periphrastic concessive I’ll grant/give you [ that X is true ], which

places the concession [ that X is true ] into a metaphorical exchange between the discourse

participants. Discourses can be understood in many cases as forms of negotiation, and the ability to

offer linguistic concessions would therefore seem to be a necessary feature of any human language.

There exists in the Japanese language a concessive construction which takes the outward

appearance of a topicalized nominal within a predication which, when used to expressive

propositional concession, behaves like a (grammaticalized)1 reduplicative construction. This paper

will be an examination of this particular construction, referred to herein as ‘Japanese concessive

reduplication’ (JCR), or ‘the koto wa concessive’, after the two morphemes involved in the

construction. The general form of JCR will be described in section 2, along with a comparative

look at the three distinct superficial elements involved in the construction: predication,

nominalization, and topicalization. The semantics and discoursal properties of concession as

involved in JCR will be discussed in section 3. An analysis of JCR as reduplication will be made in

section 4, and conclusions and references will follow.

1 While it does seem at least probable that the koto wa concessive has been grammaticalized from the combination of a topicalized nominalization and a predication involving reduplication, this paper will not explore the diachronic issues concerning the source of this construction, other than in examining the general facts regarding predication, nominalization, and topicalization in Japanese.

Page 3: Japanese concessive reduplication

126.792 Final Paper - 3 - S. Onosson

2. The Japanese koto wa concessive

2.1 General description of the koto wa concessive

The basic form of JCR can be seen in the following example, adapted from examples cited

in Okamoto (1990):

1) kono okasi wa oisii koto wa oisii kedo, takai yo. this cake T2 tasty NM T tasty but, expensive PRT “This cake is tasty, but expensive.”

JCR must contain the two morphemes koto wa, glossed as nominalizer and topic-marker,

respectively. Koto wa is preceded by a predicate construction of some kind, which may optionally

include the topic or subject (in this example, kono okasi ‘this cake’) and some other elements, and

always followed by a reduplication of the predicate itself (not including the verbal arguments or

other elements in the first clause). As well, this is normally, but not always, followed by one of a

set of sentence-linking morphemes, including kedo, noni, ga and others, which have the same

function as English but, however, etc., and also a following clause which in some way contradicts the

first clause, or runs counter to the expectations that might be inferred from the first clause. As

noted, these last elements are not always overtly present, but where absent they are inferable from

context, and the purpose of JCR seems to be to draw attention to the second clause, or to direct

the listener to actively infer it.

Okamoto (ibid.) notes that there are a few adverbial elements, such as tasika ni ‘certainly’

which may intervene between koto wa and the reduplicated predicate, but that no arguments of the

predicate or any other items from it may be repeated in the reduplication. Leaving these aside last

details for the time being, we may formulate a preliminary description of JCR as follows: 2 Abbreviations used: T = topic, NM = nominalizer, PRT = particle, 1s = first person singular pronoun, NOM = nominative case, ACC = accusative case, PST = past tense, GEN = genitive, COP = copula, NEG = negative, SUF = suffix (unglossed), HON = honorific, CAUS = causative.

Page 4: Japanese concessive reduplication

126.792 Final Paper - 4 - S. Onosson

2a) Form of JCR: [ (V) [ (W) X ] koto wa X (Y) ] ( [Z] )

2b) Minimal obligatory form of JCR: [ [ X ] koto wa X ]

In this formula (2a), X is a clausal predicate (which, in Japanese, may be a verb or one of two

classes of adjective), W includes the arguments of the predicate and V other optional elements

within the clause, Y is a member of a set of clause-final words/particles, and Z is another clause.

Clausal structure is indicated by square brackets. Ignoring the optional elements V, W, Y and Z,

we can see that the basic form of JCR is bi-clausal, as shown in (2b). Further discussion in section 4

will demonstrate that the formula in (2b) is in fact the best description of JCR, and that the other

elements in (2a) are really external to it, and uninvolved in the construction itself.

Taking into account the nature of koto as a nominalizer and wa as a topic-marker, we can

see that the overt appearance of JCR is that of a predication made upon a topicalized nominalization

of another predicate, and that both predicates in the two clauses are essentially identical (barring

any inflectional differences). However, JCR does not in fact function as a predication, but instead

serves to convey the propositional attitude of concession on the part of the speaker towards the

predication made in the interior clause. This is one of the main arguments for treating JCR as a

reduplicative construction with its own syntactic and semantic structure, and not as a compositional

structure made up of nominalization + topicalization + predication, which will be discussed further

in section 4. The following subsections will take up individually the subjects of: predication (2.2);

nominalizations including and other than koto (2.3); and topicalization (2.4). Together these will

illustrate that JCR is syntactically non-compositional in nature, despite being superficially

comprised of these disparate ‘component’ elements.

Page 5: Japanese concessive reduplication

126.792 Final Paper - 5 - S. Onosson

2.2 Predication in Japanese

Japanese is an SOV language characterized both by the presence of grammatical relations,

and prolific use of topicalization. As a result, predication can be understood in at least two ways.

First, in parallel to many other languages, predication involves a subject and a predication made of

it. Predications in Japanese involve both verbs (I include here suffixal ‘auxiliary verbs’ and the

copula, which Shibatani (1990) treats separately) and adjectives (including together both inflecting

‘canonical adjectives’ (Nishiyama, 1999) and non-inflecting ‘nominal adjectives’). These inflecting

forms of speech in Japanese (non-inflecting ‘nominal adjectives’ attach to the copula, which itself

inflects) are the only categories which select arguments, and so permit case-assignment and the

identification of grammatical roles characteristic of subject-oriented predication.

The other kind of predication is predication of a topic. This is usually discussed in the

literature as topic-comment structure, in those languages where topic is a relevant concept.

However, I believe it is useful to consider topic-comment on par with predication, and especially

so in a language like Japanese which mixes grammatical relations with topicalization so

considerably. Consider the following sentences:

3a) Watasi ga kono hon o yon-da. 1s NOM this book ACC read PST “I read this book.”

3b) Watasi wa kono hon o yon-da. 1s T this book ACC read PST “Me, I read this book.”

3c) Kono hon wa watasi ga yon-da. this book T 1s NOM read PST “This book, I read.”

3d) Kinoo wa, watasi ga kono hon o yon-da. yesterday T 1s NOM this book ACC read PST “Yesterday, I read this book.”

Page 6: Japanese concessive reduplication

126.792 Final Paper - 6 - S. Onosson

Example (3a) shows a typical two-place predicate (as per Adger, 2003: 78) in Japanese, with case-

marked arguments of the verb but no topic-marking. Clearly, this is not a topic-comment

construction. Topicalization of the subject and object, in (3b) and (3c) respectively, are usually

analyzed as having topic-comment structures. However, it may be argued that this is not

necessarily the case. Beninca and Poletto (2004) argue, based on Italian data, that many cases of

topic are in fact better identified as focus. In order to distinguish proper topics (those that fall under

TopP and not FocP), they propose that “[a]ll the elements in the Topic field share at least two

properties: (a) they are not related to a variable in the clause, differently from elements belonging

to the Focus field; and (b) they are all ‘known information’ in some sense,” (Beninca and Poletto,

2004: 63). In this sense, then, being ‘known information’ is simply not enough for an item to be

categorized as topic. In (3d) we see an example which does fulfill both of Beninca and Poletto’s

criteria: kinoo ‘yesterday’ is not an argument of yonda ‘read (past), thus fulfilling criterion (a), and it

can be understood as (always) being known information, thus fulfilling criterion (b). In contrast,

(3b) and (3c) contain ‘topics’ which are clearly derived from an argument of the verb (cf. 3a), and

thus do not meet criterion (a). According to Beninca and Poletto’s analysis, each of these ‘topics’

should be better taken as the focus of their respective sentences, and Japanese wa best understood as

a marker of either topic or focus. Certainly this is a controversial point of view, but it does help to

explain some interesting facts about Japanese, such as the fact that a sentence may contain more

than one instance of wa:

4) Kinoo wa, watasi wa kono hon o yon-da. yesterday T 1s T this book ACC read PST “Yesterday, me, I read this book.”

Taking Beninca and Poletto’s view, kinoo ‘yesterday’ in (4) should be understood as a true topic (in

other words a ‘high’ topic, generated under TopP and not moved from lower in the structure),

Page 7: Japanese concessive reduplication

126.792 Final Paper - 7 - S. Onosson

while watasi ‘I’ is in fact a focus, as it is clearly the subject of yonda ‘read (past)’3. While a thorough

examination of the applicability of Beninca and Poletto’s theory to Japanese is well beyond the

scope of this paper, we may consider their point to be valid enough to consider examples such as

(3b) and (3c) to be good examples of predication, as opposed to a separate kind of structure such as

topic-comment, and thus on par with basic predications such as (3a), while a structure such as that

in (3d) may be of a different kind, and thus not completely comparable.

2.3 Nominalization in Japanese

There are several ways in which a nominalized form may be created in Japanese. The two

most important ones identified by Martin’s (1975) seminal work, and most relevant to the

discussion herein, are no and koto. Martin discusses them thusly: “Two very common

postadnominals are used to nominalize a sentence: koto makes a general nominalizations that is

abstract, habitual, or remote; no makes a specific or definite nominalization that is single,

immediate, concrete, or directly perceivable,” (Martin, 1975: 841). The following example,

adapted from Martin (ibid.), indicates the distinction:

5) Soo kangaeru no wa toozen no koto da. Thus think NM T natural GEN NM COP ‘It is natural to think so.’, lit. ‘Thinking thus is a natural thing.’ 3 It may be the case that wa is simply a marker of ‘givenness’, and that word order is a better indicator of topichood. In (4), the order where the ‘true’ topic kinoo wa precedes watasi wa is better than the reverse: ?? Watasi wa kinoo wa kono hon o yonda. This is discussed further in section 2.3 and 2.4. Where the focus is not given/known information, a case-marking particle, perhaps in combination with word order, may be the preferred way to indicate its status. Iwasaki (1985) identifies nominative ga as a frequent indicator of ‘new’ information, while Lee (2002) argues that the relative ‘activeness’ of information is related to retention vs. optional deletion of case-marking, showing that ‘inactive’ subjects always bear ga, while ‘semiactive’ and ‘active’ subjects permit its deletion. Ono et al (2000) also argue that the usages of both ga and the accusative marker o are better understood in terms of pragmatics rather than grammatical relations, and that one of the major factors is the marking of new or unpredictable concepts. Suzuki and Ono (1991) suggest that ga is used for information in ‘the primary focus of consciousness’. Clearly, the function of case-marking in Japanese is a matter of great debate.

Page 8: Japanese concessive reduplication

126.792 Final Paper - 8 - S. Onosson

As can be seen in this very example, no has at least two distinct uses. One is to nominalize a

predicate, as the first no in (5) nominalizes soo kangaeru ‘think so/thus’ into soo kangaeru no wa

‘thinking so/thus’. Secondly, it acts as a genitive case-marker, which in this example is used to

connect the nominal-adjective toozen ‘natural’ with the abstract noun koto ‘thing’: toozen no koto ‘(a)

natural thing’. The genitive nature of no can most clearly be seen in a possessive construction

involving a pronoun:

6) watasi no hon 1s GEN book ‘my book’

Koto, on the other hand, is never used as any kind of a case marker, but originates from a common

noun which, as already discussed, means simply ‘thing’. Its usage as a nominalizer is in fact very

much like a common usage of thing in colloquial English, also present in some lexicalized

expressions such as living thing. Thing in English has a fairly broad range of usage, attaching to (and

creating) nouns and verbs, while koto is more restricted in attaching to predicates only. There is a

counterpart to koto which covers the other range of the territory of English thing: mono ‘thing’.

Raud discusses the difference between koto and mono: “[b]oth these words also function as what can

be called markers of nominalization, mono to denote the subject or object of an act, koto to express

the act itself,” (Raud, 2002: 100). Drawing on historical discussions in the Japanese linguistic

tradition, he states that “koto [is] comparable to propositions that may always be divided into the

subject and the predicate,” (ibid, p. 101). As an illustration of the differences between the two, he

sums it up thusly: “[t]he act of beating somebody up is a mono, the actual beating a koto,” (ibid., p.

100).

In relation to the discussion in section 2.2, Shibatani identifies an important contrast

regarding sentences involving koto nominalization: “sentence nominalization … takes in the

Page 9: Japanese concessive reduplication

126.792 Final Paper - 9 - S. Onosson

nominative phrase, but leaves out the topic phrase,” (Shibatani, 1990: 272). The following

examples adapted from Shibatani (ibid., (17a) and (17b)) illustrate this difference:

7a) [ Hanako ga sin-da ] koto o sir-ana-katta. Hanako NOM die-PST NM ACC know-NEG-PST ‘X did not know that Hanako died.’

7b) Hanako wa [ sin-da ] koto o sir-ana-katta. Hanako T die-PST NM ACC know-NEG-PST ‘Hanako did not know that X died.’

In (7a), the nominative Hanako is interpreted as the subject of the nominalized predicate sinda koto

‘that (X) died’, and therefore cannot be coreferential with the subject of the predicate siranakatta

‘did not know’. In (7b), Hanako is marked by wa, and as such cannot be coreferential with the

subject of sinda koto, instead being interpreted as the subject of siranakatta. Okamoto cites a parallel

example which points to an interesting contrast (Okamoto, 1990: 251):

8a) [ Taroo ga ano tegami o kai-ta ] koto wa tasika-da Taro NOM that letter ACC write-PST NM T certain-COP ‘It is certain that Taro wrote that letter.’

8b) * Taroo wa [ ano tegami o kai-ta ] koto wa tasika-da Taro T that letter ACC write-PST NM T certain-COP Intended (1): ‘Taroi, it is certain that (hei) wrote that letter’ Intended (2): ‘Taroi is certain that (Xj) wrote that letter’

Because in (8b) the wa-marked proper name cannot be interpreted as referring to the subject of the

nominalized predicate ano tegami o kaita koto ‘that X wrote that letter’, the first intended sense of

the utterance is unavailable. Also, since tasika-da ‘be certain’ cannot be predicated of an animate

(contrast with English I am certain that…), the other intended sense is unavailable, and the entire

utterance is rendered ungrammatical.

The fact, now established, that topicalization is prevented from occurring from within a

nominalized predicate is thus established, and poses a puzzle for analysis of JCR. Consider example

(1), presented here again as (9):

Page 10: Japanese concessive reduplication

126.792 Final Paper - 10 - S. Onosson

9) kono okasi wa oisii koto wa oisii kedo, takai yo. this cake T tasty NM T tasty but, expensive PRT “This cake is tasty, but expensive.”

If we consider both instances of oisii ‘tasty’ as true predicates, then the only reasonable

interpretation of this sentence is that the topic-marked kono okasi ‘this cake’ is the subject of both of

them. However, this stands in stark contrast to (7b) and (8b), where a topic-marked constituent

can not be interpreted as the subject of a nominalized predicate such as oisii koto.

Based on this evidence, the most reasonable conclusion is that JCR cannot be identified as a

nominalized predicate construction. The first ‘predicate’ in JCR does not behave like a usual

predicate, in that a topicalized constituent preceding it can be referential to one of its arguments.

One might argue that such a topic can be related to an argument of the second non-nominalized

predicate, and it is then simply the identity between the two predicates that allows the inference of

referentiality to an argument of the interior, nominalized predicate. But this, too, seems to be

incorrect. Consider the following example cited by Okamoto (1990: 251, (15)):

10) Boku wa soko e it-ta koto wa it-ta 1s T there to go-PST NM T go-PST ‘It is true that I went there.’

If we are to interpret this as a construction on par with (7) and (8), above, then we must conclude

that boku wa ‘I’ is an argument of the second, higher instance of itta ‘went’. The problem with this

concerns the relative order of the two constituents marked with wa in this sentence. Recall from

the discussion in section 2.2 that, in a double-wa sentence, it is possible to have a ‘true’ topic such

as kinoo ‘yesterday’ precede an argument-derived topic, as in example (4), presented again here:

11) Kinoo wa, watasi wa kono hon o yon-da. yesterday T 1s T this book ACC read PST “Yesterday, me, I read this book.”

Page 11: Japanese concessive reduplication

126.792 Final Paper - 11 - S. Onosson

However, reversing the order of the wa-marked constituents, while not ungrammatical, is

pragmatically odd when taken out of context:

12) ? Watasi wa, kinoo wa, kono hon o yon-da.

At the very least, (11) is considerably less marked an utterance than is (12). The two variations are

nonetheless both possible, and grammatical. In example (10), however, which also contains two

instances of wa, there is no variation possible. * Soko e itta koto wa boku wa itta and * Itta koto wa boku

wa soko e itta, where the interior predicate (either including or excluding the directional

postpositional phrase) precedes the subject-derived ‘topic’ are both equally ungrammatical.

The best conclusion that can be made, based on the evidence given, is that the entire JCR

construction acts as a single predicate, and cannot be equated with other superficially similar

nominalized constructions involving koto.

2.4 Topicalization in Japanese

The last factor to consider in discussing the elements that make up JCR, is topicalization.

In relation to the discussion on predication in 2.2, it is already apparent that a simple notion of

‘topic’ is not the best way to analyze the morpheme wa, and so it is rather presumptuous to refer to

JCR as involving topicalization at all. Numerous authors suggest that the terms ‘topic’, ‘given’,

‘active’ etc. are not sufficient to account for all uses of wa. Hinds (1987) points out numerous

instances where wa and ga may alternate for rhetorical effect. Makino (1987) argues that wa, in

contrast with ga, is ‘listener-oriented’, which means “communication … directed towards the

listener/reader (which is the ‘unmarked’ communicative orientation),” (Makino, 1987: 293).

Clancy and Downing (1987), analyzing a series of transcribed texts, conclude that “the primary

function of wa is to serve as a local cohesive device, linking textual elements of varying degrees of

contrastivity,” (Clancy & Downing, 1987: 46).

Page 12: Japanese concessive reduplication

126.792 Final Paper - 12 - S. Onosson

From this, even if we consider the morpheme wa within JCR to contribute at least in part

towards a compositional structure for the construction, there is little that can be concretely and

definitively stated about its semantic and/or discourse function. Instead, I will consider the role of

wa structurally, in terms of a syntactic position. Beninca and Poletto argue that multiple positions

for TopP are not possible, and that the structure of the “Topic/Focus portion of the CP layer”

(Beninca & Poletto, 2004: 54), cross-linguistically, may be formulated as:

13) [TopicP [FocusP [IP ] ] ]

Without delving into their argumentation any further (see also section 2.2), I will assume this

structure to be the correct one, and propose that wa marks constituents either in TopP or FocP, the

difference between the two (again, following Beninca and Poletto) being that constituents in FocP

are related to variables (equivalent to, for our purposes, raised from) within IP, while TopP

constituents are not, a distinction which is not indicated in Japanese. I will continue to use the term

‘topic’ hereafter, with the caveat that this term simply refers to constituents in Japanese marked

with wa, whether they are actually functioning as topic or focus in a given discourse.

Evidence has already been given that ‘true’ topics, unrelated to argument structure in the

IP, are preferred in initial positions, with ‘raised’ topics being preferred in a secondary position,

within a clause. While Japanese does have rather free word order with regards to constituents

preceding the verb, the unmarked order is definitely ‘High’ Topic > ‘Raised’ Topic > Subject > Other

Arguments and Peripheral Constituents, and so one may consider other, alternative orders to be

pragmatically licenced in some way. We have seen that a typical JCR clause, such as (9) or (10)

above, often contains a wa-marked initial topic, as well as the ‘predicate’ preceding koto wa, and the

main predicate immediately following and adjacent to koto wa. And yet, unlike we might expect

given the hierarchy just cited, this word order involving a wa ‘topic’ situated directly adjacent to

the predicate is not a marked structure, and in fact, as has been shown, cannot be varied. This

Page 13: Japanese concessive reduplication

126.792 Final Paper - 13 - S. Onosson

strongly suggests that, whatever the nature, function, and structure of wa in a typical Japanese

sentence, it is rather different from what is happening in JCR, and that the morpheme wa present in

JCR should not be considered to be an indication of topic, focus, or anything else which wa may

indicate in other usages.

3. Semantic and discoursal properties of the Japanese koto wa concessive

Aside from the overt syntactic/morphological elements which make up JCR, discussed in

detail in section 2, there is the semantic component to consider as well. Given that it has been

established throughout the preceding that the overt characteristics of JCR are not in any way

indicative of its syntactic status and properties, we shall not attempt to link its semantic

characteristics to those of koto or wa. Indeed, the semantics of JCR are completely unrelated to

those of its overt components.

The usage of JCR can only be understood from within a particular discourse context. This

is true on several levels. The predication involved in JCR must be, in some respect, given

information. It is not acceptable to utter a JCR construction, such as any of the examples cited

above, ‘out of the blue’ with no context for the listener to draw on. As always, context is a

variable thing, and it may not be necessary in all cases for givenness to be overtly established by a

linguistic device in a discourse. Shared knowledge and experience will play a role here, and

anything within that realm or which the speaker may reasonably (or perhaps unreasonably?) assume

the listener to be able to infer or recall in some way might be a good candidate for JCR. Anything

beyond this, however, is simply impossible, if not strictly speaking ungrammatical.

Beyond the possibilities revolving around the given status of a predication in JCR, there is

also the discoursal positioning of JCR itself. JCR does not make a good ‘opener’ in a discourse. It

functions as an act of concession, whereby the speaker actively concedes a point of fact or opinion

Page 14: Japanese concessive reduplication

126.792 Final Paper - 14 - S. Onosson

to the other discourse participant. Clearly, this will not work very well if there is not an ongoing

discourse. There are of course many possibilities as to what constitutes a discourse. One can easily

imagine a conversation continuing on from some point in the past, be it minutes, hours, even days

earlier, and thus the possibility of a seemingly incongruous JCR utterance opening up such a

continuation of a discourse. But a true non sequitir JCR must be considered ungrammatical, and

will not be understood in such a context, any more than it would make sense for an English speaker

to walk up to a total stranger and say “It’s true that there’s only a small chance of rain today, but I

still decided to bring my umbrella.”

These facts regarding JCR mark it as a different kind of construction than nominalizations

and topicalizations in Japanese. The semantics are in no way relatable. Nominalizations do not

suffer from restrictions on their givenness nor their discoursal position. Only the nature of topic

may be related in some way to JCR as topics, too, are restricted in certain ways regarding where

they may occur in a discourse. But here the connection must be that both JCR and topics in general

make reference to given information, however that is construed in a specific context, and it is only

this characteristic which unites them, rather than the presence of wa in JCR which, as has been

shown (section 2.4), does not otherwise function in a characteristic manner within JCR. We may

safely conclude that the nature of the semantics and discoursal properties of JCR are unique to it,

and not composed from the properties of its ‘components’.

A further point which may be made here is that JCR is not completely unique cross-

linguistically. There are other semantically aberrant (non-compositional) constructions involving

reduplication which carry a specific discoursal function. Sadat-Tehrani’s (2002) analysis of the

Persion ‘Indifference-ke Construction’ (IKC) is relevant here. Sadat-Tehrani analyzes this

construction specifically as a reduplication, within the framework of Parallel Architecture. This

framework allows for the adoption of a ‘Modality tier’ wherein the particular discoursal

Page 15: Japanese concessive reduplication

126.792 Final Paper - 15 - S. Onosson

characteristics of the IKC (the expression of indifference on the part of the speaker) may be

expressed, as these characteristics are entirely unrelated to the component parts of the IKC, just as

the semantic and discoursal nature of JCR are completely unrelated to its component parts (aside

from the common nature of givenness noted for both JCR and topicalization in general). English

shm-reduplication, as described by Nevins & Vaux (2003) and Grohmann & Nevins (2004), carries a

similar yet distinct propositional attitude, that of pejoration on the part of the speaker towards the

referent of the reduplication. Saeed Ghaniabadi (personal communication) has also noted that

echo-reduplication in Persian may carry a propositional attitude for some speakers, the same as that

of English shm-reduplication. There are thus precedents, in several reduplications in non-related

language, for our treatment and analysis of JCR as a reduplication focused on the expression of

propositional attitude.

4. The koto wa concessive as reduplication

4.1 Factors in treating the koto wa concessive as reduplication: preliminaries

Kazutaka (2005), in an incomplete listing, identifies three kinds of reduplication in

Japanese: ‘renyookei’ reduplication, plural reduplication, and mimetic reduplication. These are all

word-level processes, with semantic associations typical of reduplications cross-linguistically. None

of the major works on Japanese, such as Martin (1976) and Shibatani (1990), address JCR amongst

other reduplications. In fact, JCR has not been noted in any great depth as a reduplication at all.

Martin (1976: 843-4) makes note of it briefly, referring to it as the ‘Iterative Concessive’ with only

a succinct description and no in-depth analysis. Okamoto (1990), in the only English-language

discussion of the construction of any length, while referring to it as reduplication, merely (but

importantly) addresses its non-compositional nature, taking its reduplicative status for granted.

Page 16: Japanese concessive reduplication

126.792 Final Paper - 16 - S. Onosson

Two important points established in the preceding sections, however, point towards the

appropriateness of dealing with JCR as reduplication. These are: 1) JCR is non-compositional (see

section 2); and 2) JCR is comparable to other reduplications cross-linguistically (see section 3). I

believe that these two points, taken together, justify a treatment of JCR as a reduplication.

The first point is relevant in regards to the classic treatise by Moravcsik (1978) on

reduplication. That author identifies one of the fundamental characteristics of any reduplication,

which, to paraphrase, is that equivalence in form does not equate to equivalence in meaning.

Reduplications, by their very nature, involve a phonological (and/or morphological and/or

syntactic) repetition of some form, but the semantic import of this repetition can not be related to

the meaning of the repeated form itself. Thus, reduplication is characteristically and fundamentally

non-compositional in nature.

The second point is equally important. Many reduplications have been shown to cover

much of the same semantic territory cross-linguistically (ibid.): for example, plurality (of some

kind) has long been noted to be one of its major semantic functions. Discoursal/modal functions,

on the other hand, have not been generally established as one of the major functions of

reduplication cross-linguistically, but the fact that such usages have begun to be recognized, and

especially in non-related (to Japanese) languages such as Persian and English, allows for at least the

acceptability of proposing a reduplicative analysis of JCR, itself having a function primarily

discoursal in nature. In the absence of this kind of cross-linguistic evidence, such an analysis would

at least be less tenable, if not outright inappropriate. As it stands, we may be satisfied that JCR, as

a reduplication, does not belong to a class unique to itself.

The general form of JCR was made early on in section 2.1, repeated here as (14):

14a) Form of JCR: [ (V) [ (W) X ] koto wa X (Y) ] ( [Z] )

14b) Minimal obligatory form of JCR: [ [ X ] koto wa X ]

Page 17: Japanese concessive reduplication

126.792 Final Paper - 17 - S. Onosson

While (14a) includes many items generally present in many concrete instances of the JCR, the

formula in (14b) makes a more proper general characterization of the construction, excluding those

items not properly part of the construction itself, however often they may actually occur alongside

it. There are two explanations for this, relating firstly to items V and W, and secondly to items Y

and Z together.

The position labeled W in (14a) subsumes the arguments of the base predicate (the first X),

such as soko e ‘(to) there’, an argument of it-ta ‘went’ in example (10), repeated here as (15):

15) Boku wa soko e it-ta koto wa it-ta 1s T there to go-PST NM T go-PST ‘It is true that I went there.’

Also potentially present at position V in (14a) are topicalizations, whether raised from arguments as

in Boku wa ‘I (topic)’ in 15, or otherwise generated (see the discussion in section 2.4). Elements in

either position V or W are typical of non-JCR clauses in Japanese. Also typically, topics and verbal

arguments are often optionally omitted in discourse. Thus, the syntactic and discoursal

characteristics of the items in positions V and W are neither unique to JCR, nor do they acquire any

special status when used in JCR. They retain all their typical qualities, including their relationship

(or lack thereof) to the predicate, and the status of their optionality (in general, related to their

status vis-à-vis given information). Therefore, it seems unnecessary and incorrect to include these

elements, either properly or improperly, within a description of JCR, as everything about them

falls out from general principles of Japanese grammar unrelated to JCR. Therefore, our

description of JCR will exclude them, as in (14b).

The final two elements which deserve some explanation are labeled Y and Z in (14a).

These are indeed related to each other, and furthermore to the discoursal function of JCR, and yet

should not be included within a correct description the construction itself. Example (1) is re-

presented in (16) to illustrate:

Page 18: Japanese concessive reduplication

126.792 Final Paper - 18 - S. Onosson

16) kono okasi wa oisii koto wa oisii kedo, takai yo. this cake T tasty NM T tasty but, expensive PRT “This cake is tasty, but expensive.”

The clause at the end, takai yo ‘(It’s) expensive’, comprises the position labeled Z in (14a). This is a

‘tag’ clause which is either present alongside or implied by JCR, and serves as a counterpoint to the

discoursal concession made by the use of JCR. The function of JCR is really to introduce this

second clause, albeit in a soft, polite manner, hence its concessive nature. The point of using JCR

is, therefore, all about this second clause. But we should not confuse the nature of this relationship

by identifying a clause at Z as a part of JCR. The nature of such a clause is in no way tied to or

restricted by JCR, a caveat to this being that the clause at Z should not logically follow from the

clause at X. This, however, has nothing to do with the Japanese-internal characteristics of JCR, but

is instead a function of its concessive nature. English, too, follows the same restriction, as seen in

the oddness of a statement such as ?? “It’s true it’s raining, but it’s wet outside.” Such a statement

fails to function as a concession precisely because it is non-argumentative by being patently true.

Without an argumentative follow-up, the point of making a concession is entirely lost. Following

from this, both JCR and English concessives require the second clause to be contradictory in some

way to the first clause (and thus argumentative). Otherwise, the discoursal function of concession

is lost. In English, “It’s true it’s delicious, but it’s cheap” may make sense as a simple statement, if

not as a concession. The Japanese JCR equivalent *Oisii koto wa oisii kedo, yasui yo, however, is quite

ungrammatical. This is due to the nature of the JCR precisely as a concessive construction – it

makes no sense with any other implied meaning. The English “It’s true…” concessive is not a

grammaticalized construction, and so an English speaker is free to utilize it in a variety of usages

unavailable to JCR. These facts regarding the nature of concession and the second clause in a

concessive construction point towards excluding position Z in (14a) from a description of JCR

itself, although JCR may be accurately described as having the purpose of introducing such a clause.

Page 19: Japanese concessive reduplication

126.792 Final Paper - 19 - S. Onosson

This brings us to position Y, which includes elements such as kedo ‘but’ in (16) and other

similar morphemes which function to join two contradictory clauses together. From the preceding

discussion, this position can be seen to fall out naturally from the juxtaposition of the two clauses,

JCR and its antecedent clause, and general principles of Japanese grammar which allow for a variety

of clause-final ‘particles’. Only those particles which serve as clause linkers and which carry

contradictory semantics (comparable to English but, however, etc.) are allowable in this position,

which is a characteristic not of JCR but rather of the nature of the relationship between the two

clauses between which Y intervenes, as can be seen from precisely similar restrictions regarding but,

however, etc. in English. As Z should not be included in a proper description of JCR, neither should

Y which introduces Z. Rather, they are both present alongside JCR because of the nature of

linguistic concession itself, JCR merely being a specific grammatical expression of concession.

Having established that none of V, W, Y or Z should be included in a description of JCR, we are

left with the form given in (14b) as a complete and accurate description. The final task in our

analysis of this construction is to specify the form of the reduplication involved in JCR.

4.2 A reduplication analysis of the koto wa concessive

Taking the initial predicate in JCR to be the base, and the following one the reduplicant,

we may re-state our formula for JCR as:

17) JCR: [ [ XB ] koto wa XR ]

The subscripts indicate the base and reduplicant. Okamoto’s (1990) analysis of JCR shows that, for

a number of inflectional features, the base and reduplicant are required to be identical. The three

inflectional paradigms used to demonstrate this are the ‘subject honorific’, the passive, and the

causative, and we may further include tense as another demonstration of base-reduplicant identity.

Page 20: Japanese concessive reduplication

126.792 Final Paper - 20 - S. Onosson

Inflectionally, tense in Japanese is indicated with a simple past/non-past distinction

involving the presence/absence of the past tense suffix -ta, much as in English. In every example I

have been able to find regarding JCR, both base and reduplicant carry identical tense marking. This

particular fact is not addressed by Okamoto (ibid.) or any other work cited here that addresses JCR

specifically, so we may take it as helpful but inconclusive evidence.

The ‘subject honorific’ in Japanese is formed by a complex set of morphemes, including

the honorific prefix o- attached to the verb stem, the postposition ni ‘to, at, for, etc.’ following the

verb stem, as well as this entire complex being followed by another verb, naru ‘to become’.

Okamoto (ibid., p. 252) compares JCR with compound verbs in Japanese, noting that for

compounds the honorific prefix attaches to the first verb in the compound, with the postposition

and naru following the second verb. In contrast, JCR requires all three to be present in both the

base and reduplicant. Okamoto’s example (17) (ibid.) cited here as (18) demonstrates this:

18a) *Sensei wa tegami o o-yon-da koto wa yom-i ni nat-ta. teacher T letter ACC HON-read-PST NM T read-SUF to become-PST “It is true that the teacher read the letter.”

18b) Sensei wa tegami o o-yom-i ni nat-ta koto wa yom-i ni nat-ta. teacher T letter ACC HON-read-SUF to become-PST NM T read-SUF to become-PST “It is true that the teacher read the letter.”

As indicated, non-identity in (18a) regarding the subject honorific form between the base and

reduplicant in JCR is ungrammatical.

The inflectional passive and causative in Japanese are both formed in a similar manner, as a

suffix immediately following the verb stem and preceding tense, and so we will only examine one

of them here. Okamoto (ibid., pp. 252-253) examines both and demonstrates that they behave

identically in relation to JCR. The causative is formed with the suffix -(s)ase being added to the

verb stem and, like the subject honorific, it must appear on both the base and reduplicant in JCR

(example taken from Okamoto’s (ibid.) no. (20)):

Page 21: Japanese concessive reduplication

126.792 Final Paper - 21 - S. Onosson

19a) *Taroo ni kusuri o non-da koto wa nom-ase-ta. Taroo to medicine ACC drink-PST NM T drink-CAUS-PST “It is true that (I) had Taroo take the medicine.”

19b) Taroo ni kusuri o nom-ase-ta koto wa nom-ase-ta. Taroo to medicine ACC drink-CAUS-PST NM T drink-CAUS-PST “It is true that (I) had Taroo take the medicine.”

Here again, non-identity of inflection between base and reduplicant in (19a) is ungrammatical. As

stated, the passive behaves in a completely identical manner to the causative.

One may conclude, then, that inflectional morphology must be copied entirely when

forming the reduplicant from the base. There is no optionality allowed for here. As such, the

approach required in explaining JCR is somewhat simpler than that of Morphological Doubling

Theory proposed by Inkelas (2005), which is useful in explaining reduplication which have some

kind of non-identity or optionality regarding the target of the reduplication. JCR does not fit either

requirement, having complete morphological and phonological identity on the part of the base and

the reduplicant. However, JCR is also more complex in regards to more traditional approaches to

reduplication usually handled strictly via phonology, such as within the framework of Optimality

Theory, among others. This is because, as demonstrated by the case of the subject honorific

construction (example (18) above), the material copied extends beyond simply the verb and its

inflectional morphology. In the case of the subject honorific construction, the base verb, its

inflectional morphology as well as the honorific prefix4, a subsequent postposition, and the verb

naru (as well as its inflectional morphology – e.g. past tense) are copied entirely. All of these

elements are part of the subject honorific construction, but they are well beyond the word-level,

and even beyond the initial VP, as ni heads its own PP contained within a second VP headed by

naru:

4 The status of the honorific prefix as either inflectional or derivational will remain uninvestigated here.

Page 22: Japanese concessive reduplication

126.792 Final Paper - 22 - S. Onosson

20) A simplified structure for the subject honorific:

VP2

PP V

VP1 P naru

verb ni

Inflectional morphology including the passive and the causative, as well as the honorific prefix (see

footnote 4, above) are all subsumed within VP1 at the position simply labeled verb in (20). Recall

that there is no optional allowance to target the lower VP1 within JCR.

Recall too, from example (18), that tense is marked on the higher verb, naru, in the subject

honorific construction. We may assume this to fall out from general syntactic principles regarding

the structural relationship between TP and VP (as per Adger, 2003), and some Japanese-specific

syntactic rules presumably involving feature checking which cause the tense morphology to end up

attaching to the verb in the highest VP5.

This raises the question as to what the target of JCR is? The two likeliest possibilities at

this point are either TP, or VP (always, as shown for the subject honorific construction, the highest

VP). Because tense morphology always ends up attached to a verb, we might conclude that VP is

actually the target. There is actually conflicting evidence in regards to this, however. The fact that

the PP headed by ni in the subject honorific construction is copied in JCR points to the correctness

of claiming VP as the target. However, it has already been shown that core verbal arguments, also

within the VP, are never copied. Example (10) is repeated again here:

21) Boku wa soko e it-ta koto wa it-ta 1s T there to go-PST NM T go-PST ‘It is true that I went there.’ 5 Further analysis of Japanese syntax is well beyond the scope of this paper.

Page 23: Japanese concessive reduplication

126.792 Final Paper - 23 - S. Onosson

In (21), soko e ‘to there’ is a core argument of it-ta ‘went’, and yet cannot appear alongside the

reduplicated it-ta: *Boku wa soko e it-ta koto wa soko e it-ta. This would indicate that only the V, and

not the entire VP, is targeted by JCR.

This is a puzzle for analysis. Is the target of JCR a word or a phrase? Optionality is not a

hallmark of JCR, and even for these two respective cases (subject honorific construction and core

arguments) there is no optionality: the subject honorific must include the PP in the reduplicant, and

core arguments can never be included. Perhaps the correct analysis of this situation is that the

subject honorific construction, like JCR, is itself non-compositional in nature, and that its syntactic

structure below the VP is opaque to JCR and other similar processes. Ghomeshi et al (2004: 338)

point out that within parallel architecture “there is no reason to confine the syntactic structure of

lexical items to X0 size: idioms may be XPmin or larger, and productive morphological affixes may

be treated as lexical items syntactically smaller than X0”. While subject honorification is probably

not an idiom, the parallel architecture model offers a range of possibilities regarding categorization

in this way. Taken from this point of view, the conclusion is that JCR targets the highest V0 in the

structure, along with all of its attendant morphology, but never anything above this, with the caveat

that specific grammatical constructions such as the subject honorific contain non-transparent

structures at this level, and so these are targeted by JCR, to which they are invisible. This would

seem to explain all the data presented in regards to the items included and excluded by the

reduplication.

Despite all of this, I will not attempt a complete parallel architecture-style analysis of JCR

in this paper. Despite the solutions offered by the model, I am not firmly convinced that it is the

best solution. While the availability of a ‘modality tier’ is obviously helpful, the problem (and

solution offered by parallel architecture) relating to the target of JCR, specifically in regards to the

subject honorific construction, seems too slight to justify adoption of this model. It may be argued

Page 24: Japanese concessive reduplication

126.792 Final Paper - 24 - S. Onosson

instead that the problem lies with the correct analysis of the subject honorific construction itself,

something which will not be attempted here. As such, leaving aside this one small problem, I

believe that a much simpler analysis of JCR will work well. Given the non-characteristic function

of the morphemes koto and wa within JCR, it may be appropriate to characterize these as affixes, as

they operate in this particular construction. A concise formula for JCR may then be made as:

22) JCR: V+koto+wa+RED

In this formulation, V is a verb which forms the base for the reduplication, and RED is a

reduplicative morpheme which is an exact copy of V. Koto and wa are JCR-specific affixes which

carry none of the information related to their other uses as independent morphemes6. Such an

analysis is further supported by Okamoto’s (1990: 252) comments regarding the intonation

involved with JCR: “the pitch pattern of the koto wa construction differs from that of a compound

verb: The latter has three accentual peaks, while the former has only one.”

5. Conclusions

This paper has demonstrated that JCR is a discourse-focused construction, whose primary

role is to act as a form of linguistic concession in order to introduce an argumentative statement by

the speaker in a polite, less argumentative fashion. Analysis of the superficial ‘components’ of JCR

has shown JCR to be essentially non-compositional in nature, and in this regard it should best be

described as a grammatical (possibly grammaticized) construction. This fits in well with the

traditional approach to reduplication taken by Moravcsik and others, which is that reduplications

are not simply more than the sum of their parts, but rather that reduplications are something other

than the sum of their parts. As such, a reduplication analysis of JCR is justified. Regarding the 6 Since, as noted previously, JCR may have been grammaticalized from a simple collocation of its component elements, it is likely that at one time koto and wa may have in fact contributed compositionally to the construction, whereas now they do not.

Page 25: Japanese concessive reduplication

126.792 Final Paper - 25 - S. Onosson

specific analysis of the reduplication involved in JCR, it has been clearly shown that the base and

reduplicant are required to be identical, but that determining the specific target of JCR is

somewhat problematic. While the parallel architecture model has offered a partial solution here,

the actual problem regarding the specific target may in fact not be related to the formulation of JCR

itself, and so a non-syntactic solution to the description of JCR has instead been offered.

Hopefully, the argumentation and descriptions offered in this paper may prove useful in further

work both on grammatical constructions in Japanese, and in the analysis of reduplication in other

languages.

6. References

Adger, David. 2003. Core Syntax: A Minimalist Approach. Oxford: Oxford University Press. Beninca, Paola and Cecilia Poletto. 2004. “Topic, Focus, and V2: Defining the CP Sublayers” in Luigi Rizzi (editor) The Structure of CP and IP: The Cartography of Syntactic Structures, Volume 2. Oxford: Oxford University Press Clancy, Patricia M. and Pamela Downing. 1987. “The use of wa as a cohesion marker in Japanese oral narratives” in Hinds et al, 1987. Ghomeshi, Jila, Ray Jackendoff, Nicole Rosen and Kevin Russell. 2004. “Contrastive focus reduplication in English (the salad-salad paper)” in Natural Language & Linguistic Theory 22: 307-357. Grohmann, Kleanthes K. and Andrew Ira Nevins. 2004. “On the syntactic expression of pejorative mood” in Linguistic Variation Yearbook 4 (2004), 143-179. Hinds, John. 1987. “Thematization, assumed familiarity, staging, and syntactic binding in Japanese” in Hinds et al, 1987. Hinds, John, Senko K. Maynard and Shoichi Iwasaki (editors). 1987. Perspectives on topicalization: the case of Japanese Wa. Amsterdam/Philadelphia: John Benjamins Publishing Company. Inkelas, Sharon. 2005. “Morphological Doubling Theory: Evidence for morphological doubling in reduplication” in Bernard Hurch and Veronika Mattes (editors) Studies on Reduplication. Berlin, New York: Mouton de Gruyter.

Page 26: Japanese concessive reduplication

126.792 Final Paper - 26 - S. Onosson

Iwasaki, Shoichi. 1985. “The ‘given A constraint’ and the Japanese particle ga” in S. Delancey and R. Tomlin (editors) Proceedings of the First Annual Pacific Linguistics Conference. Eugene: Department of Linguistics, University of Oregon. Kazutaka, Kurisu. 2005. “Gradient Prosody in Japanese” in Journal of East Asian Linguistics 14: 175-226. Lee, Kiri. 2002. “Nominative case-marker deletion in spoken Japanese: an analysis from the perspective of information structure” in Journal of Pragmatics 34: 683-709. Makino, Seiichi. 1987. “How relevant is a functional notion of communicative orientation to ga and wa?” in Hinds et al, 1987. Marantz, Alec. 1982. “Re Reduplication” in Linguistic Inquiry vol. 13, 3: 435-497. Martin, Samuel E. 1976. A reference grammar of Japanese. New Haven and London: Yale University Press. Maynard, Senko. K. 1987. “Thematization as a staging device in the Japanese narrative” in Hinds et al, 1987. Moravcsik, Edith. 1978. “Reduplicative Constructions” in Joseph H. Greenberg (editor) Universals of Human Language, pp. 298-334. Stanford: Stanford University Press. Namai, Kenichi. 2002. “The Word Status of Japanese Adjectives” in Linguistic Inquiry vol. 33, 2: 340-349. Nevins, Andrew and Bert Vaux. 2003. “Metalinguistic, shmetlanguistic: the phonology of shm-reduplication” in Proceedings of CLS 39. Nishiyama, Kunio. 1999. “Adjectives and the copulas in Japanese” in Journal of East Asian Linguistics 8, 183-222. Okamoto, Shigeko. 1990. “Reduplicated verbs in Japanese as grammatical constructions” in Kira Kall, Jean-Pierre Koenig, Michael Meacham, Sondra Reinman, and Laurel Sutton (editors) Proceedings of the Sixteenth Annual Meeting of the Berkely Linguistic Society, 248-256. Berkely: Berkeley Linguistic Society. Ono, Tusyoshi, Sandra A. Thompson and Ryoko Suzuki. 2000. “The pragmatic nature of the so-called subject marker ga in Japanese: evidence from conversation” in Discourse Studies Vol. 2(1): 55-84. Raud, Rein. 2002. “Objects and Events: linguistics and philosophical notions of ‘thingness’” in Asian Philosophy, Vol. 12, No. 2. Sadat-Tehrani, Nima. 2002. “The ‘Indifference-ke Construction’ in Modern Conversational Persian” in Linguistica atlantica 24 (2002-3) 43-69.

Page 27: Japanese concessive reduplication

126.792 Final Paper - 27 - S. Onosson

Shibatani, Masayoshi. 1990. The languages of Japan. Cambridge: Cambridge University Press. Suzuki, Ryoki and Yoshi Ono. 1991. “Japanese ga, Spotlighting, and Intransitives in Spoken Narratives” in Patricia M. Clancy and Sandra A. Thompson (editors) Asian Discourse and Grammar: UCSB Working Papers Vol. 3: 144-152. Travis, Lisa deMena. 2001. “The Syntax of Reduplication” in Proceedings of NELS 31, Minjoo Kim and Uri Strauss (editors), GLSA, University of Massachusetts, Amherst, pp. 455-469.