Construction It
Resume:
Japanese Reciprocal Constructions and Binding Theory
Chizuru Nakao
University of Tokyo
****-***@*****-***.**.**
Abstract
This paper investigates Japanese reciprocal constructions (the otagai construction and the V-aw
construction), which are classified into several subtypes. As for the reciprocal pronoun otagai, otagai
works as an anaphor in an object position and it works as a logophor in other positions. As for the
verbal morpheme aw, aw which yields a reciprocal reading is an affix which saturates one of the
arguments of the verb it attaches to, while aw which yields a competitive reading is a verb which
selects a clausal complement. These variations can be explained within Reinhart and Reuland s (1993)
Binding Theory.
Keywords: Binding Theory, reciprocal constructions, anaphor, logophor,
1. Introduction
Japanese has two different reciprocal constructions: the otagai construction and the V-aw
construction. The former employs the nominal element otagai as in (1a) and the latter
employs the verbal morpheme aw as in (1b) to express reciprocity.
(1) a. [Taro to Jiro]i-ga otagaii-o tatai-ta.
Taro and Jiro-NOM each-other-ACC hit-past
[Taro and Jiro]i hit each otheri.
b. [Taro to Jiro]i-ga ei tataki-aw-ta.
Taro and Jiro-NOM hit-AW-past
In this paper, we will observe the basic facts of these two constructions. Based on the
observations, I will show that the otagai construction and the V-aw construction are classified
into several subtypes; analyses should be given for each type of otagai and aw. As for otagai,
I will argue that otagai works as a reciprocal anaphor in some cases, but that it works as a
logophor in other cases. As for the status of aw, I will give supporting evidence for Ishii
(1989), who claims that it is an affix. I will also claim that there is another kind of V-aw
construction, where aw is a verb which takes a clausal complement. Based on these
This paper is based on Nakao (2002).
17
observations, I will go on to argue that the behavior of otagai and aw can be adequately
explained by Reinhart and Reuland s (1993) (henceforth, R&R) Binding Conditions.
In Section 2, we will observe the basic properties of otagai and discuss in what
circumstances it works as an anaphor or a logophor. In Section 3, observing the properties of
aw, I will classify the V-aw construction into three subtypes and show how each of them is
analyzed. In Section 4, I will show that the analyses in Sections 2 and 3 are compatible with
R&R s Binding Theory, but not with Chomsky s (1981) Binding Theory. I will make
concluding remarks in Section 5.
2. The Otagai Construction
2.1. Properties of Otagai
In this subsection, let us focus on the properties of otagai. The nominal element otagai
appears in various syntactic positions such as the direct object position (2a), the indirect
object position (2b), the object position of a postposition (2c), and the possessor position
within the object NP (2d). In these examples, otagai takes its antecedent within a clause.
(2) a. [Taro to Jiro]i-ga otagaii-o tatai-ta. (=(1a))
Taro and Jiro-NOM each-other-ACC hit-past
[Taro and Jiro]i hit each otheri.
b. [Taro to Jiro]i-ga otagaii-ni seki-o yuzut-ta.
Taro and Jiro-NOM each-other-to seat-ACC give-past
[Taro and Jiro]i gave a seat to each otheri.
c. [Taro to Jiro]i-ga otagaii-ni booru-o ket-ta.
Taro and Jiro-NOM each-other-to ball-ACC kick-past
[Taro and Jiro]i kicked the ball to each otheri.
d. [Taro to Jiro]i-ga [otagaii-no hahaoya]j-o home-ta.
Taro and Jiro-NOM each-other-GEN mother-ACC praise-past
[Taro and Jiro]i praised [each otheri s mothers]j.
As Imani and Peters (1996) point out, otagai in the possessor position allows readings
other than the reciprocal reading. For example, (3) yields four different readings in (3a-d),
depending on the situation.1
As C. Tancredi (p.c.) points out, (3c) and (3d) entail both (3a) and (3b); they may not be treated as
1
completely separated readings. If we treat (3c) and (3d) as special cases of reciprocal and reflexive
readings, in any case, otagai in (3) is different from that in (2a-c) in that it allows the reflexive
reading.
18
(3) [Mary to John]i-ga [otagaii-no kodomo]j-o yuuenti-ni tureteit-ta.
Mary and John-NOM each-other-GEN child-ACC park-LOC take-past
a. Mary took John s children to the park, and he took hers there. (reciprocal)
b. Mary took her children to the park, and John took his there. (reflexive)
c. Mary took her and John s children to the park, and John took the children there.
(collective)
d. Mary took her children to the park and also took John s there, and he did the
same. (distributive) (Imani and Peters 1996: 100)
On the other hand, otagai in the direct object position, the indirect object position or the
object position of a postposition yields only the reciprocal reading and does not allow the
other readings.
Otagai does not always obey the strict locality condition; there are cases where otagai
takes a long-distance antecedent. First, it is observed that otagai in the embedded subject
position can be bound by the matrix subject as in (4), although, otagai in the embedded object
position cannot take the matrix subject as its antecedent as shown in (5).
(4) [John to Mary]i-ga [otagaii-ga Bill-o seme-ta to] omow-ta.
John and Mary-NOM each-other-NOM Bill-ACC blame-past that think-past
John thought that Mary blamed Bill and Mary thought that John blamed Bill.
(Nishigauchi 1992: 159)
(5) *[John to Mary]i-ga [Billj-ga otagaii-o seme-ta to] omow-ta.
John and Mary-NOM Bill-NOM each-other-ACC blame-past that think-past
[John and Mary]i thought that Billj blamed each otheri.
Second, when otagai is further embedded within the object NP in the embedded clause, the
acceptability increases. In the sentence in (6), otagai can take the matrix subject John to Mary
(John and Mary) as its antecedent.
(6) ?[John to Mary]i-ga [Bill-ga [otagaii-no kodomo]j-o seme-ta to]
John and Mary-NOM Bill-NOM each-other-GEN child-ACC blame-past that
omow-ta.
think-past
John thought that Bill blamed Mary s child and Mary thought that Bill blamed John s
child.
Otagai normally takes a c-commanding antecedent as seen in the above examples. With
the point-of-view interpretation, however, antecedent-less otagai sometimes occurs in the
19
subject position as in (7).
(7) Otagai-ga ki-o-tukeru beki-da.
each-other-NOM be-careful should-pres
Each of us/them should be careful. (Imani and Peters 1996: 102)
In this case, otagai denotes each individual of a group which is salient in the discourse.2 I will
term the reading which antecedent-less otagai yields the each-of-us/them reading.
2.2. Problems
Traditionally, it has been widely held that otagai is a reciprocal pronoun, hence an anaphor,
just like English each other is. At least I accept the view that otagai is a nominal element
because as we saw in (2), otagai appears in a Case position; it is assigned accusative Case in
(2a), dative Case in (2b), oblique Case in (2c) and genitive Case in (2d).
However, the data in Section 2.1. show that occurrences of otagai in different syntactic
positions behave differently. The problem arises as to whether we can treat all kinds of otagai
in the same way or not. First, as for its syntactic behavior, otagai requires a local antecedent
in some cases and it does not in the other cases. Second, taking a look at its meaning, we see
that otagai sometimes yields various readings other than the reciprocal reading and sometimes
yields only the reciprocal reading. In the next subsection, I will examine what properties
otagai exhibits in what circumstances and show that some kinds of otagai should be treated as
a logophor, rather than an anaphor.
2.3. Proposal
2.3.1. Otagai in an object position
As we have seen in (3), otagai in the possessor position of object NP yields various
readings such as a reciprocal reading, a reflexive reading, a collective reading and a
distributive reading. The other examples of otagai with a local antecedent, otagai in (2a-c),
however, yield only a reciprocal reading.
(2) a. [Taro to Jiro]i-ga otagaii-o tatai-ta. (=(1a))
Taro and Jiro-NOM each-other-ACC hit-past
[Taro and Jiro]i hit each otheri.
Because of the point-of-view restriction, these individuals must be animate objects.
2
20
b. [Taro to Jiro]i-ga otagaii-ni seki-o yuzut-ta.
Taro and Jiro-NOM each-other-to seat-ACC give-past
[Taro and Jiro]i gave a seat to each otheri.
c. [Taro to Jiro]i-ga otagaii-ni booru-o ket-ta.
Taro and Jiro-NOM each-other-to ball-ACC kick-past
[Taro and Jiro]i kicked the ball to each otheri.
d. [Taro to Jiro]i-ga [otagaii-no hahaoya]j-o home-ta.
Taro and Jiro-NOM each-other-GEN mother-ACC praise-past
[Taro and Jiro]i praised [each otheri s mothers]j.
In addition, otagai in (2a-c) always requires a local antecedent, while otagai in the possessor
position of an object sometimes takes a long-distance antecedent as shown in (6); as shown in
(5), otagai in a direct object position cannot take a long-distance antecedent and neither can
otagai in the indirect object position or the oblique position, as shown in (8).
(8) a. *[Taro to Jiro]i-ga [Hanakoj-ga otagaii-ni seki-o yuzut-ta to] omow-ta.
Taro and Jiro-NOM Hanako-NOM each-other-to seat-ACC give-past that think-past
[Taro and Jiro]i thought that Hanakoj gave a seat to each otheri.
b. *[Taro to Jiro]i-ga [Hanakoi-ga otagaii-ni booru-o ket-ta to] omow-ta.
Taro and Jiro-NOM Hanako-NOM each-other-to ball-ACC kick-past that think-past
[Taro and Jiro]i thought that Hanakoi kicked the ball to each otheri.
In (2a-c), otagai is in the object position of the main verb, while it is embedded within the
object in (2d). (We can analyze the verb keru (kick) and the postposition ni (to) in (2c) as
making a constituent and select otagai as the object. See R&R.) Let us tentatively generalize
that otagai in an object position forces the reciprocal reading and that otagai in an object
position requires a local antecedent, and confirm the generalization by looking at other
examples.
2.3.2. Otagai in a non-object position
Otagai in the possessor position within an object NP yields four different readings. The
same four readings are available when otagai takes a long-distance antecedent. For example,
(4), where otagai is in the embedded subject position, and (6), where otagai is in the
possessor position inside the embedded object, are actually interpreted in various ways as
shown in (4 ) and (6 ) depending on the situation.
21
(4 ) [John to Mary]i-ga [otagaii-ga Bill-o seme-ta to] omow-ta.
John and Mary-NOM each-other-NOM Bill-ACC blame-past that think-past
a. John thought that Mary blamed Bill and Mary thought that John blamed Bill.
(reciprocal)
b. John thought that he blamed Bill and Mary thought that she blamed Bill.
(reflexive)
c. John thought that he and Mary blamed Bill and Mary thought that John and she
blamed Bill. (collective)
d. John thought that he blamed Bill and that Mary also blamed Bill, and Mary had the
same thought. (distributive)
(6 ) ?[John to Mary]i-ga [Bill-ga [otagaii-no kodomo]j-o seme-ta to]
John and Mary-NOM Bill-NOM each-other-GEN child-ACC blame-past that
omow-ta.
think-past
a. John thought that Bill blamed Mary s child and Mary thought that Bill blamed
John s child. (reciprocal)
b. John thought that Bill blamed John s child and Mary thought that Bill blamed
Mary s child. (reflexive)
c. John thought that Bill blamed John s and Mary s children and Mary thought that
Bill blamed the children. (collective)
d. John thought that Bill blamed John s child and that Bill also blamed Mary s child
and Mary had the same thought. (distributive)
We can see from the data in (3), (4 ) and (6 ) that otagai yields not only a reciprocal reading
but also other readings when it resides in a non-object position. In these cases, it may not be
appropriate to call otagai a normal reciprocal anaphor since its meaning is not limited to the
reciprocal one; it should not be treated in the same way as otagai in an object position.
Besides, otagai in the embedded subject position and otagai in the possessor position can take
a long-distance antecedent; in this respect, too, they are different from otagai in an object
position.
Here, I would like to propose that such a kind of otagai should be analyzed as a logophor,
rather than a reciprocal anaphor. Logophors are defined as pronouns such that the choice of
the antecedent is not determined syntactically (Sells and Wasow (1999)). Here, I would like
to assume distribution of logophoric otagai is not determined by Binding Theory, which
determines distributions of anaphors. This line of explanation is compatible with R&R s
Binding Theory, which we will see in Section 4.
22
2.3.3. Otagai in the matrix subject position
The antecedent-less otagai in (7), which only allows the each-of-us/them reading, is at
first sight problematic for our generalization; although it is in the subject position, it does not
yield any of the four readings available to otagai in the embedded subject position. However,
such a problem does not arise if we assume that this type of otagai is not in the canonical
subject position, but rather in a focus position. In short, we can treat this type of otagai as a
focus logophor, which is yet another kind of a logophor.
Note that this kind of otagai cannot be topicalized as in (7 b), although, in Japanese,
subjects frequently undergo topicalization.3
(7 ) a. Otagai-ga ki-o-tukeru beki-da. (=(7))
each-other-NOM be-careful should-pres
Each of us/them should be careful.
b. *Otagai-wa ki-o-tukeru beki-da.
each-other-TOP be-careful should-pres (Imani and Peters 1996: 102)
We assume that otagai in (7) must be moved to the focus position and it cannot reside in the
topic position. Indeed, antecedent-less otagai forces the contrastive reading such as each of
us, but not only one of us.
2.3.4. Summary
From the point of view of its syntactic behavior and meaning, otagai can be classified into
three subtypes.
(9) a. otagai in an object position: an anaphor
b. otagai in the embedded subject position and the possessor position: a logophor
c. otagai in the matrix subject position: a focus logophor
If we treat all the types of otagai as a mere reciprocal anaphor, we cannot explain why it
exhibits the variety of syntactic and semantic behavior observed.
3
Otagai in other examples can be topicalized, as exemplified in (ib).
(i) a. [Taro to Jiro]i-ga otagaii-wa home-ta. (cf. (1a))
Taro and Jiro-NOM each-other-TOP praise-past
Each otheri, [Taro and Jiro]i praised (but theyi didn t praise the others).
b. [John to Mary]i-ga [otagaii-wa Bill-o seme-ta to] omow-ta. (cf. (4))
John and Mary-NOM each-other-TOP Bill-ACC blame-past that think-past
John thought that Mary (but not the others) blamed Bill and Mary thought that John (but not
the others) blamed Bill.
23
3. The V-aw Construction
3.1. Properties of Aw
Let us consider the properties of the V-aw construction next. (10a-d) are examples of the
V-aw construction and have roughly the same meaning as (2a-d).
(10) a. [Taro to Jiro]i-ga ei tataki-aw-ta. (=(1b))
Taro and Jiro-NOM hit-AW-past
[Taro and Jiro]i hit each otheri.
b. [Taro to Jiro]i-ga ei seki-o yuzuri-aw-ta.
Taro and Jiro-NOM seat-ACC give-AW-past
[Taro and Jiro]i gave a seat to each otheri.
c. [Taro to Jiro]i-ga ei booru-o keri-aw-ta.
Taro and Jiro-NOM ball-ACC kick-AW-past
[Taro and Jiro]i kicked the ball to each otheri.
d. [Taro to Jiro]i-ga [ei hahaoya]j-o home-aw-ta.
Taro and Jiro-NOM mother-ACC praise-AW-past
[Taro and Jiro]i praised [each otheri s mothers]j.
As Nishigauchi (1992) points out, there must be a gap in the reciprocalized argument
position in the V-aw construction. For example, the verb tataku (hit) usually has one internal
argument, that is, a direct object. When the reciprocal morpheme aw is attached to the verb as
in (10a), the object of tataku must not be overtly expressed and there must be a gap (e) in the
potential object position. The gap licensed by aw can appear in various syntactic positions:
the direct object position (10a), the indirect object position (10b), the object position of a
postposition (10c), and the possessor position inside an object NP (10d). In this respect, the
distribution of the gap is similar to that of otagai which takes a local antecedent.
When the reciprocalized argument is a possessor NP as in (10d), the number of the verb s
argument is not affected even when aw is attached to the verb; the direct object of the verb
homeru (praise) is overtly expressed as the noun hahaoya (mother) in (10d). In this case, the
object of Taro s praising is Jiro s mother, not Jiro himself (and vice versa), unlike the normal
reciprocal sentences in (10a-c). Thus, we will call the reciprocal construction exemplified in
(10a-c) the direct reciprocal V-aw construction and that in (10d) the indirect V-aw
reciprocal construction.
It is observed that unlike otagai, aw obeys the strict locality condition; when aw is
attached to the matrix verb, it does not license a gap in the embedded subject position (11a) or
the embedded object position (11b).
24
(11) a. *[Taro to Jiro]i-ga [IP ei sensei-o home-ta to] omoi-aw-ta.
Taro and Jiro-NOM teacher-ACC praise-past that think-AW-past
[Taro and Jiro]i thought that each otheri praised the teacher.
b. *[Taro to Jiro]i-ga [IP sensei-ga ei home-ta to] omoi-aw-ta.
Taro and Jiro-NOM teacher-NOM praise-past that think-AW-past
[Taro and Jiro]i thought that the teacher praised each otheri.
Even when aw is attached to the embedded verb, the gap licensed there does not behave like
otagai in the embedded clause; the embedded subject NP cannot be the reciprocalized
argument.4
(12) a. #[Taro to Jiro]i-ga [IP ei Hanako-o home-aw-ta to] omow-ta.
Taro and Jiro-NOM Hanako-ACC praise-AW-past that think-past
[Taro and Jiro]i thought that each otheri praised Hanako.
b. *[Taro to Jiro]i-ga [IP Hanako-ga ei home-aw-ta to] omow-ta.
Taro and Jiro-NOM Hanako-NOM praise-AW-past that think-past
[Taro and Jiro]i thought that Hanako praised each otheri.
The V-aw construction with a reciprocal reading needs a gap in the reciprocalized
argument position. However, the gap-less V-aw construction is allowed under different
readings. In examples in (13), the morpheme aw gives the readings different from the
reciprocal reading, while all of the examples of the V-aw construction with a gap require a
reciprocal reading.
(13) a. [Taro to Jiro]-ga yorokobi-aw-ta.
Taro and Jiro-NOM be-pleased-AW-past
Taro and Jiro showed the joy together.
b. [Taro to Jiro]-ga ookina oto-de sutereo-o kake-aw-ta.
Taro and Jiro-NOM loud sound-in stereo-ACC play-AW-past
Taro and Jiro competitively/collaboratively played stereos at a high volume.
4
# indicates that the sentence is not acceptable in the intended reading. (12a) is only grammatical in
the competitive reading. In this case, the null element in the embedded subject position is assumed to
be pro. Thus, (ia) can be paraphrased as (ib).
(i) a. [Taro to Jiro]i-ga [IP proi Hanako-o home-aw-ta to] omow-ta.
Taro and Jiro-NOM teacher-ACC praise-AW-past that think-past
b. [Taro to Jiro]i-ga [IP karerai-ga Hanako-o home-aw-ta to] omow-ta.
Taro and Jiro-NOM they-NOM teacher-ACC praise-AW-past that think-past
Taro and Jiro thought that they competitively/collaboratively praised Hanako.
25
There is no reciprocalized argument in these sentences, but the V-aw construction is still
allowed. (13a) means that Taro and Jiro showed joy at the same time. (13b) means that Taro
and Jiro played stereos competitively, or in collaboration. The readings of the gap-less V-aw
construction are classified into three: (i) a simultaneous reading, (ii) a collaborative reading,
and (iii) a competitive reading. Let us term this gap-less reciprocal construction the
competitive V-aw construction. 5
3.2. Previous Analyses and Problems
As for the direct reciprocal V-aw construction, there are two opposing hypotheses to
explain the nature of the morpheme aw and the gap it licenses. Nishigauchi (1992) claims,
adapting the analysis of each other in Heim, Lasnik and May (1991) to Japanese, that the
V-aw construction involves two operators: the distributor and the reciprocator. More
specifically, aw works as the distributor and the gap is the trace of the null reciprocator.
(14) [IP [John to Mary]i-ga [IP ti [VP1 ti [VP1 Opj [VP2 ti [V 2 tj aisi]] aw]] te-iru]]
John and Mary-NOM love-AW-prog
[John and Mary]i love each otheri.
Under this hypothesis, the gap position is syntactically active as an A -trace.
Ishii (1989), however, argues that aw is an affix that saturates a syntactic position of the
verb it attaches to.
(15) [IP [John to Mary]i-ga [VP ti [V aisi-aw]] te-iru]
In his view, the V-aw construction is different from the otagai construction in that the former
does not have any reciprocalized argument position which is syntactically active.
In order to see which of the two hypotheses is empirically adequate, we should test the
V-aw construction for the syntactically visible reciprocalized argument position. In Section
3.3.1, I will give supporting evidence for Ishii (1989) and show that the gap position in the
V-aw direct reciprocal construction is not syntactically active, and that aw is an affix which
saturates one argument position of the verb it attaches to.
Then, another question arises as to how we should analyze aw in the indirect reciprocal
V-aw construction and the competitive V-aw construction. As we have seen, aw in these
constructions does not reduce the number of the arguments of the verb it attaches to. To
answer the question, I will propose the analysis of the indirect reciprocal V-aw construction
5
Tonoike (1991) terms this kind of construction kyousou sougobun (competitive reciprocals) in
Japanese.
26
and the competitive V-aw construction. In Section 3.3.2, I will claim that aw in the indirect
reciprocal V-aw construction is the same affix as aw in the direct reciprocal V-aw construction,
showing that the gap in the indirect reciprocal V-aw construction is indeed in an argument
position of the verb V. In Section 3.3.2, I will argue that aw in the competitive V-aw
construction is a verb which selects a clausal complement. Postulating the two aws of
different categories, we can account for many behavioral differences between them.
3.3. Proposals
3.3.1. Supporting evidence for Ishii (1989)
In this section, I will give two pieces of supporting evidence for Ishii s (1989) analysis that
aw in the direct reciprocal V-aw construction is an affix which saturates one of the argument
positions of the verb it attaches to.
3.3.1.1. Distributive reading versus collective reading
The first argument concerns the interpretation of the reflexive pronoun zibun (self) and its
plural form zibun-tati (self-pl). Japanese is a language without overt number agreement; a
bare NP such as hon (book) can mean a/the book or (the) books, depending on the
situation where it is used. However, Japanese optionally allows the plural marker tati when
the plural NP is animate. For example, while sensei (teacher) can mean either a/the teacher
or (the) teachers, sensei-tati (teacher-pl) can only refer to more than one teacher.6
Interestingly, although the plural marker tati is optional, zibun and zibun-tati give different
interpretations when they have plural antecedents; zibun must always take an atomic
individual as the antecedent. Consider (16).
(16) a. [Taro to Jiro]i-ga [PP zibuni-no heya-de] benkyoosi-ta.
Taro and Jiro-NOM self-GEN room-in study-past
[Taro and Jiro]i studied in selfi s room.
b. [Taro to Jiro]i-ga [PP zibun-tatii-no heya-de] benkyoosi-ta.
Taro and Jiro-NOM self-pl-GEN room-in study-past
[Taro and Jiro]i studied in theiri room.
(16a) and (16b) illustrate different situations. (16a) means that Taro studied in Taro s room
6
The expression such as sensei-tati (teacher-pl) has another interpretation where it denotes a group
consisting of the teacher(s) and others. This interpretation is irrelevant here, so I discuss only tati as
a plural marker.
27
and Jiro studied in Jiro s room, while (16b) means that Taro and Jiro studied together in the
same room which they own jointly. In other words, (16a) yields the distributive reading and
(16b) yields the collective reading of the plural subject. The same is true in the case of
transitive verbs as seen in (17).
(17) a. [Taro to Jiro]i-ga [PP zibuni-no heya-de] hon-o yon-da.
Taro and Jiro-NOM self-GEN room-in book-ACC read-past
[Taro and Jiro]i read books in selfi s room.
b. [Taro to Jiro]i-ga [PP zibun-tatii-no heya-de] hon-o yon-da.
Taro and Jiro-NOM self-pl-GEN room-in book-ACC read-past
[Taro and Jiro]i read books in theiri room.
When we use an inherently reciprocal intransitive verb such as kenkasuru (fight), however,
the distributive reading is blocked, as in (18a), because the subject NP of the verb kenkasuru
is obligatorily plural; one cannot fight by oneself.
(18) a. *[Taro to Jiro]i-ga [PP zibuni-no heya-de] kenkasi-ta.
Taro and Jiro-NOM self-GEN room-in fight-past
[Taro and Jiro]i fought in selfi s room.
b. [Taro to Jiro]i-ga [PP zibun-tatii-no heya-de] kenkasi-ta.
Taro and Jiro-NOM self-pl-GEN room-in fight-past
[Taro and Jiro]i fought in theiri room.
Bearing in mind the difference between the singular and plural forms of zibun, let us turn
now to the relevant examples of the reciprocal constructions. In the case of the otagai
construction, the distributive reading is possible as the acceptability of (19a) indicates.
(19) a. [Taro to Jiro]i-ga [PP zibuni-no heya-de] otagaii-o tatai-ta.
Taro and Jiro-NOM self-GEN room-in each-other-ACC hit-past
[Taro and Jiro]i hit each otheri in selfi s room.
b. [Taro to Jiro]i-ga [PP zibuni-tati-no heya-de] otagaii-o tatai-ta.
Taro and Jiro-NOM self-pl-GEN room-in each-other-ACC hit-past
[Taro and Jiro]i hit each otheri in theiri room.
However, the corresponding example of the V-aw reciprocal construction does not allow the
distributive reading; the plural subject NP cannot be the antecedent of the singular zibun as
shown in (20a).
28
(20) a. *[Taro to Jiro]i-ga [PP zibuni-no heya-de] ei tataki-aw-ta.
Taro and Jiro-NOM self-GEN room-in hit-AW-past
[Taro and Jiro]i hit each otheri in selfi s room.
b. [Taro to Jiro]i-ga [PP zibun-tatii-no heya-de] ei tataki-aw-ta.
Taro and Jiro-NOM self-tati-GEN room-in hit-AW-past
[Taro and Jiro]i hit each otheri in theiri room.
As shown in (19b) and (20b), the collective reading is obtained in both the otagai construction
and the V-aw construction.
In this respect, the complex verb tataki-aw (hit-AW) is similar to inherently reciprocal
verbs such as kenkasuru (fight), deaw (meet) and kakkonsuru (marry). This leads us to claim
that aw is an affix. If, the gap in (20a) were syntactically visible, the internal structure of the
embedded VP would be no different from the VP otagai-o tataku (hit each other) in (19a).
3.3.1.2. Object comparison readings
The second argument concerns the possibility of an object comparison reading in a
comparative deletion sentences. Darlymple, Mchombo and Peters (1994) argue that the
Chiche a reciprocal morpheme should be treated as an intransitivizing affix. One basis of
their claim is that an object comparison reading in comparative deletion sentences is allowed
in the case of reflexive verbs, but not in the cases of reciprocal verbs, as shown in (21).7
(21) a. Alenjei -ma-dz -ny z- kup s asodzij
2-hunters 2SM-hab-REFL-despise-FV exceeding 2-fishermen
The huntersi despise themselvesi more than the fishermanj.
(subject comparison/object comparison)
b. Alenjei -ma-ny z- n- kup s asodzij
2-hunters 2SM-hab-despise-RECIP-FV exceeding 2-fishermen
The huntersi despise each otheri more than the fishermanj.
(subject comparison/*object comparison)
(Darlymple, Mchombo and Peters 1994: 155)
When a reflexive verb is used in the comparative deletion sentence as in (21a), the sentence
allows two different readings: the subject comparison reading and the object comparison
reading. On the other hand, the object comparison reading is not obtained in (21b). This
indicates that there is no syntactic object in the first sentence and the verb despise becomes
intransitive when the reciprocal marker is attached to it.
The numbers in the Chiche a glosses indicate the gender classes of the nouns.
7
29
The same discussion also holds of the Japanese verbal morpheme aw. The V-aw
construction does not allow the object comparison reading in comparative deletion sentences.
(22) a. Sensei-tatii-ga seito-tatij-yorimo sakini otagaii-o hihansi-ta.
teacher-pl-NOM student-pl-than earlier each-other-ACC criticize-past
The teachersi criticized each otheri earlier than the studentsj.
(subject comparison/object comparison)
b. Sensei-tatii-ga seito-tatij-yorimo sakini ei hihansi-aw-ta.
teacher-pl-NOM student-pl-than earlier criticize-AW-past
(subject comparison/*object comparison)
This leads us to claim that the object of hihansuru (criticize) is not syntactically visible in this
case, and aw is attached to the verb as an affix.
3.3.2. The indirect reciprocal V-aw construction as an EPC
3.3.2.1. The structure
Having made clear the status of aw in the direct reciprocal V-aw construction, let us move
on to the next question, that is, How should the indirect reciprocal V-aw construction and the
competitive V-aw construction be analyzed? First, I propose in this subsection that we should
analyze the indirect reciprocal V-aw construction as an External Possessor Construction
(EPC). Under this analysis, indirect reciprocal aw is explained as the same type of aw as
direct reciprocal aw; it is an intransitivizing affix which saturates an argument position of the
attached verb.
Thus far, we have assumed that the indirect reciprocal V-aw construction licenses a gap in
the possessor position of an argument NP, as shown in (23a). In this subsection, however, I
would like to argue that the indirect reciprocal sentence (10d) has the internal structure in
(23b), where the gap is in another argument position external to the possessee NP hahaoya
(mother).
(23) a. *[Taro to Jiro]i-ga [ei hahaoya]j-o home-aw-ta.
Taro and Jiro-NOM mother-ACC praise-AW-past
b. [Taro to Jiro]i-ga ei [hahaoya]j-o home-aw-ta.
Taro and Jiro-NOM mother-ACC praise-AW-past
In (23b), [Taro to Jiro]i and the gap are realized as coarguments of the verb homeru (praise);
the gap licensed by aw is in an argument position of the verb. Thus, indirect reciprocal aw is
30
analyzed as a morpheme that turns one of its arguments into a gap, exactly as direct reciprocal
aw is.
Note that in (23b), the gap is interpreted as the possessor of the object NP hahaoya
(mother), but the gap is not realized in the possessor position of this NP. In this respect, (23b)
is much the same as the structure of so-called External Possessor Constructions (EPCs),
which are observed in many languages.8 In EPCs, the possessor NP is realized as another
argument which is external to the possessee NP. In order to show that the indirect reciprocal
V-aw construction has the structure as in (23b), I would like to show that it is analyzed as a
kind of EPC.
In the following subsections, I will take up Vergnaud and Zubizarreta s (1992) study of the
French EPC (henceforth, V&Z). In V&Z, the differences between the EPC and the Internal
Possessor Construction (IPC) are discussed in detail. I will compare Japanese indirect
reciprocal constructions with the French EPC and show that the indirect reciprocal V-aw
construction, but not the indirect reciprocal otagai construction, exhibits the properties similar
to those of the French EPC. Thus we can support the hypothesis that the gap in (10d) resides
in a position external to the possessee NP as illustrated in (23b).
3.3.2.2. The French IPC and EPC
V&Z discuss two inalienable constructions in French: the IPC and the EPC. The EPC and
the IPC are exemplified in (24) and (25), respectively.
(24) a. Le medicine a radiographi [l estomac] aux enfants. (EPC)
the doctor has X-rayed [the-stomach] to-the children
The doctor has X-rayed the children s stomachs.
b. Le medicine leuri a radiographi [l estomac] ti. (EPC with cliticization)
the doctor them has X-rayed [the-stomach]
The doctor has X-rayed their stomachs. (V&Z: 597)
(25) Le medicine a radiographi [leur estomac]. (IPC)
the doctor has X-rayed [their(sg) stomach]
The doctor has X-rayed their stomachs. (V&Z: 602)
In (24a, b), an inalienable noun estomac (stomach) is interpreted as a part of another NP
enfants (children) or leur (them). As shown in (24a), the inalienable noun is in the direct
object position and the possessor NP is expressed as another dative argument of the verb in
the EPC. This dative object can be cliticized as in (24b). (25) is an example of the IPC, where
the possessor NP leur (them) is realized within the NP headed by the inalienable noun
For the cross-linguistic variation in EPCs, see Payne and Barshi (1999), among others.
8
31
estomac.9
Inalienable nouns such as body-part nouns are dependent entities; they must be inherently
defined as a part of another NP. There are other nouns, such as kinship nouns, which function
as inalienable nouns by extension. In the IPC, an inalienable noun and another NP which it
depends on are represented within one argument position as shown in (25). On the other hand,
the EPC is a construction in which an inalienable object NP is interpreted as a possessee of
another argument that is external to it as shown in (24). The EPC is allowed only when the
possession of an inalienable noun is involved.
V&Z point out that there are three differences between the EPC and the IPC in French.
First, as Kayne (1975) notes, the inalienable noun in the EPC must be singular even when it is
associated with a plural possessor NP. The examples (26a, b) illustrate this point. In contrast,
the inalienable noun in the IPC may be singular as in (25) or plural as in (26c) when the
possessor NP is plural. This is the first difference between the EPC and the IPC.
(26) a. *Le medicine a radiographi [les estomacs] aux enfants. (EPC)
the doctor has X-rayed [the(pl) stomachs] to-the children
The doctor has X-rayed the children s stomachs.
b. *Le medicine leuri a radiographi [les estomacs] ti. (EPC with cliticization)
the doctor them has X-rayed [the(pl) stomachs]
The doctor has X-rayed their stomachs.
c. Le medicine a radiographi [leurs estomacs]. (IPC)
the doctor has X-rayed [their stomachs]
The doctor has X-rayed their stomachs. (V&Z: 602)
Second, although l estomac (the stomach) in the EPC examples (24a, b) is realized in the
singular form, it is implied that the total number of the stomachs is plural because this noun is
associated with a plural possessor NP. The singular inalienable noun in (24b) can be the
antecedent of a plural pronoun as in (27).
(27) Le medicine leuri a radiographi [l estomac]j ti
the doctor them has X-rayed [the-stomacs]
et il a constat qu ilsj abaient des images normal.
and he has seen that they had DET images normal
The doctor has X-rayed [theiri stomachs]j
and he saw that theyj looked normal. (V&Z: 599)
9
The pronoun leur (their) in (25) cannot cliticize as in (i).
(i) *Le medicine leuri a radiographi [ti estomac].
the doctor their has X-rayed stomach
The doctor has X-rayed their stomachs.
32
In (27), the distributive reading is yielded within the EPC. V&Z call this phenomenon the
distributivity effect, which is only seen in the EPC.10
Third, as Kayne (1975) notes, there is a restriction on the sort of modifiers which modify
the inalienable noun in the EPC. The inalienable noun in the IPC can be modified by any
modifier. To be concrete, the inalienable noun in the EPC cannot be modified by an appositive
modifier as shown in (28a), while the same noun in the IPC example (28b) can.
(28) a. *Pierre lui a lav les mains sales. (EPC)
Pierre to-him past washed the hands dirty
Pierre washed his dirty hands.
b. Pierre a lav ses mains sales. (IPC)
Pierre past washed his hands dirty (V&Z: 603)
The adjective sale (dirty) in (28) is appositive; the plurality of the inalienable noun mains
(hands) indicates that both of the two hands are dirty. The inalienable noun in the EPC must
not be modified by such a modifier.
To summarize, the EPC in French shows three properties which are not seen in the IPC.
First, the inalienable noun in the EPC must always be realized in the singular form. Second,
the EPC shows the distributivity effect. Third, the inalienable noun in the EPC cannot be
modified by appositive modifiers. From now on, I would like to use these three properties as
diagnostics for the EPC.
3.3.2.3. The indirect reciprocal V-aw construction
I would like to show in this subsection that the indirect reciprocal V-aw construction, but
not the indirect reciprocal otagai construction, exhibits the three properties of the EPC. This
supports the view I have proposed in (24b) that the gap in the indirect reciprocal V-aw
construction is not in the possessor position of the object NP, but in another argument position
of the verb.
First, let us consider the data concerning grammatical number. In otagai indirect
reciprocals such as (29a), the plural marker tati can optionally be attached to the NP
[otagai-no hahaoya] (each other s mothers).
10
The first and the second differences between the EPC and the IPC are correlated. Because of the
distributivity effect, body-part nouns such as estomac must not have the plural form in the EPC,
because the plural estomacs (stomachs) cannot be distributed to each members of the group-denoting
NP; one person cannot have more than one stomach. Thus the first difference is derived from the
second difference.
33
(29) a. [Taro to Jiro]i-ga otagai-noi hahaoyaj(-tati)-o kenasi-ta.
Taro and Jiro-NOM each-other-GEN mother(-pl)-ACC speak-ill-of-AW-past
Taro and Jiro spoke ill of each other s mothers.
b. [Taro to Jiro]i-ga ei hahaoyaj( -tati)-o kenasi-aw-ta.
Taro and Jiro-NOM mother(-pl)-ACC speak-ill-of-AW-past
On the other hand, in indirect reciprocal V-aw construction such as (29b), when the plural
marker is attached to the inalienable noun hahaoya (mother), the sentence is awkward; the
inalienable noun should be singular even when there are two mothers, that is, Taro s mother
and Jiro s mother, in the context (Taro and Jiro are assumed not to be brothers).
Second, the indirect reciprocal V-aw construction shows the distributivity effect. For
example, look at (30).11
(30) [Taro to Jiro]i-ga ei hahaoyaj-o home-aw-ta ga,
Taro and Jiro-NOM mother-ACC praise-AW-past but
kanozyoraj-wa sore-o kii-te-i-nakat-ta.
they(fem)-TOP that-ACC hear-prog-not-past
[Taro and Jiro]i praised [each other s mothers]j, but theyj(fem) didn t hear that.
The singular hahaoya (mother) in (29b) can be the antecedent of the plural pronoun
kanozyora (them) as shown in (30); the inalienable noun is distributed to both Taro and Jiro in
this example.
Third, let us examine what kind of modifiers can modify the inalienable nouns in the two
indirect reciprocal constructions. Compare the indirect reciprocal V-aw sentences in (31b) and
with the indirect reciprocal otagai sentences in (31a).
(31) a. [Taro to Jiro]i-ga [otagai-noi [erumesu-no yoohuku]]j-o kenasi-ta.
Taro and Jiro-NOM each-other-GEN Hermes-GEN clothes-ACC speak-ill-of-past
Taro and Jiro spoke ill of each other s Hermes clothes.
b. [Taro to Jiro]i-ga ei [erumesu-no yoohuku]j-o kenasi-aw-ta.
Taro and Jiro-NOM Hermes-GEN clothes-ACC speak-ill-of-AW-past
The corresponding otagai sentence employs the plural pronoun kanozyora (they(fem)), too, but for
11
the different reason.
(i) [Taro to Jiro]i-ga [otagai-noi hahaoya]j-o home-aw-ta ga,
Taro and Jiro-NOM each-other-GEN mother-ACC praise-AW-past but
kanozyoraj-wa sore-o kii-te-i-nakat-ta.
they(fem)-TOP that-ACC hear-prog-not-past
[Taro and Jiro]i praised [each other s mothers]j, but theyj(fem) didn t hear that.
As we saw in (29a), the hahaoya in this example is considered to be grammatically plural from the
first place.
34
In (31a), the only possible interpretation is that both Taro and Jiro are wearing Hermes clothes
and they spoke ill of each other s Hermes clothes. However, in (31b), the sentence is degraded
under the same interpretation, and the competitive reading is preferred that Taro and Jiro
competitively spoke ill of some (contextually salient) Hermes clothes, which may be Taro s or
Jiro s or someone else s, or of Hermes clothes in general. The modifier in (31) is an appositive
modifier, not a restrictive modifiers, because one person usually wears only one set of clothes
at the same time.12 Therefore, these data show that the indirect reciprocal V-aw construction is
not compatible with appositive modifiers. In this respect, too, the indirect reciprocal V-aw
construction is similar to the French EPC.
As we have seen thus far, all the three diagnostics show that the indirect reciprocal V-aw
construction is a kind of EPC, while the indirect reciprocal otagai construction is not. Besides,
both the indirect reciprocal V-aw construction and the French EPC are allowed only when a
kind of inalienable possession is involved.13 Given these facts, one may say that the indirect
reciprocal V-aw construction is an EPC, and hence, has an internal structure as in (24b), not
the structure in (24a).
Under this analysis, we can analyze the affix aw in the direct reciprocal construction and
aw in the indirect reciprocal construction uniformly, because aw licenses a gap in one
argument position of the attached verb in both constructions. In the direct reciprocal sentence
(32a), the direct object position is a gap. In the indirect reciprocal sentence (32b), the
argument position in which the possessor NP of the inalienable noun resides is a gap.
(32) a. [Taro to Jiro]i-ga ei tataki-aw-ta.
Taro and Jiro-NOM hit-AW-past
Taro and Jiro hit each other.
b. [Taro to Jiro]i-ga ei [hahaoya]j-o home-aw-ta. (=(10))
Taro and Jiro-NOM mother-ACC praise-AW-past
Taro and Jiro praised each other s mothers.
Restrictive modifiers are compatible with this construction. For example, in (i), it is indicated that
12
only one of the two shoulders of each person was hit.
(i) [Taro to Jiro]i-ga ei [hidari-no kata]j-o tataki-aw-ta.
Taro and Jiro-NOM left-GEN shoulder-ACC hit-AW-past
[Taro and Jiro]i hit each otheri s left shouldersi.
13 (i), where the possession of an alienable noun hon (book) is involved, allows reciprocal reading only
marginally.
(i) # [Taro to Jiro]i-ga ei honj-o yomi-aw-ta.
Taro and Jiro-NOM book-ACC read-AW-past
[Taro and Jiro]i read [each otheri s books]j.
35
3.3.3. Competitive aw as a verb
3.3.3.1. Competitive aw and reciprocal aw
Next, let us consider the status of aw in the competitive V-aw construction. Unlike aw in
direct and indirect reciprocal sentences, competitive aw does not yield a reciprocal reading.
As seen in Section 3.1, it requires the reading where each individual in the group does the
same thing (i) simultaneously, (ii) competitively, or (iii) collaboratively. Since it is not a
reciprocal morpheme, I would like to assume that this kind of aw is a different lexical item
from aw in the direct and indirect reciprocal V-aw constructions.
Let us assume that aw in the competitive V-aw construction is a non-affixal verb, which
-selects a group-denoting subject NP and a clausal complement. The structure of (33a) is
represented as (33b).
(33) a. [Taro to Jiro]-ga yorokobi-aw-ta. (= (13a))
Taro and Jiro-NOM be-pleased-AW-past
Taro and Jiro showed the joy together.
b. IP
NP I
[Taro to Jiro]i-ga VP I
IP V ta
NP VP aw
PROi yorokobi
In the structure in (33b), the subject Taro to Jiro is the -subject of the verb aw and thus its
semantic properties are selectionally constrained by aw.14 More specifically, the subject in
(33b) has to be, say, synchronizers, collaborators, or competitors. The embedded
subject is the agent of the verb yorokobu (be pleased), and is interpreted as coreferential with
the synchronizers, etc. Let us take it to be a PRO that is controlled by the matrix subject.
On the other hand, our claim that the reciprocal aw is an affix indicates the structure of
(34a) is represented as (34b). Thus we postulate two kinds of aw.
I would like to specify the name of the -role given here in future research.
14
36
(34) a. [Taro to Jiro]-ga tataki-aw-ta. (=(8b))
Taro and Jiro-NOM hit-RARE-past.
Taro was hit by Jiro.
b. IP
NP I
[Taro to Jiro]i-ga VP I
NP V ta
ei tataki-aw
3.3.3.2. Animacy constraint
In Sections 3.3.3.2. and 3.3.3.3, I will give two pieces of evidence for my claim that
competitive aw -selects the subject but direct/indirect reciprocal aw does not.
Imani and Peters (1996) note that in the V-aw constructions, some predicates allow
inanimate subjects while others do not. For example, when aw attaches to the verb tatu
(stand), the inanimate noun ki (tree(s)) cannot be the subject of this verb as in (35a), while the
animate noun heisi (soldier(s)) can be its subject as in (35b).
(35) a. *Ki-ga 300-meter-okini tati-aw-te-iru.
tree-NOM 300-meter-at-the-interval stand-AW-pres
The trees (competitively/collaboratively) stand at interval 300 meters.
b. Heisi-tati-ga 300-meter-okini tati-aw-te-iru.
soldier-pl-NOM 300-meter-at-the-interval stand-AW-pres
The soldiers stand at interval 300 meters. (Imani and Peters 1996: 108)
The verb tatu (stand) itself of course allows an inanimate subject as shown in (36).
(36) Ki-ga 300-meter-okini tat-te-iru.
tree-NOM 300-meter-at-the-interval stand-pres
The trees stand at interval 300 meters.
Besides, the inanimate noun ki (tree(s)) can be the subject of yorikakar-aw (lean-AW) as in
(37).
37
(37) Kii-ga ei yorikakari-aw-te-iru.
tree-NOM lean-AW-pres
The trees lean against each other. (cf. Imani and Peters 1996: 108)
In our view, (37) is an example of the direct reciprocal V-aw construction, while (35a-b) are
examples of the competitive V-aw construction. This is indicated by the fact that yorikakaru
(lean) is a two-place predicate and tatu (stand) is a one-place predicate. If we turn (37b) into a
indirect reciprocal V-aw sentence as in (38), where the gap is located in the possessor position
of the dative object, the sentence is still acceptable with an inanimate object.
(38) Kii-ga [ei miki]-ni yorikakari-aw-te-iru.
tree-NOM trunk-DAT lean-AW-pres
The trees lean against each other s trunks.
These facts show that only the competitive V-aw construction prohibits inanimate subjects.
Since the subject of the competitive V-aw construction is -selected by aw, this is due to
the inability of inanimate nouns to be assigned the role of synchronizer, collaborator, or
competer. If we analyze competitive aw on a par with reciprocal aw, we cannot explain the
facts about inanimate subjects.
3.3.3.3. Historical data
Another piece of evidence is based on Japanese historical data: the two types of aw came
to be used in different periods. Before we consider aw in the V-aw construction, let me draw
your attention to aw which is not accompanied with another verb. Aw is also used as a main
verb, which has several meaning such as meet, match, and agree/be the same. 15 The
sentences in (39) are several examples of the main verb aw.
(39) a. [Taro to Jiro]-ga aw-ta.
Taro and Jiro-NOM meet-past
Taro and Jiro met.
b. [Sono huku to sono kutu]-ga yoku aw.
the clothes and the shoes-NOM well match
The clothes and the shoes match.
15
Aw with the meaning meet is represented by a different kanji from aw with other meanings in
Modern Japanese. Nevertheless, they are originally the same word and Japanese dictionaries treat
them within one lexical entry.
38
c. [Taro to Jiro-no kotae]-ga aw-ta.
Taro and Jiro-GEN answer-NOM agree-past
Taro s answer and Jiro s answer agreed.
From the discussion thus far, we can see that there are three types of aw, and that competitive
aw yields three meanings. The types of aw are summarized in (40).16
(40) a. main verb aw; meet, match, agree, etc.
b. competitive aw;
(i) the simultaneous meaning
(ii) the collaborative meaning
(iii) the competitive meaning
c. direct/indirect reciprocal aw; the reciprocal meaning
Now, let us see when each type of aw came to be used. Main verb aw (40a) is attested in
Manyousyuu, the oldest Japanese anthology in 8c. Aw in the competitive V-aw construction
with the simultaneous meaning (40bi) is also attested in this anthology. On the other hand, the
V-aw construction with collaborative and competitive meaning (40bii, iii) appeared later.
These meaning are attested in the literature in 10-11c. The reciprocal meaning of the V-aw
construction (40c) is not attested until about 14c. (41a-e) are examples of the five types of aw
listed in (40).17
(41) a. Suso-no utikae awa-ne-domo (Manyousyuu; 8c)
hem-GEN front of the kimono AW-not-though
Though the front hem of the kimono does not fit well,
b. Himo-no o-no itugari-ai-te (Manyousyuu; 8c)
string-GEN life-NOM be linked-AW-te18
The strings of life were linked together, and
c. Dainagon-o sosiri-ai-tari. (Taketori Monogatari; 10c)
Dainagon-ACC speak ill of-AW-complete
(They) spoke ill of Dainagon collaboratively.
16
As for the reciprocal pronoun otagai, it came to be used in about 9c. It is derived from the deverbal
noun tagai difference, which is in turn derived from the inherently reciprocal verb tagau differ.
17
The examples in (41) are taken from Japanese dictionaries. (41a-c) are from Shogakukan Kokugo
Daijiten, (41d) is from Kojien 4th edition, and (41e) is from Jidaibetsu Kokugo Daijiten:
Muromachi-jidai hen 1.
18 An auxiliary verb te is used to connect the sentence to another sentence.
39
d. Kakemono-domo nado ni-naku-te idomi-ae-ri.19 (Genzi Monogatari; 11c)
prize-pl euphemistic two-not-te challenge-AW-complete
The prizes were peerless and (they) competitively challenged (for them).
e. Tada ko-wa oya-o sute kiri-ai (Taiheiki; 14c)
just child-TOP parent-ACC abandon slay-AW
The child just abandoned the parent and (they) slew each other
When we assume that competitive aw is a verb and reciprocal aw is an affix, it can be said
that the former is closer to the main verb aw because it has its own -role structure.
Reciprocal aw is closer to functional items such as auxiliaries in that it does not assign any
-roles to the arguments. Given that the derivation of a functional item from a lexical item is a
common process of grammaticalization, the fact that competitive aw appeared earlier than
reciprocal aw supports our view; the two types of aw in the V-aw construction are different
items and the reciprocal aw has a more functional nature than competitive aw.
3.4. Summary
In this section, I have shown that the V-aw construction is classified into three subtypes,
which are summarized in (42).
(42) a. the direct reciprocal V-aw construction
b. the indirect reciprocal V-aw construction
c. the competitive V-aw construction
Among these subtypes, aw in the direct reciprocal V-aw construction and the indirect
reciprocal V-aw construction can be analyzed as an affix which saturates an argument position
of the verb V, and aw in the competitive V-aw construction can be analyzed as a verb which
selects a clausal complement.
4. Binding Theory
The properties of reciprocal elements have traditionally been accounted for by Chomsky s
(1981, 1986) Binding Condition A. However, interestingly, the observations in Section 2 and
Section 3 cast doubt on the adequacy of the traditional Binding Conditions in (43).
(43) a. Binding Condition A: An anaphor must be bound in its governing category.
b. Binding Condition B: A pronominal must be free in its governing category.
The postposition of euphemism nado is used to express the speaker s politeness.
19
40
Adopting the Binding Condition A in (43a) to account for Japanese reciprocal
constructions, one empirical problem and one theoretical problem arise. First, it is empirically
problematic because (43a) does not explain the fact that the reciprocal pronoun otagai
sometimes takes a long-distance antecedent and it sometimes takes no syntactic antecedent.
We have to say that it can be exempt from the Binding Condition A in some cases, but it is not
clear at all how to account for these exceptions while still maintaining a Binding Theoretic
analysis as in (43) in the core cases.
Second, Binding Condition A in (43a) is a condition on anaphors, that is, nominal elements,
and not reciprocal elements in general. It is theoretically undesirable because the reciprocal
verbal affix aw cannot be in the scope of the Binding Conditions, although it yields many of
the same reciprocal reading as reciprocal pronouns do.
Instead, I would like to propose extending Reinhart and Reuland s (1993) Binding
Conditions to the reciprocal predicate. R&R redefine the Binding Condition as conditions
imposed on predicates, not on arguments. They proposed the Binding Condition in (45) on the
basis of the notions defined in (44).
(44) a. A predicate is reflexive iff two of its arguments are coindexed.
b. A predicate is reflexive-marked iff either P is lexically reflexive or one of P s
argument is a self-anaphor. (R&R: 663)
(45) a. Binding Condition A: A reflexive-marked predicate is reflexive.
b. Binding Condition B: A reflexive predicate is reflexive-marked. (R&R: 670-671)
Let us refer to this kind of treatment of reflexive constructions as the predicate-centered
approach to reflexive constructions, as opposed to the argument-centered approach such as
(43).
In order to extend the predicate-centered approach to reciprocal constructions, let us
reformulate R&R s Binding Condition in (45), replacing the notions reflexive and
reflexive-marked with reciprocal and reciprocal-marked, respectively.20
(46) a. A predicate is reciprocal iff
(i) two of its arguments are coindexed, and
(ii) these arguments refer to a group-denoting NP.
b. A predicate is reciprocal-marked iff either P is lexically reciprocal or one of P s
arguments is a reciprocal anaphor.
When the subject is a group-denoting NP, a predicate with two or more coindexed coarguments is
20
interpreted as either a reflexive predicate (44a) or a reciprocal predicate (46a). I would like to discuss
the precise condition on which a predicate is interpreted as reciprocal, not reflexive, in future research.
Contact this candidate