Splitting a coordination with “ with ”

By investigating data on Japanese particle to ‘with/and’, this paper provides a unified analysis of three types of construction containing the particle to. We argue that a single to particle coordinating two elements is a conjunction particle (single to construction), while to selecting the coordination phrase (multiple to construction) and to adjoined to a single element (split to construction) are comitative postpositions. Under the proposal, the split to construction and the multiple to construction share the same underlying structure containing the single to construction. We also show that Russian coordinative and discontinuous comitatives are derived in a similar way to Japanese, following Ionin and Matushansky (2002). Our proposal suggests that a discontinuous comitative phrase and its associate universally form a constituent in a base structure.

(3) Multiple to construction [ Haru to Aki to]-ga Natsu-o home-ta.Haru TO Aki TO-NOM Natsu-ACC praise-PAST 'Haru and Aki praised Natsu.'In (1), to forms a coordination phrase consisting of two proper names.We refer to this type of coordination phrase as the single to construction.Two conjuncts in (1) can be separated from each other, as shown in (2).In this case, one conjunct is marked by a case marker, and the other combines with to.We refer to this type of example as the split to construction.As shown in (3), it is possible to repeat the particle to, forming a single constituent.The coordination phrase in (3) is marked by the nominative case -ga.According to Stassen (2000), similar constructions are also observed in Burmese.Tibetan, Newari, and Canela-Krahô.We refer to this type of construction as the multiple to construction.

Previous analyses.
It has been argued that the single to construction is derivationally related to the split to construction or the multiple to construction.Kuno (1973) claims that the single to construction is derived from the split to construction by movement of the to phrase to the front of its associate.A schematic derivation of Kuno's analysis is illustrated in (4).In (4), to is introduced into the structure as a comitative postposition, but not a conjunction particle (see also Kasai and Takahashi 2001).
In contrast, Nishigauchi (1992) proposes that the split to construction is derived from the single to construction by movement of the to phrase, as shown in (5). (5) Single to: In (5), two elements form a coordination phrase, and then the first conjunct moves out of the coordination phrase in tandem with to.When the remnant coordination phrase containing the second conjunct moves over the moved to phrase, we obtain the word order given in (2).According to Nishigauchi's (1992) analysis, to in the split to construction is not a comitative postposition, unlike Kuno' (1973) analysis.Rather, it forms a coordination phrase taking two constituents.
As for the single to construction and the multiple to construction, Hiraiwa (2014) proposes that the multiple to construction is constructed based on the single to construction.Following Chino (2013), he assumes that logical connectives are head-initial in Japanese.In the single to construction, the first conjunct occupies the specifier of a projection headed by the conjunction particle, while the second conjunct appears in the complement of the particle.In the multiple to construction, another conjunction particle takes the single to construction as its complement.The complement to phrase undergoes head-to-spec movement, as shown in (6).

(6)
Single to: There are three previous approaches to to constructions in Japanese in this section.On the one hand, Kuno (1973) and Nishigauchi (1992) argue that the single to construction and the split to construction contains the same to particle.On the other hand, Hiraiwa (2014) assumes that the single to construction and the multiple to construction use the same conjunction particle.

Proposal.
In contrast to the previous approaches to the to constructions, we offer a unified analysis of the three to constructions in (1-3).We propose that the split to construction and the multiple to construction are derived from the same underlying structure.Importantly, the shared underlying structure contains the single to construction.Moreover, we argue that to in the single to construction is a conjunction particle, whereas to in the split to construction and the rightmost to in the multiple to construction are comitative postpositions.
Our proposal is illustrated in ( 7) and ( 8).Here, & stands for a conjunction particle, and we represent a comitative particle as WITH.7) is a representation for the single to construction.Here, to is a conjunction particle and it projects its own coordination phrase.Following Hiraiwa (2014), we assume that the conjunction particle is head initial in Japanese.(8a) is the underlying representation for the split to construction and the multiple to construction.In (8a), the comitative postposition takes the coordination phrase in (7) as its complement.(8b) is a derivation of the split to construction.Here, the comitative phrase is basegenerated in a theta-position, and only the first conjunct moves out of the coordination phrase to a case-position, while leaving the second element in-situ.In the remnant coordination phrase, the conjunction particle is not phonologically realized, as represented in (8b).Japanese makes use of several conjunction particles, depending on the item preceding a conjunction particle.For example, when two verb phrases are coordinated, te is used as a conjunction particle, but not to.Based on this property of Japanese conjunction particle, we assume that the conjunction particle to becomes phonologically null when it follows unpronounced elements such as a deleted copy of a moved constituent.(8c) is a derivation of the multiple to construction.Here, the comitative phrase is base-generated in a theta-position just like (8b), but the whole comitative phrase moves to a case-position, instead of the first conjunct.Note that Japanese postpositional phrases can be Case-marked, as can be seen in ( 9). ( 9) [rokuzyus-sai kara](-ga) zinsei-wa motto omosiroku naru.sixty-age from-NOM life-TOP more interesting become 'Life becomes more interesting from sixty.' 4. Support.In this section, we provide evidence for our proposal, by comparing it with the previous analyses explained in section 2. 4.1.NATURE OF TO.We first offer support for the assumption that to in the split to construction and the rightmost to particle in the multiple to construction are comitative postpositions.Among the three to constructions in question, only the single to construction does not block the association between an NP and a numeral quantifier, as shown in (10a).In contrast, the association is blocked in the multiple to construction, as in (10b).
[single to construction] student and teacher-NOM 3-CL come-PAST 'Three students and three teachers came.' 'Three people came; they are students and teachers.'b. *[ Gakusei to sensei to]-ga san-nin ki-ta.
[multiple to construction] student and teacher with-NOM 3-CL come-PAST Int.'Three students and three teachers came.' Int.'Three people came; they are students and teachers.'In (10a), the numeral quantifier can be associated with the coordination phrase, meaning that three people came and they are students and teachers.Of importance here is that (10a) can receive an interpretation in which 'three' is associated with 'teachers'.Both readings are impossible in (10b), and the sentence is unacceptable.The unacceptability of (10b) arises because the association between san-nin 'three-CL' and the nouns inside the to phrase is blocked.Under Hiraiwa's (2014) analysis, it is not clear how to capture the unacceptability of (10b).(10a) shows that phrases headed by the conjunction particle do not block the association between nouns and numeral quantifiers.He analyzes both two tos in (10b) as conjunction particles, and there should be no difference between (10a) and (10b), regarding the association between nouns and numeral quantifiers, contrary to the fact.It may be worth noting here that three-member coordination allows the association, as shown in ( 11).
(11) [&P Gakusei to sensei to zimuin]-ga san-nin ki-ta.[three-membercoordination] student and teacher and office.worker-NOM3-CL come-PAST 'Three students, three teachers, and three office workers came.' 'Three people came; they are a student, a teacher, and an office worker.'Unlike the multiple to construction, the three-member coordination phrase in (11) lacks the rightmost to particle directly followed by the nominative case marker.If two tos in the multiple to construction were the same as tos in the three-member coordination, the contrast between (10b) and ( 11) is unexpected.
In contrast to Hiraiwa's (2014) analysis, the present analysis can capture the contrast in (10).We argue that the association between an NP and a numeral quantifier is blocked in (10b) because the rightmost to in (10b) is a comitative postposition.Miyagawa (1989) argues that a numeral quantifier can be associated with a case-marked NP as in (12a), but not with an NP in a postpositional phrase, as in (12b).
(13) *[PP [&P Gakusei to sensei] to]-ga san-nin ki-ta.= (10b) student and teacher with-NOM 3-CL come-PAST In (13), the comitative postposition to takes a coordination phrase as its complement.The numeral quantifier in (10b) cannot be associated with the nouns in the coordination phrase or the coordination phrase itself because they occur in the postpositional phrase headed by the comitative postposition.On the other hand, ( 11) is acceptable because the three-member coordination phrase does not occur with a comitative postposition.
Moreover, numeral quantifiers provide evidence that to in the split to construction is also a comitative postposition.As shown in ( 14), the association between 'three' and 'teachers' cannot be achieved in the split to construction.
[split to construction] John-NOM teacher with 3-CL come-PAST 'John came with three teachers.'According to our analysis, ( 14) is predicted to be unacceptable like (12b) because the comitative postpositional phrase in (14) blocks the association of numeral quantifiers.
Further support for the assumption that there are two types of to in Japanese comes from the selectional property of to.As shown in (15), to in the single to construction is different from to in the split to construction and the rightmost to in the multiple to construction, with respect to the selectional property.
(15) a. Haru-ga [ [PP Aki-kara]  In the single to construction, two PPs can be coordinated as in (15a).However, coordination of two PPs is ungrammatical in the multiple to constructions, as in (15b).Similarly, to cannot combine with a PP in the split to construction as in (15c).We suggest that the unacceptability of (15b,c) arises because the comitative postpositions in these examples fail to select an NP as its complement.We assume that conjunction particles do not interfere with selection properties of coordinated items.
The data above show that the split to construction and the multiple to construction are comitative constructions in the sense that they contain a comitative postposition.On the other hand, the single to construction is a coordination phrase headed by a conjunction particle.4.2.DISTRIBUTIVITY.It is well-known that comitative constructions disallow distributive readings (McNally 1993).For example, the comitative preposition with in English cannot express distributive readings, as in ( 16), whereas the conjunction particle and can, as in ( 17).If the lack of the distributive reading is a property of comitative phrases, the contrast in (18) shows that the split to construction and the multiple to construction contain a comitative phrase, whereas the single to construction does not.
The lack of distributive readings in multiple to constructions and split to constructions is further confirmed by using a counting adverbial.In ( 19), the counting adverbial go-kai 'five times' is used.Only the single to construction in (19a) allows a reading in which Haru and Aki praised Natsu five times separately.Under this interpretation, there are ten praising events in total.The fact that the multiple to construction in (19b) and the split to construction in (19c) disallow the ten-event reading can be seen as evidence that they are comitative constructions.

RECIPROCALS.
There is another piece of evidence for the current analysis, in connection to reciprocal expressions.As observed by Frajzyngier (1999), Dimitriadis (2008) and Siloni (2012), when reciprocal constructions contain a discontinuous comitative phrase as in (20a), reciprocity must hold between the set denoted by a subject noun phase and the set denoted by a comitative phrase.On the other hand, in reciprocal constructions containing a coordination phrase like (20b), a reciprocal relation can be established within the set denoted by each conjunct in the coordination phrase.
[Hebrew] the-boys kissedREC with the-girls 'The boys kissed the girls, and the girls kissed the boys.' *'The boys kissed each other, and the girls kissed each other.' b. ha-yeladim ve-ha-yeladot hitnašku.
the-boys and-the-girls kissedREC 'The boys kissed the girls, and the girls kissed the boys.' 'The boys kissed each other, and the girls kissed each other.' (Siloni 2012: 297) As reported by Tatsumi (2017), a similar contrast holds between the split to construction and the single to construction, which we analyze as a discontinuous comitative construction and a coordination construction, respectively.In Japanese, when a verb is compounded with the reciprocal verb aw, which originally means 'meet', 'fit' or 'match', the resulting compound verb bears a reciprocal interpretation.In a split to construction containing a reciprocal verbal compound, the reciprocal relation holds only between the set denoted by a subject phase and the set denoted by an NP marked by to like (20a), as shown in (21a).On the other hand, when a single to construction occurs with a reciprocal verbal compound, the resulting sentence becomes ambiguous like (20b), as shown in (21b).In addition to the interpretation where a reciprocal abusing event holds between teachers and students, (21b) can receive another interpretation in which teachers abused teachers and students abused students.
teacher and student-NOM abuse-REC-PAST 'Teachers abused students, and students abused teachers.''Teachers abused each other, and students abused each other.' (Tatsumi 2017: 540) Note that multiple to constructions behave like split to constructions in this respect, as shown in (22).Like (21a), ( 22) is true only when teachers bears a reciprocal relationship with students.
(22) [Kyoosi to gakusei to]-ga nonosiri-at-ta.teacher and student with-nom abuse-REC-PAST OK 'Teachers abused students, and students abused teachers.'*'Teachers abused each other, and students abused each other.'Remember that reciprocal constructions with a discontinuous comitative phrase unambiguously receive an interpretation in which a reciprocal relation is established between a subject phrase and a comitative phrase.The unambiguity of (21a) and ( 22) can be seen as support for our proposal that the split to construction and the multiple to construction contain a comitative phrase.In other words, the comitative postposition to makes (21a) and ( 22) unambiguous.
We have seen that the multiple and split to constructions have a comitative postposition and they behave different from the single to constructions.This is unexpected under the previous approaches illustrated in Section 2. Kuno (1973) and Nishigauchi (1993) attempt to unify the single to constructions and the split to constructions.Hiraiwa (2014) argues that the multiple to constructions is constructed based on the single to phrase, analyzing to in the multiple to construction and the single to construction as a conjunction particle.4.4.DERIVATION.In this subsection, we show that the multiple to constructions and the split to constructions are derived from the same underlying structure by syntactic movement.Although the split to phrases in the examples above are associated with the subject phrase, there is a case where a split to phrase is associated with an object phrase, as in (23a).This type of association is possible only with verbs that can select a multiple to phrase as its object phrase.As shown in (23b), kuraberu 'compare' can take a multiple to phrase as its object phrase.As illustrated in (24a), homeru 'praise' cannot select a multiple to phrase as its object phrase.When the same verb is used as a matrix predicate, the split to phrase cannot be associated with the object, as in (24b).The split to phrase in (24b) is associated only with the subject phrase.5. Implication.We have seen in Section 4.4 that associations of the split to phrase are not free.For example, it associates with an object phrase in some cases, but not in other cases (cf. 23-24).Under our proposal, associations of the split comitative phrase follow from what it can co-occur with as a constituent.The idea here is that associations of the discontinuous comitative phrase are compositionally determined.We would like to suggest that this compositional way of analyzing the discontinuous comitative construction is universal.Although it is beyond the scope of this paper to give complete cross-linguistic studies on comitative constructions, we would like to show that Russian comitatives also make use of the compositional association of discontinuous comitative phrases.
Russian has two types of comitative construction as in ( 28) and ( 29).The verb in (28) exhibits plural agreement and the comitative phrase cannot be split from its associate.On the other hand, the verb in (29) shows singular agreement and the comitative phrase can be discontinuous.We call the former the coordinative comitative construction and the latter the discontinuous comitative construction.
(28) Coordinative comitatives in Russian (Larson 2014: 15) a.It has been discussed in the literature whether the discontinuous phrase is base-generated as an adjunction to a VP or its associate (Ionin and Matushansky 2002, Larson 2014, McNally 1993, Vassilieva 2001 a.o).Ionin and Matushansky (2002) argue against the VP-adjunction approach because discontinuous comitative phrases can be associated not only with subjects but also with direct objects, indirect objects, or possessives, as in (30).
( If discontinuous comitative phrases are to be adjoined to a VP, it would be unclear how to determine whether a VP-adjoined comitative associates with subjects, direct objects, indirect objects, or possessives.Instead, Ionin and Matushansky (2002) and Larson (2014) claim that the coordinate comitative construction and the discontinuous comitative construction share the same underlying structure.(31) illustrates Ionin and Matushansky's (2002) analysis on the coordinate comitative construction in (28) and the discontinuous comitative construction in (29).
(31) a. [vP (31a) is the underlying structure for ( 28) and ( 29).In (31a), the coordinated DPs by s 'with' is base-generated in a theta-position.In (31b), the whole coordination phrase moves from Spec,vP to Spec,TP and the coordinative comitative construction is derived.In (31c), on the other hand, the discontinuous comitative construction is derived via movement of the first DP Maša.
Notice that their proposal for Russian comitative constructions are similar to our proposal in Section 3.There are two types of comitative constructions in both Russian and Japanese, and they differ in what has moved from a theta-position: one involves movement of the whole element base-generated in the theta-position, whereas the other is derived by moving only the first element.Given this resemblance in analyses, it is expected that there are similarities between Japanese and Russian comitatives.This expectation is borne out.As noted by Vassilieva (2001), Russian split comitatives cannot appear with a verb that disallows a collective reading like know, as in (32a), while coordinative comitatives can, as in (32b).
(32) a. *Pelagija s Mitrofaniem znala, kto ubijca.Pelagia with Mitrofanij knew.SG who murderer b.Pelagija s Mitrofaniem znali, kto ubijca.Pelagia with Mitrofanij knew.PL who murderer 'Pelagia and Mitrofanij knew who the murderer is.' (Ionin & Matushansky 2002: 258) Likewise, Japanese split to construction does not allow the verb know as its predicate, whereas the multiple to construction does, as shown below: 6. Summary.This paper has provided the unified analysis of three types of construction containing the particle to in Japanese.We have proposed (i) that to in the single to construction is a conjunction particle, whereas to in the split to construction and the rightmost to in the multiple to construction are comitative postpositions, and (ii) that the comitative postposition to takes a coordination phrase headed by the conjunction particle to as its complement.We have also argued that the split to construction and the multiple to construction share the base structure containing the single to construction.Under our proposal, the split to construction is derived once it syntactically relates with its associate as a constituent.The proposal implies that a comitative phrase and its associate are universally a constituent in an underlying structure.We have seen that this compositional analysis of comitative phrases is also taken in Russian comitative constructions (Ionin andMatushansky 2002, cf. Larson 2014).

(
16) a. *John lives separately with Bill.b.John lives together with Bill.(Tang 2011:141) (17) a. John and Bill live separately.b.John and Bill live together.(Tang 2011:141) The multiple to construction and the split to construction cannot have distributive readings.(18a,b) is true only when there is a single event of coming.On the other hand, the single to construction in (18c) allows the multiple-event reading in which Haru and Aki came separately.(18) a. Haru to Aki to-ga ki-ta.[multiple to construction] Haru and Aki with-NOM come-PAST 'Haru and Aki came together.'*'Haru and Aki came separately.' b.Haru-ga Aki to ki-ta.[split to construction] Haru-NOM Aki with come-PAST 'Haru and Aki came together.'*'Haru and Aki came separately.' c.Haru to Aki-ga ki-ta.[single to construction] Haru and Aki-NOM come-PAST 'Haru and Aki came together.''Haru and Aki came separately.'