A Grammatical Source of Judgmental Fluctuation

Total Page:16

File Type:pdf, Size:1020Kb

A Grammatical Source of Judgmental Fluctuation

Kaken Ellipsis Workshop December 19-22, 2002 Kyushu University [The materials added in April 2006 for distribution at USC in LING 635 are in footnotes, as in #1, #2, etc. Minor corrections made in April 2006 are not indicated. English gloss and translations are not provided for most of the Japanese examples in the original handout, and they are added in the April 2006 version. In the April 2006 version, there is a "comment" on the title, stating "Title need to be worked on," which is recorded here and the comment itself has been deleted. The materials added in November 2012 are in footnotes, as in ##1, ##2, etc. In the November 2012 version, English gloss and translations are provided for more of the Japanese example. November, 2012. HH]

A Grammatical Source of Judgmental Fluctuation: A Case Study of Verbless Conjunction and Backward Multiple Stripping in Japanese Hajime Hoji USC

1. Mukai 2002 The properties in (0) are contrasted in Mukai 2002##1 with what is observed in the case of CM-stripping as given in (0).1

(0) Two observations of verbless conjunction noted in Mukai 2002: a. Subjacency effects are not observed. b. Honorification mismatch is not allowed.

(0) In the case of CM-stripping: a. Subjacency effects are observed. b. Honorification mismatch is allowed.

(0a) is accounted for in Hoji 1990 and Fukaya & Hoji 1999 by hypothesizing the involvement of movement in the derivation of CM-stripping. (0b) can be understood as an instance of what is referred to in Fiengo & May 1994 as vehicle change, a property of a certain type of ellipsis, such as VPE and stripping in English, under the assumption that CM-stripping is an instance of such ellipsis. On the basis of (0), Mukai (2002) proposes that verbless conjunction is derived by String Deletion at PF, i.e., (0a) is due to the absence of movement in the derivation of verbless conjunction and (0b) is because verbless conjunction is not an instance of ellipsis that has the property of vehicle change.

##1 I am pretty sure that "Mukai 2002" here is Mukai, Emi (2003) "On verbless conjunction in Japanese," in NELS 33, pp. 205-224, or its draft version. In her March 31, 2007 posting (chat2009 [30861]), Emi notes that Koizumi 2000: Appendix A (pp. 274-277) suggests that there may be a deletion operation that can delete non-constituent(s) only if there is sosite 'and', the operation that Mukai 2003 proposes (leaving aside the "only if …" part. Emi further points out in the posting that Mukai 2003 should have cited Koizumi 2000. (i) belowis from Koizumi 2000, being followed by (ii), according to Emi. (In the original the materials in < … > are "placed in strike-out.") (i) (Koizumi 2000, (122), according to Emi) Mary-wa Stanford-no < igakubu-ni hairi >, sosite Mary-TOP Stanford-GEN < medical school-to enter >, and John-wa Harvard-no, igakubu-ni haitta. John-TOP Harvard-GEN medical school-to entered Lit. 'Mary < entered > Stanford's < medical school >, and John entered Harvard's medical school.' (ii) "In (122), the materials placed in strike-out are optional. Since they are not a syntactic constituent, and since it is generally impossible to syntactically extract "noun + postposition" stranding a genitive modifier of that noun, whatever process responsible for this optionality is not sensitive to syntactic constituent, and it cannot be a syntactic movement." Emi also notes in the same posting that Takano 2002 provides examples with onazi 'same' and tigau 'different' under the sentence-internal reading in the sense of Carlson 1987, as problematic cases to the Verbless conjunction analysis.

Carlson, G (1987) "Same and Different: some consequences for Syntax and Semantics," Linguistics and Philosophy 10, 531-565. Koizumi, M (2000) "String Vacuous Overt Verb Raising," Journal of East Asian Linguistics 9, 227-285. Takano, Y (2002) "Surprising constituents," Journal of East Asian Linguistics 11, 243-301. 1 (i) a. Verbless conjunction A-ga B-cm, (sosite) C-ga D-cm V b. Multiple CM-stripping C-ga D-cm V. (Sosite) A-ga B-cm (da) 2. What happened at USC: Part I Some Japanese speakers (2 out of the 4 in my syntax class at USC), however, did not find particularly clear effects of (0b). 3 out of the 3 Korean speakers in the class reported the same judgments on the Korean analogues.

3. What happened at USC: Part II 3.1. "Intonation breaks" and disambiguation Saito (1987: 317) reports that Kuno (1973) "notes that (42) is acceptable when one reads this sentence assuming the bracketing in (43a), but not when one reads it assuming the structure in (43b)."2

(0) (Saito 1987: (42)) (The translation by HH) Taroo-ga Hanako-o Ziroo-ga Natuko-o Saburoo-ga Akiko-o butta 'Taro Hanako, Ziro Natsuko, and Saburo hit Akiko'

(0) (Saito 1987: (43)) a. [Taroo-ga Hanako-o, Ziroo-ga Natuko-o, Saburoo-ga Akiko-o] butta b. [Taroo-ga Hanako-o], [Ziroo-ga Natuko-o], [Saburoo-ga Akiko-o butta]

I do not share Kuno's (1973) intuitions reported in Saito 1987. But whether we read (0) as in (0a) or (0b) does seem to affect our intuitions in regards to the effects alluded to in (0) and (0).

(0) A-ga B-cm, (sosite) C-ga D-cm ... V

(0) a. A-ga B-cm (sosite) C-ga D-cm, PAUSE ... V b. A-ga B-cm, PAUSE (sosite) C-ga D-cm ... V

3.2. Possible disambiguation

Examples like (0) are what is reported in Mukai 2002 as being unacceptable (due to the honorification mismatch.

(0) Boku-ga tempura-o, sosite sensei-ga osusi-o (mina-no miteru tokoro-de) o-tukuri-ni-natta. I-NOM tempura-ACC and teacher-NOM sushi-ACC (all-GEN is:watching place-at) HON-make-past 'I tempura, and the teacher honoric-made sushi (before everyone).'

As noted above, some speakers did not find them to be particularly bad while others found them to be unacceptable. I am not sure how I read the relevant examples in the class when I got these responses. But I know that I did not pay attention to the intonation breaks. It is quite likely that the speakers read the relevant examples (in Japanese and in Korean) on their own and gave their judgments. Now, when I read (0) as in (0), even those speakers who had initially found (0) to be acceptable found it to be unacceptable.

(0) *Boku-ga tempura-o sosite sensei-ga osusi-o PAUSE I-TOP tempura-ACC and teacher-NOM sushi-ACC (mina-no miteru tokoro-de) o-tukuri-ni-natta. everyone-GEN is:watching place-at HON-make-past (Lit) 'I tempura, and the teacher sushi, made-honorific (in front of everyone watching it)'.

Cf. *Boku to sensei-wa osusi-o (mina-no miteru tokoro-de) o-tukuri-ni-natta.3 I and teacher-top sushi-acc everyone-GEN is:watching place-at HON-make-past 'The teacher and I made-honorific sushi.'

(0) Boku-ga tempura-o sosite Bill-ga susi-o PAUSE I-TOP tempura-ACC and Bill-NOM sushi-ACC (mina-no miteru tokoro-de) tukutta. everyone-GEN is:watching place-at made (Lit) 'I tempura and Bill sushi, made (in front of everyone watching it)'.

2 It is not clear, given the exposition in Saito 1987: 316-317, whether it is Kuno 1973a or Kuno 1973b that is meant here. It is possible that both Kuno 1973a and Kuno 1973b are intended. 3 The marking is on acceptability.

Ellipsis Workshop 12/2002 Hoji 2 Cf. Boku to Bill-wa (mina-no miteru tokoro-de) susi-o tukutta. I and teacher-top everyone-GEN is:watching place-at sushi-acc made 'Bill and I made sushi.'

If we read (0) as in (0), on the other hand, those who initially rejected (0) come to find it (more) acceptable.

(0) Boku-ga tempura-o PAUSE I-TOP tempura-ACC sosite sensei-ga osusi-o (mina-no miteru tokoro-de) o-tukuri-ni-natta. and teacher-NOM sushi-ACC everyone-GEN is:watching place-at HON-make-past (Lit) 'I tempura; and the teacher made sushi (in front of everyone watching it)'

The contrast seems fairly clear between (0) and (0) (and also between their Korean analogues).

3.3. Initial Hypothesis

(0) a. Verbless conjunction is structurally ambiguous; one involves String Deletion and the other is an instance of backward multiple stripping. b. We can achieve disambiguation by means of an intonation break. When it is read as in (0a), (0) is associated with the structural representation that involves String Deletion, but when it is read as in (0b), it is associated with the one that involves movement (and is an instance of backward multiple stripping).4

I repeat (0) and (0) here.

(0) A-ga B-cm, (sosite) C-ga D-cm ... V

(0) a. A-ga B-cm (sosite) C-ga D-cm, PAUSE ... V b. A-ga B-cm, PAUSE (sosite) C-ga D-cm ... V

3.4. Predictions

(0) Predictions (I): a. Even those who initially failed to detect the effects of honorification mismatch in verbless conjunction do find the effects much more clearly if the example is read as in (0a). b. Even those who initially did not tolerate honorification mismatch in verbless conjunction come to tolerate it much more readily if the example is read as in (0b).

The predictions (0) have been confirmed at USC. We also make the following predictions.

(0) Predictions (II): a. Even those speakers who initially failed to detect subjacency effects in verbless conjunction, whose judgments are reported in Mukai 2002, do find subjacency effects if the example is read as in (0b). b. Even those who might initially detect subjacency effects in verbless conjunction would come to accept 'subjacency violation cases' if the example is read as in (0a).

(0) a. John-ga tempura-o (sosite) Bill-ga susi-o, PAUSE John-NOM tempura-ACC and Bill-nom sushi-ACC zyoozuni tukuru hito-o sitteiru. skillfully make person-ACC know (Lit) 'John tempura and Bill sushi, know a person who can prepare __ skillfully.'

b. [John-ga tempura-o], PAUSE John-NOM tempura-ACC sosite Bill-ga susi-o zyoozuni tukuru hito-o sitteiru. and Bill-nom sushi-ACC skillfully make person-ACC know (Lit) ' 'John tempura and Bill sushi, know a person who can prepare __ skillfully.'

Contrary to the expectation recorded in (0a), the speakers in my USC syntax class found (0b) to be fairly acceptable.

4 Given what Saito (1987) reports Kuno (1973) states, we are led to conclude that Kuno (1973) does not accept backward multiple stripping.

Ellipsis Workshop 12/2002 Hoji 3 It is not obvious that the intonation break alone is sufficient for the relevant disambiguation, i.e., it is not clear how exact the correspondence is between a particular intonation property and a structural description associated with a given phonetic string. The fact that the prediction in (0a) is not confirmed, at least in any compelling way, indicates that the intonation break alone cannot serve as a sufficiently robust means to attain structural disambiguation.

4. What I was going to say in Fukuoka 4.1. Generalizations restated and (another) prediction

(0) Two observations of verbless conjunction noted in Mukai 2002: a. Subjacency violations are not observed. b. Honorification mismatch is not allowed.

(0) In the case of CM-stripping: a. Subjacency violations are observed. b. Honorification mismatch is allowed.

(0) Two properties of CM-verbless conjunction derived by String Deletion: a. Island effects are not observed. b. Honorification mismatch is not allowed.

(0) Two properties of verbless conjunction as an instance of multiple CM-stripping: a. Island effects are observed. b. Honorification mismatch is allowed.

(0) Prediction (III): Honorification mismatch and the absence of island effects cannot obtain simultaneously in verbless conjunction.

The unacceptability of examples like (0) and (0) is in support of the prediction in (0).

(0) *[[Boku-wa tempura-o] sosite [sensei-wa osusi-o]] PAUSE I-TOP tempura-ACC and teacher-TOP sushi-ACC zyoozuni tukuru hito-o gozonzida. skillfully make person-ACC HON:know (Lit) 'I tempura and the teacher sushi, know-HON a person who prepares __ well.'

(0) *[[Boku-wa tempura-o], PAUSE sosite I-TOP tempura-ACC and [sensei-wa osusi-o]] zyoozuni tukuru hito-o gozonzida. teacher-TOP sushi-ACC skillfully make person-ACC HON:know (Lit) 'I tempura, and the teacher knows-HON a person who prepares __ well.'

If we remove one of the two "factors," we can obtain examples that are not unacceptable. The examples in (0) do not involve "honorification mismatach," and the one in (0) does not involve an island.

(0) a. [[syatyoo-wa tempura-o] sosite [sensei-wa osusi-o]] PAUSE president-TOP tempura-ACC and teacher-TOP sushi-ACC zyoozuni tukuru hito-o gozonzida. skillfully make person-ACC HON:know (Lit) 'The president tempura and the teacher sushi, know-HON a person who prepares __ well.'

b. [[Boku-wa tempura-o] sosite [Bill-wa susi-o]] PAUSE I-TOP tempura-ACC and Bill-TOP sushi-ACC zyoozuni tukuru hito-o sitteiru. skillfully make person-ACC know (Lit) 'I tempura and Bill sushi, know a person who prepares __ well.'

(0) Boku-wa tempura-o, PAUSE sosite sensei-wa osusi-o otukurininatta.##2 I-TOP tempura-ACC and teacher-TOP sushi-ACC HON:made

##2 See (0) for a similar example where ga is used in place of wa.

Ellipsis Workshop 12/2002 Hoji 4 (Lit) 'I tempura, and the teacher prepared-HON sushi.

4.2. Further predictions 4.2.1. Vehicle change We have considered honorification mismatch as an instance of predicate vehicle change. What about argument vehicle change?

(0) a. John-ga NY-kara, PAUSE sosite John-NOM NY-from and Mary-ga Chicago-kara kanozyo-no tomodati-o syootaisita. Mary-NOM Chicago-from she-GEN friend-ACC invited 'John (invited his friend) from NY, and Mary invited her friend from Chicago.' b. *John-ga NY-kara sosite Mary-ga Chicago-kara, PAUSE John-NOM NY-from and Mary-NOM Chicago-from kanozyo-no tomodati-o syootaisita. she-GEN friend-ACC invited

Although the contrast seems clear, it would not be surprising if some speakers find (0b) to be acceptable (taking it to be an instance of backward multiple stripping), given that the intonation break by itself is not sufficient for the intended disambiguation (for some speakers). It is, however, predicted that even those who accept (0b) do not accept (0), no matter how it is pronounced, under the assumption that verbless conjunction is only two-ways ambiguous.##4

(0) *John-wa NY-kara sosite Mary-wa John-top NY-from and Mary-top

[NP [ ec Chicago-kara kanozyo-o tazunetekita] tomodati]-o Chicago-from she-acc visited friend-acc syokuzi-ni sasotta. meal-to invited 'John invited ([(a) friend(s) who had visited him) from NY] (to dinner), and Mary invited [(a) friend(s) who had visited her from Chicago] to dinner.'

(0) a. *John-wa NY-kara, PAUSE sosite Mary-wa [NP [ ec Chicago-kara kanozyo-o tazunetekita] tomodati]-o syokuzi-ni sasotta.

b. *John-wa NY-kara sosite Mary-wa [NP [ ec Chicago-kara, PAUSE kanozyo-o tazunetekita] tomodati]-o syokuzi-ni sasotta.

The examples in (0) differs minimally from those in (0), with Jane replacing John.

(0) a. Jane-ga NY-kara sosite Mary-ga [NP [ ec Chicago-kara, PAUSE kanozyo-o tazunetekita] tomodati]-o syokuzi-ni sasotta (koto) b. Jane-wa NY-kara sosite Mary-wa [NP [ ec Chicago-kara, PAUSE kanozyo-o tazunetekita] tomodati]-o syokuzi-ni sasotta.

4.2.2. Parallelism between backward multiple stripping and forward (multiple) stripping Given the view being pursued here that backward multiple stripping involves the same grammatical operations as forward multiple stripping, we predict that the contrast between (0a) and (0a), repeated here, will be replicated in their forward multiple stripping counterparts.

(0) a. John-ga NY-kara, sosite Mary-ga Chicago-kara kanozyo-no tomodati-o syootaisita.

(0) a. *John-wa NY-kara, sosite Mary-wa [NP [ ec Chicago-kara kanozyo-o tazunetekita] tomodati]-o syokuzi-ni sasotta.

(0) a. Mary-wa Chicago-kara kanozyo-no tomodati-o syootaisita; sosite John-wa NY-kara (da). b. *Mary-wa [NP [ ec Chicago-kara kanozyo-o tazunetekita] tomodati]-o syokuzi-ni sasotta, sosite John-wa NY-kara (da).#1

##4 The use of wa in some examples and that of ga in others do not affect the relevant empirical point in question. The status of (0) and that of (0), for example, should not be affected by the choice between wa and ga, leaving aside how hopeless (0) may turn out to be. #1 With the possibility of the 'E-type pronoun strategy', introduced in Merchant 2002 and adopted in Fukaya 2002?, (0b) can be acceptable, with the interpretation of the second conjunct corresponding to something like (i). (i) As for John, [the person who he invited for dinner] was him from NY.

Ellipsis Workshop 12/2002 Hoji 5 4.2.3. Parallelism with CM-comparatives#2

4.3. Remaining issues 4.3.1. 'Non-constituent deletion' 4.3.2. Mix readings 4.3.3. Conjunction and sloppy readings with kare#3 ##5

(0) a. *John to Bill-ga kare-no heya-o soozisita (koto) John and Bill-NOM he-gen room-acc cleaned 'Intended as: Each of John and Bill cleaned his own room.' b. ?*John-ni yorimo sakini Bill-ni PAUSE John-DAT than early Bill-DAT sensei-ga kare-no heya-o soozisaseta (koto) teacher-NOM he-GEN room-ACC clean:caused (With the sloppy reading) 'Intended as: The teacher made Bill clean his room earlier than the teacher made John clean his room.' c. John-ni yorimo sakini PAUSE sensei-ga Bill-ni kare-no heya-o soozisaseta (koto) (With the sloppy reading)

4.3.4. Other issues

5. What happened in Fukuoka 5.1. Predicate vehicle change

(0) a. Boku-wa syussya syatyoo-wa taisyamae-ni I-TOP coming:to:company president-TOP leaving:company:prior-at kanarazu koohii-o nomu.5 always coffee-ACC drink 'I (always drink coffee pre-)company-show-up, and the president always drinks coffee pre-company-leaving'. More natural translation: 'I (always drink coffee before) coming to the company, and the president always drinks coffee before leaving the company.'

b. *Boku-wa syussya syatyoo-wa taisyamae-ni kanarazu koohii-o mesiagaru. (The (b) example differs from the (a) example only in that the form of the verb "drink" is in its honorific form in (b).)

With the genitive no missing before mae, it is perhaps reasonable to conclude that syussyamae (出社前) and taisyamae (退社前) are not phrasal. Whether it is phrasal or not, what is "missing" in the first conjunct in (0a) and (0b) is clearly not a constituent. The status of (0b) shows that non-constituent 'deletion' is not compatible with (i.e., does not "license") honorification mismatch, as observed earlier in section 4.1. A pause and the addition of sosite 'and' cannot seem to save (0b). "║" stands for a pause.

(0) *Boku-wa syussya ║ sosite syatyoo-wa taisyamae-ni kanarazu koohii-o mesiagaru.

This is as expected in light of the incompatibility of String Deletion and honorification mismatach (an instance of vehicle change). The status of (0) seems to mirror that of (0).

#2 Given that cm-stripping and cm-comparative have the same formal properties, we expect cm-comparative to exhibit similar, if not the identical, properties as cm-stripping, including multiple cm-comparatives of the form in (i). (i) NP1cm NP2cm yori … An expanded version of the present discussion will include the relevant materials. #3 The point of this subsection is that two NPs contained in the 'remnant' of multiple (backward) stripping behaves like a conjoined NP, but they do not behave as such in String Deletion. As indicated in (0a), BVA(John to Bill, kare) is not possible. And the status of (0b) is very much like that of (0a), and (0b) contrasts sharply with (0c). BVA(Toyota to Nissan, soko) is possible. As expected, the "Toyota … Nissan" counterparts of (0b) and (0c) [with soko replacing kare 11/2012, HH] are both acceptable. This has interesting implications on various analyses of 'surprising constituents' such as those by Takano Kawazoe, and Koizumi. ##5 Even with, or possibly because of, the remarks in #3, it is not clear what is intended by (0). We don't even have String Deletion cases or cases of multiple (backward) stripping in (0). I suppose that I meant to compare String Deletion and multiple (backward) stripping with cases where the conjoined NP intended to serve as an antecedent of kare in (0a) and two "types" of comparative ellipsis. The exposition in this subsection seems to be just too abbreviated. 5 Examples of verbless conjunction with 'non-constituent deletion' are discussed in Mukai 2002.

Ellipsis Workshop 12/2002 Hoji 6 (0) *Syatyoo-wa taisyamae-ni kanarazu koohii-o mesiagaru; boku-wa syussya (da).

The status of (0) seems to indicate that the status of (0) is independent of honorification mismatch.#4

(0) *Syatyoo-wa taisyamae-ni kanarazu koohii-o nomu, boku-wa syussya (da).

(0) does not have honorification mismatch; but it is still clearly unacceptable, in contrast to the examples in (0).##4a

(0) a. Syatyoo-wa taisyamae-ni kanarazu koohii-o nomu, boku-wa syussyamae-ni (da). b. Syatyoo-wa taisyamae-ni kanarazu koohii-o mesiagaru, boku-wa syussyamae-ni (da).

5.2. Argument vehicle change

(0) John-wa syussya ║ Mary-wa taisyamae-ni kanozyo-no yuuzin-ni denwasita.##6 (*Sloppy reading) (0) John-wa syussyamae-ni ║ sosite Mary-wa taisyamae-ni kanozyo-no yuuzin-ni denwasita. (OKSloppy reading)6

It is not clear how acceptable the sloppy reading is in examples like (0), but I suspect that it is more readily acceptable than the sloppy reading in (0).

(0) ?John-wa syussya Bill-wa taisyamae-ni kare-no yuuzin-ni denwasita. (with the Sloppy reading)

The status of the sloppy reading in (0) and (0) seems to mirror that of the distributive reading in (0a) and (0b).#5

(0) a. *John to Mary-wa taisyamae-ni kanozyo-no yuuzin-ni denwasita. b. ??/?*/*John to Bill-wa taisyamae-ni kare-no yuuzin-ni denwasita.

--Can kare be B of FD(A, B)? --Perhaps yes. But, for some reason (which is perhaps related to the somewhat unique nature of the modern-day use of kare), it cannot seem to be B of FD(t, B) if the t is the trace of a plural-denoting expression. --This leads us to expect reconstruction effects with kare, as long as its antecedent is not plural-denoting. This seems to be a correct result.#6

Boku 'I' cannot seem to be used as a bound variable.

(0) *[boku dake]-ga boku-no tiimu-ni toohyoosita (koto) 'only I voted for my team'

==> Boku cannot be B of FD(A, B). Can boku be I-indexed?#7

#4 I am not sure what was intended here, but my guess is as follows. --Honorific mismatch is possible only in an ellipsis context. --There is a restriction on multiple stripping in regard to [The sentence was not even completed… 11/2012, HH] ##4a The examples in (i) were provided around here in the original file, but I am not sure what the intended point was. (i) a. Boku-wa syussyamae-ni ║ sosite syatyoo-wa taisyamae-ni kanarazu koohii-o mesiagaru. b. *Boku-wa syussyamae-ni sosite syatyoo-wa taisyamae-ni ║ kanarazu koohii-o mesiagaru. ##6 The placement of the pause elsewhere would not improve the acceptability with the intended sloppy reading. 6 The status of (i) is not entirely clear. (i) John-wa syussyamae-ni sosite Mary-wa taisyamae-ni ║ kanozyo-no yuuzin-ni denwasita. It seems to depend upon how the intonation break disambiguates the relevant structural ambiguity. Some speakers, including myself, find the sloppy reading in (i) difficult or not possible. Similarly, many speakers do not find the sloppy reading available in (0), contrary to what is reported here. #5 The acceptability marking on (0b) differs from what is noted in (0a); see also note #3 above. I am pretty sure that the marking on (0b) reflects the fact that some speakers (possibly only one) at the Fukuoka workshop accepted BVA(John to Bill, kare) without any problems, and some to varying degrees. This reminds us of the observation in Hoji et al. (GLOW) that for a very small group of speakers kare 'he' can be 0-indexed in the terms of Hoji et al. (GLOW) (not indexed in the terms of Ueyama 1998). (Hoji et al. (GLOW) presents Ueyama's (1998) theory of anaphoric relations slightly differently from the original for the purpose of exposition. #6 I do not remember what I had in mind here. I suspect that I was considering reconstruction effects involving a sloppy reading, rather than BVA(A, B) where A corresponds to 'which N', but I am not sure. #7 It is not entirely clear what was intended by the examples in (0)-(0). The absence of the judgments on these examples makes it difficult to figure out the original intension. The examples in (0) do not seem to be constructed correctly. I will try to figure it out later.

Ellipsis Workshop 12/2002 Hoji 7 (0) John-ni yorimo sakini sensei-wa boku-ni boku-no heya-o soozi-saseta 'Professor made me-DAT clean my room before John-DAT'

(0) a. John-ni yorimo sakini sensei-wa boku-ni PAUSE boku-no heya-o soozi-saseta

b. John-ni yorimo sakini sensei-wa PAUSE boku-ni boku-no heya-o soozi-saseta

(0) a. John-o osieta sensei-ni yorimo sakini gakutyoo-wa Bill-o osieta sensei-ni kare-ga totemo yuusyuu datta to iwaseta. —Sloppy okay.

b. Boku-o osieta sensei-ni yorimo sakini gakutyoo-wa Bill-o osieta sensei-ni kare-ga totemo yuusyuu datta to mitomesaseta.

5.3. Subjacency

(0) (Cf. (0) and (0).)#8 a. *[Boku-ga tempura-o subayaku] [sensei-ga susi-o zyoozuni tukuru hito-o gozonzida]. I-NOM tempura-ACC quickly teacher-NOM sushi-ACC skillfully make person-ACC know:honorific

b. [John-wa tempura-o subayaku] [Mary-wa susi-o zyoozuni tukuru hito-o sitteiru]. John-TOP tempura-ACC quickly Mary-TOP sushi-ACC skillfully make person-ACC know

6. Summary

(0) Clear correlations are observed among: a. the possibility of vehicle change b. subjacency effects c. the possibility of 'non-constituent deletion'

We have discussed an instance of predicate vehicle change (honorification mismatch) and one of argument vehicle change. The instances of 'non-constituent deletion' are (i) the syussya 'reporting to the company' example (with syussya being the remnant), (ii) the relative clause example (with the relative clause, and not including the head of the relative clause, being the remnant) and (iii) the NP-no example.7 The factors in (0) also seem to affect the choice between the two representations, but their effects are less clear.

(0) a. intonation break b. the presence and absence of sosite

The considerations summarized above suggest that there are two distinct types of verbless conjunction. I have suggested that (i) one type is derived by String Deletion, as argued in Mukai 2002, and (ii) the other is an instance of multiple stripping. The multiple stripping in verbless conjunction shares all the formal properties with 'regular' multiple stripping, with the only crucial difference being that the former is 'backward'. Recall that CM-ellipsis requires a linguistic antecedent, as an instance of surface anaphora. 8 It seems that the linguistic antecedent requirement can be satisfied in two ways; (i) by the relevant ellipsis site appearing at PF after what can serve as its antecedent and (ii) by the relevant ellipsis site appearing in the same sentence as what can serve as its antecedent though not necessarily appearing after it. The linguistic antecedent requirement is satisfied in backward multiple stripping differently than in forward multiple stripping in the sense just noted; and what is involved in the satisfaction of the condition in the case of the former might result in some degree of its marginality for some speakers. Aside from this difference, we expect backward multiple stripping and forward multiple stripping to share their properties and we have observed that they do.

(0) Summary of the generalizations (which I believe can be established):

#8 The point intended here seems to be made already in an earlier discussion on (0) and (0), and it is not clear what new point was intended here. The two examples in (0) do not form a minimal pair, to begin with. The intended contrast should be illustrated by (i) instead of (0a) on the one hand and (0b) on the other, or by a similar pair with the nominative marker replacing the topic marker. (i) *[John-wa tempura-o subayaku] [sensei-wa susi-o zyoozuni tukuru hito-o gozonzida] 7 Examples of (ii) and (iii) are not provided here. 8 I suppress the issues raised and addressed in Merchant 2002.

Ellipsis Workshop 12/2002 Hoji 8 MS: multiple stripping SD: verbless conjunction based on String Deletion: a. Vehicle change is possible in MS but not in SD. b. Subjacency effects are observed in MS but not in SD. c. 'Non-constituent deletion' is possible in SD but not in MS.

The empirical demonstration of the validity of the propositions in (0) has been attempted by showing the following.

--A weaker form of the demonstration ('weaker' because the effects of the factors in (0) are not as clear as one might want/expect):

(0) The intonation break and the presence or the absence of sosite correlates with the properties in (0).

--A stronger form of the demonstration:

(0) a. Subjacency effects are clearly observed (when an island is involved in the structure in question) in verbless conjunction only if vehicle change is 'attempted'.#9 b. Vehicle change is possible only if the relevant structure in question does not involve 'non-constituent deletion'.

Crucial in the demonstration noted above is (i) an independent demonstration that vehicle change is possible with certain lexical items (and (ii) some not totally ad hoc theoretical characterization of the relevant observation).

(0) Predicate vehicle change

(0) Argument vehicle change a. kare 'he' b. kanozyo 'she' c. boku 'me'

(0) The initial generalization/observation (which has not yet been confirmed with a high degree of repeatability) Vehicle change is possible with (0a) and (0b), but not with (0c).

7. Some more thoughts 7.1. FD-based BVA, co-I-indexation-based BVA, and Mix readings

--boku as B of FD(A, B)? ==> No. --FD-based sloppy readings with boku? ==> No. --Co-I-indexation-based sloppy readings with boku? ==> No. --Co-I-indexation-based sloppy readings with kare? ==> Yes. --FD-based sloppy readings with kare? ==> Yes. Then, kare as B of FD(A, B)? ==> Yes (as long as A of FD(A, kare) is not the trace of an NP/DP that is 'plural-denoting').

--Verbless conjunction based on String Deletion does not give rise to the Mix reading pattern, but backward multiple stripping does.

(0) Toyota to Nissan-ga [ soko-ga soko-no .... to] hapyoosita (koto) announced

Assumption: The Mix reading pattern arises only based on FD.

(0) T1-ga [soko1-ga soko1-no ... to] V N2-ga [soko2-ga soko2-no ... to] V

(0) T1-ga [soko1-ga soko1-no ... to] V

N2-ga [soko1-ga soko1-no ... to] V

(0) T1-ga [soko1-ga soko1-no ... to] V N2-ga [soko2-ga soko1-no ... to] V

#9 With "only if," it is being claimed that there is no other way to force the "multiple stripping" analysis over the "String Deletion" analysis of a given verbless conjunction. Whether such is indeed the case will have to be determined on the basis of further research and the proposition in (0a) must be understood as a reflection of our (well, my, I guess) current understanding of the relevant issues.

Ellipsis Workshop 12/2002 Hoji 9 FD(t, soko(-ga)), T1 and soko(-o) are co-I indexed.

(0) T1-ga [soko1-ga soko2-no ... to] V N2-ga [soko1-ga soko1-no ... to] V

FD(t, soko(-ga)), N1 and soko(-o) are co-I indexed.

(0) John-wa syussya Bill-wa taisya-zi-ni kare-ga kare-no buka-o sikatta to itta. (The Mix reading pattern does not obtain, no matter how we read this.)

(0) a. John-wa syussyazi-ni, ║ sosite Bill-wa taisya-zi-ni kare-ga kare-no buka-o sikatta to itta.

b. John-wa syussyazi-ni, ║ sosite Bill-wa taisya-zi-ni kare-no zyoosi-ga kare-o sikatta to itta. (The Mix reading pattern obtains.)

7.2. Remaining issues (Multiple stripping in English)

(0) *Bill to NY, and Susan went to Paris.

If (0) is indeed out, why is it that English does not allow backward multiple stripping?

(0) Susan went to Paris, {and(/but)} Bill to NY.

Can (0) be analyzed as an instance of multiple stripping (in addition to being an instance of gapping)? If the answer is no, what could be the reason why English does not allow multiple stripping (while Japanese does)?

Ellipsis Workshop 12/2002 Hoji 10

Recommended publications