April 2006 version 3.0
H. Hoji
Vehicle Change: Its instances and characterization in Fiengo and May 1994
[This is an incomplete summary of what is meant by vehicle change and how it is used in Fiengo and May 1994. This document should be read with A. Ueyama's 1997 handout and the one prepared by Candice and Soyoung for LING 635, spring of 2006, both of which have been uploaded in the GGES Library.]
1.Instances of vehicle change
1.1.Plural vs. singular, person
1.1.1.Examples (I)
(1)(Based on an example in footnote 7 (p. 101))
Everyone saw his mother, and Oscar did, too
= '… and Oscar saw their mothers, too'
(2)(Based on footnote 7: (i) (p. 101))
If every student revises his paper, then I won't have to.
= '…, then I won't have to revise their papers.
1.1.2.Remarks (I)
--The E-type pronoun analysis
--But not clear to me whether this should really be an instance of an E-type link…
--For example, what about the PF precedence requirement?
--Need to look at JP analogues.
1.1.3.Examples (II)
(3)(Based on F&M's (23)[1])
Max and Oscar said they are going to Europe this summer, and Sally did too.
'…, and Sally said she is going to …
(4)a.(F&M's (61))
I turned in my assignment, but most of the other students didn't
b.(F&M, p. 218)
I've never heard myself sing, but you have
1.1.4.Remarks (II)
(5)a."Let us call this aspect of reconstruction, where syntactic form that expresses a given argument is in some sense altered among the tokens of the reconstruction, vehicle change. What is the domain of vehicle change? Simply put, vehicle change is closed under invariance of indices. That is, in a reconstruction, a nominal can take any syntactic form so long as its indexical structure (type and value) is unchanged (modulo identity for -occurrences[2])." (p. 218)
b."…, we can discern two sorts of vehicle change that fall these parameters.
The first sort of vehicle change occurs when there are various ways of realizing one sort of expression—for example, the indiscernability of pronoun, reflexive, and PRO with respect to the nominal typology." (pp. 218-219)
--Then they go on to suggest:
c."This indiscernibility of lexical and null forms of a cell extends to vehicle change between a name and a variable (trace)."
--and cite an example like (6a), representing it as in (6b).[3]
(6)a.(F&M's (64))
John kissed Mary, but I wonder who Harry did.
b.(F&M's (65))
John kissed Mary1, but I wonder who1 Harry kissed e1
(7)"The second sort of vehicle change is indiscernibility of values within syntactic feature paradigms." (p. 220)
--By 'syntactic features', they must at least mean 'agreement features', but they also mean feature responsible for the distinction between someone and anyone. They also seem to mean by 'syntactic features' binding theoretic features. But then it is not clear why they call this the 'second sort of vehicle change', distinct from the first sort noted in (5b)…
(8)(F&M's (66), p. 220, attributing it to W. Ladusaw)[4]
a.Max talked to someone, but Oscar didn't
b.Max didn't talk to anyone, but Oscar did
(9)(F&M's (67), p. 220)
Mary loves John, and he thinks that Sally does, too
(10)"Proper names have the feature [-pronoun] … [but] it can be realized as [+pronoun]. Let us call this product of vehicle change the pronominal correlate of the name." (p. 221)
--But what does (10) mean? How does a Name enter the Numeration? Would it enter the Numeration, optionally as [−pronoun]? But what would "a Name having the feature [−pronoun] mean if it could enter the Numeration with the feature [+pronoun]?
--The answer to the question seems to be that it actually enters the Numeration as [−pronoun]. They state (11), perhaps one of the most crucial points to bear in mind.
(11)"[the] diacritic [for the pronominal correlate of a Name, pNP] is not part of the syntactic structure of pronominal correlates. In this regard, it is only notational, as opposed to the markings for pronominal feature and indexical type…," (footnote 23)
--Given "product of vehicle change" in (10), one might get an impression that vehicle change is an operation that yields an output. "[V]ehicle change is operative" in their footnote 24 also gives one that impression. But vehicle change is said to be an "aspect of reconstruction" (see (5a)) and reconstruction is not an operation…
--They also mention "the reconstructed NP is a pronominal." (p. 221) This raises at least two questions. This gives the impression that the NP under discussion has been "reconstructed". But, as noted, "reconstruction" in question is not an operation…[5]
1.2.Crucial paradigms of vehicle change (I): a pronominal correlate of a Name
(12)(F&M's (67), with the italics added)
Mary loves John, and he thinks that Sally does, too
(13)(F&M's (71), with the italics added)
a.Mary hit John, and he did too
b.Mary introduced John to everyone, and he did, too
(14)(F&M's (73), with the italics added)
a.Mary fired Max, although he shouldn't have
b.Mary fired Max, although heshouldn't have been
--(14b) is meant to be an Eg corresponding to the Eg* in (14b) under the reading Max=he. But when would (15) below be felicitous, anyway?
(15)Mary fired Max, although Max shouldn't have fired himself.
So, to make the intended point, we must first construct an Eg like (16), and see if its corresponding Eg* (such as (17)) is as unacceptable as predicted.
(16)Mary fired Max, although Max would never have fired himself in a similar situation.
(17)Mary fired Max, although he would never have in a similar situation.
--How hopeless is (17), anyway?
--More crucially, if we allow the reflexive form of the pronominal correlate of a Name, as in F&M's account of (18), it is not clear why the "Max shouldn't have fired Max" readingshould be disallowed in (14b)? This issue relates to their view of anaphoric relations and how they propose to express them; cf. section 1 of A. Ueyama's handout, to be distributed later.
1.3.Crucial paradigms of vehicle change (II): the reflexive form of the pronominal correlate of a Name
(18)(F&M's (79), p. 224))
a.I shaved John because he wouldn't
b.I disliked hearing Yeats read in an English accent, and he did, too
c.Mary believes Max to be heroic, and he does, too
F&M's account of the acceptability of (18) is that "the elided NP can be the reflexive form of the pronominal correlate of the overt NP John," which "yield[s]" "the representation in [(19)]."[6] (p. 224)
(19)(F&M's (80))
I shaved John1 because he1 wouldn't shave pJohn1+self
--So, how does "the reflexive form of the pronominal correlate of a Name" enter the Numeration? As something like John-self? Is that in the Lexicon? See footnote 7 above.
--Suppose that we allow a form like John-self and it can be 'taken as' its pronominal correlate (plus -self). But there should be a restriction in that case that prevents a form like John-self from appearing on the surface.
1.4.Crucial paradigms of vehicle change (III): a pronominal correlate of a reflexive
(20)(F&M's (46a), (47a), (48a), and (49a), with the italic being added here)
a.Burr defended himself against the accusations better than Luther Martin could have
b.Bush voted for himself, but Barbara didn't
c.Yeats disliked hearing himself read in an English accent, and I did, too
d.Max believes himself to be heroic, and Mary does, too
1.5.Crucial paradigms of vehicle change (IV): a reflexive form of a pronoun (or would it be called a reflexive correlate of a pronoun with a diacritic like ahim?)
(21)(F&M's (46b), (47b), (48b), and (49b), with the italic being added here)
a.Luther Martin defended Burr against the accusations better than he could have
b.Barbara voted for him, but Bush didn't
c.I disliked hearing him read in an English accent, and Yeats did too
d.Mary believes him to be heroic, and Max does, too
1.6.The 'grandmother' readings
I fail to follow the discussion here.
(22)a.(F&M's (75), p. 222)
Mary corrected her mother's mistakes before she did
b.(F&M p. 222)
Mary corrected her mother's mistakes, and she did too
--It is claimed that the second conjunct of (22b) cannot have the interpretation in (23).
(23)Mary's mother corrected her mother's mistakes.
--And I did not follow their reasoning.
--Furthermore, they state in their footnote 25, the reading represented in (23) becomes available with pragmatic and lexical adjustments. And if so, the unavailability of the reading in question should not be attributed to a structural reason. Here they would object, it seems, and might insist that it is due to a structural reason, noting, as they 'conjecture' in footnote 25, that when the reading is available, the two NPs that are crucially being considered for being coreferential are not co-indexed (i.e., with the coreference arising without co-indexation)… Hmm. Where is falsifiability?
2.Reconstruction and i-copies
2.1.Reconstruction
F&M's use of reconstruction is not totally consistent throughout the book. They initially introduce it as an identity relation (pp. 93-94), and then later refer to it as "a set of token structures under a syntactic identity condition"(p. 191), and also as "an identity condition" (p. 192).[7] They also use it as if reconstruction is a process, as in "the first person pronoun would be reconstructed as a first person singular pronoun." (p. 218)
According to the first remark (well, based on the Index in F&M) on reconstructions, it seems to be a descriptive term used as in (24). (p. 94)
(24)Occurrences of some one sub-phrase maker over a given terminal vocabulary are reconstructions of one another.[8]
(24) is introduced in the context of stating "the narrowest structural identity condition that will do the job [of determining whether given portions of distinct sentences count as being identical]. In this sense, it is perhaps safe to assume that the notion reconstruction of in F&M is indeed meant to express structural identity. See also (25).
(25)"It is in this manner, then, that the formal identity condition on indices—i-copies—contributes to a more general notion of syntactic identity for a language, setting a necessary, but not sufficient, condition for reconstruction, and hence for the characterization for the information structure of connected parallel discourse." (p. 96)
(26)"Because they are redundant, reconstructions need not be repeated in each subsequent utterance of a conversation."[9] (p. 97)
So, for now, I want to conclude as in (27) although reconstruction is often used in F&M without having an overt object (as in reconstruction of x).
(27)Reconstruction is an identity relation and is to be used as A is a reconstruction of B, A and B are reconstructions of each other, A, B, and C are reconstructions of one another.
2.2.I-copies
I-copies is a notion that is meant to express a relation between two or more indexical dependencies, as in (28) while indexical dependencies are said to be of the form in (29). (F&M (p. 95))
(28)"Indexical dependencies ID and ID' are i-copies if and only if ID and ID' vary from each other in at most the value of I."[10] (p. 95)
(29)<(c1, c2, ..., cn), I , SD
(30)The sequence (c1, c2, ..., cn)
"The elements of the sequence (c1, c2, ..., cn) are bearers of the occurrences of the index I; such sequences must contain one and only one element bearing an -occurrence of I—the antecedent—and at least one bearing a -occurrence of I—the dependents." (p. 52)
(31)Indices
"[I]ndices … are complex objects, consisting of a value and a type." (p. 83)
More information about and some illustration of Fiengo & May's proposal are provided in A. Ueyama's handout uploaded in the Library ("Handouts_Plus/Handouts_on_Fiengo_and_May_1994").
2.3.Reconstruction and ellipsis
(32)(p. 236)
a."A reconstruction is a set of occurrences of a (sub-)phrase marker over a given terminal vocabulary."
b."When a sentence contains a reconstruction, it is possible to dispense with the phonetic realization of otherwise present syntactic structure of some of the occurrences."
c."However, ellipsis and reconstruction are distinct notions, and their relation may be complex."
I take (32b) as stating that having a reconstruction is a sufficient condition for ellipsis[11], and if that is what is intended, the relation between reconstruction and ellipsis can be articulated, rather than leaving it as vague as in (32c).
3.Some Eg*s in F&M 1994 and their analogues in Japanese
3.1.Unavailability of a sloppy reading due to the absence of parallelism (I)
The sloppy reading as indicated in (33b) is said to be unavailable for (33a).
(33)a.(F&M: chap. 3, (28))
Max's mother saw him, and Oscar said Mary did, too.
b.Max's mother saw Max, and Oscar said Mary saw Oscar.
Fiengo and May's (1944: chap. 3) account is as follows: (i) a sloppy reading arises only with the use of (a pronoun with) a -occurrence[12](of an index) 'as a sloppy pronoun', (ii) -occurrences can appear in reconstructions only if the relevant dependencies are i-copies, and (iii) the ID in (34b) and the one in (34c) are not i-copies of each other.[13]
(34)(F&M: chap. 3, (29); the brackets in (a) and and in (b) and (c) have been added here)
a.*Max1's mother [VP saw him1β], and Oscar2 said Mary [VP saw him2β]
b.<(Max, himβ), 1, <NP,N,V,NP>
c.<(Oscar, himβ), 2, <NP,V,NP,V,NP>
Presumably, the anaphoric relations as indicated in (35) are possible, where the elided VP in (33) is realized overtly.
(35)(Cf. (33a).)
Max1's mother saw him1, and Oscar2 said Mary saw him2, too.
Also missing in (33a) is the reading in (36), where the him in the second conjunct 'refers to' someone other than Max or Oscar.
(36)Max1's mother saw him1, and Oscar2 said Mary saw him3 (too).
Just as the reading in (36) is missing in (33a), so the readings in (37a) and (38a)are missing in (37b) and (38b), respectively.
(37)a.Max1 invited his1 roommate, and Oscar2 invited his3 roommate (too).
b.Max1 invited his1 roommate, and Oscar2 did too. (=>Does not have the reading in (37a).)
c.Max1 invited his1 roommate, and Oscar2 did it too.
(38)a.Max1 invited his1 roommate, but Oscar2 didn't invite his3 roommate.
b.Max1 invited his1 roommate, but Oscar2 didn't. (=>Does not have the reading in (38a).)
c.Max1 invited his1 roommate, but Oscar2 didn't do it.
We might note that the missing readings are available or at least not impossible in (37c)/(38c), given an appropriate context for the determination of what is intended by it. The missing readings in (36), (37a) and (38a) all involve the referential use of a pronoun, so to speak; i.e., his3 is not 'dependent on' Max1 or Oscar1. A satisfactory account of the unavailability of the reading in (35) for (33a) must therefore state not only (i) how the dependency as indicated in (34a) is blocked in (33a) but also (ii) how the coreference between Oscar2 and him2 is blocked in (33a), with him2 being 'referential'. An explanation of the latter should also account for the missing reading in (36), (37a) and (38a). One such account of the latter can be as follows, in the terms of Fiengo and May 1994.[14] Assuming that (33a), repeated here, is represented as in (39) at LF, leaving aside the details deliberately, the interpretation in (36), also repeated here, is not possible for (33a) because the two VPs in (36) (see (40)) cannot be reconstructions of each other.
(33)a.(F&M: chap. 3, (28))
Max's mother saw him, and Oscar said Mary did, too.
(39)Max's mother saw him, and Oscar said Mary saw him, too.
(36)Max1's mother saw him1, and Oscar2 said Mary saw him3 (too).
(40)a.[VP saw him1]
b.[VP saw him3]
The two VPs are not reconstructions of each other (i) if one pronoun is an -occurrence and the other a -occurrence, or (ii) if the two pronouns are both -occurrences. Finally, the two VPs are not reconstructions of each other (iii) even if both pronouns are -occurrences since the two IDs that they are part of are not i-copies of each other. See the Ueyama handout on Fiengo and May 1994.
Now, if the NOC in Japanese were to be analyzed always on a par with VPE in English, we would expect that the missing readings noted above in the English VPE examples would also be missing in the corresponding NOC examples in Japanese. The Japanese NOC examples, however, actually behave more like do it examples such as (37c) and (38c) than like VPE examples, clearly indicating that the NOC in Japanese cannot always be analyzed on a par with VPE in English. This conclusion is consistent with the independent observation that the NOC in Japanese can be used felicitously without any linguistic antecedent, in sharp contrast to the unmarked uses of VPE in English.
3.2.Some details
Before proceeding, I would like to address some details that we should be paying attention to in evaluating the relevant empirical materials. Consider again (33a) and the missing reading for it in (35), repeated here.
(33)(F&M: chap. 3, (28))
Max's mother saw him, and Oscar said Mary did, too.
(35)Max1's mother saw him1, and Oscar2 said Mary saw him2, too.
One might consider (41) as the direct Japanese NOC analogue of (33).
(41)Max1-no hahaoya-ga kare1-o mita, sosite Oscar-ga [Mary-mo ec mita] to itta.
'Max1's mother saw him1, and Oscar2 said [Mary also] saw ec.'
It is difficult to accept the interpretation in (35) for (41); the most salient reading of the second conjunct in (41) is "Oscar said Mary saw Max," given that the first conjunct means "Max's mother saw Max."[15] One might suggest that the missing reading in (35) for (41) can be taken as evidence for analyzing the Japanese NOC on a par with VPE in English.
We must again ask whether the claim under discussion is (42a) or (42b).
(42)a.The Japanese NOC must always be analyzed on a par with VPE in English.
b.The Japanese NOC can be analyzed on a par with VPE in English.
If the claim is (42b), it is not falsifiable for the reasons that we have discussed in class, and the absence of the reading in (35) for (41) is only compatible with it, and does not provide particularly strong evidence for it. I.e., we must be told how else the NOC in Japanese can be analyzed and how the reading in (35) is blocked for (41) under such alternative analysis/ses of the NOC in Japanese.
Under the claim in(42a), on the other hand, the absence of the reading in (35) for (41) seems to provide direct evidence for it. I.e., it would be accounted for in the same way as the missing reading in (35) for English (33a). Such a claim, however, makes negative predictions elsewhere. For example, it predicts that the Japanese NOC analogue of (37b) and (38b) behave like (37b) and (38b), rather than like (37c) and (38c) (or some other surface forms that are distinct from VPE and are presumably instances of deep anaphora), disallowing the readings in (37a) and (38a), respectively, even under a context that allows such readings for (37c) and (38c). (37) and (38) are repeated here.
(37)a.Max1 invited his1 roommate, and Oscar2 invited his3 roommate (too).
b.Max1 invited his1 roommate, and Oscar2 did too. (=>Does not have the reading in (37a).)
c.Max1 invited his1 roommate, and Oscar2 did it too.
(38)a.Max1 invited his1 roommate, but Oscar2 didn't invite his3 roommate.
b.Max1 invited his1 roommate, but Oscar2 didn't. (=>Does not have the reading in (38a).)
c.Max1 invited his1 roommate, but Oscar2 didn't do it.
One such context where (37c) allows the interpretations in (37a) is as follows: There has been a prior arrangement made between Max and Oscar that Max would invitehis own roommate and Oscar in turn would invite Bill's roommate, because of some special relations among the individuals involved.
Returning to (41), consider the Japanese example in (43), whichis structurally identical to (41).
(43)Max-no sensei-ga kare-o hometigitta, suruto
Max-gen teacher-nom he-acc praised-to-the-sky and:then
Oscar-wa [Mary-ga __ hometigitta nitigainai] to itta.
Oscar-top Mary-nom praised-to-the-sky must:be:the:case that said
'Max's teacher praised him to the sky; and Oscar said that Mary must have praised__ to the sky.'