Chapter 2 Literature Review
2.2 The Reduction Analysis
2.2.5 Hsieh (2017)
Although the proposals in Liu (1996) and Chung (2006) can perfectly account for the
examples used in their papers, the three problems encountered by the Reduction Analysis, as
presented in Xiang (2003), are still not solved by them. In order to argue for the Reduction
Analysis, Hsieh (2017) proposes some possible solutions to these problems, which are
summarized as follows.
To begin with, Hsieh suggests that the distribution of dou in the bi-comparatives can be
explained with MaxElide proposed in Merchant (2008), as shown in (90).
(90) MaxElide (Merchant 2008)
Let XP be an elided constituent containing an A’-trace. Let YP be a possible target for
ellipsis. YP must not properly contain XP.
In other words, MaxElide regulates that ellipsis always deletes the maximal possible targets.
In this way, the paradigm below in (91) - (93) can be accounted for with MaxElide as well as
the appropriate places where dou is base-generated.
(91) mei-ge nusheng dou bi mei-ge nansheng (*dou) congming every-CL girl all BI every-CL boy all smart ‘Every girl is smarter than every boy.’
(92) mei-ge nusheng *(dou) bi Zhangsan (*dou) congming every-CL girl all BI Zhangsan all smart
‘Every girl is smarter than Zhangsan.’
(93) Zhangsan bi mei-ge nusheng ?/*(dou) congming Zhangsan BI every-CL girl all smart
‘Zhangsan is smarter than every girl.’
Take (91) as an example. It has the structure shown in (94) below.21
(94)
21 In Hsieh (2017), dash-line boxing marks the elided constituent, and the solid-line boxing indicates its antecedent.
Here, dou is assumed to be base-generated under AP. Based on MaxElide, the ellipsis operation
targets at AP; thus, the dou in the bi-constituent is elided (in the dash-line boxing). Then, the
remaining dou (in the solid-line boxing) moves to adjoin to TP in order to be in the same
Spell-out domain with the subject which it modified. This is why dou does not appear in the
bi-constituent in the surface structure of this sentence. According to Hsieh (2017), dou cannot be
base-generated under vP in this sentence since ellipsis only targets maximal projections.
However, the maximal projection here is the AP rather than the vP. Therefore, MaxElide targets
at AP and the dou adjoined to vP cannot be elided, which causes ungrammaticality because we
have two tokens of dou in the surface structure, contrary to the fact. Sentences (92) and (93)
can be analyzed in a similar way. The only difference is that in these two sentences, dou is
assumed to be base-generated under vP rather than AP. Consider the structure of (92) in (95).
(95)
As mentioned above, dou here is base-generated within vP, instead of AP. If this is not the case,
there will be no maximal projection that can be deleted. As a result, the predicate inside the
bi-constituent cannot be elided, which leads to ungrammaticality. However, if dou is under vP,
the AP will be the maximal projection that satisfies MaxElide. With the deletion of the
predicate in the bi-constituent and the movement of dou to the TP domain, the correct surface
word order can be successfully derived. Then, let’s see what happens in (96), the structure of
the sentence in (93).
(96)
Again, in this structure, dou is base-generated under vP. If it is in the AP, the predicate within
the bi-constituent cannot be deleted. Same as (95), MaxElide is satisfied by the ellipsis of AP,
the maximal projection. Then, we can derive the grammatical sentence.
As for the second and third problem, i.e., the lack of subcomparative and embedded
standard, Hsieh (2017) adopts two theoretical notions, Agree and the rescuing effect from
PF-deletion. For the former, Hsieh adopts Wurmbrand’s (2014) definition of Agree, and the latter
has been proposed in Boskovic (2011), Merchant (2009), and many others. Moreover, there
are three assumptions. Firstly, all functional clausal heads (T, Mod, Asp…) have an
interpretable T(ense)-feature (iT) which is typically valued; all verbal heads (V, A…) have an
uninterpretable T-feature (uT) which is typically unvalued. Secondly, an unvalued feature is
not allowed at the interface (PF/LF), so it must undergo Agree with the closet valued feature.
Thirdly, Hsieh assumes that the bi-constituent contains a vP, and lacks all the higher functional
heads.22 With the two notions and the three assumptions, the two ungrammatical sentences in
(97a) and (98a) can be explained. Consider (97a) first, which is an example of subcomparatives.
Hsieh provides its structure in (97b).
(97) a. *zhe-zhang zhuozi bi na-shan men kuan chang
this-CL table BI that-CL door wide long
Intended: ‘The table is longer than the door is wide.’
b.
22 According to Hsieh (2017), the assumption that bi takes a small clause (vP) as its complement and lacks all the higher functional heads is inspired by Pancheva’s (2006) analysis of certain types of phrasal comparatives in Slavic languages.
As we can see, there is an uninterpretable feature on the predicate kuan ‘wide’. With the
assumption that the bi-constituent lacks all the higher functional heads, the uninterpretable
feature on the predicate cannot be valued inside the PP formed by bi. The only possible head
with interpretable valued feature is the T head; however, the feature valuation is blocked by
the PP, which is an adjunct island. Then, since the uninterpretable unvalued feature cannot be
valued, it becomes a fatal feature. Hence, the derivation crashes, and the sentence is
ungrammatical. Now, we can turn to (98), an example of embedded standard.
(98) a. *zhe-zhang zhuozi bi Lisi renwei (ta) kuan
this-CL table BI Lisi think it wide
Intended: ‘This table is wider than Lisi thinks it is,’
b.
The same logic can be applied to the explanation for the ungrammaticality of (98). The
uninterpretable unvalued feature on the predicate renwei ‘think’ cannot be valued inside the
bi-constituent, nor can it get valued by the interpretable feature on the T head because of the
PP adjunct. Thus, it turns out to be a fatal feature, which crashes the derivation at PF.
With the three solutions provided, Hsieh (2017) suggests that the reduction analysis has a
greater advantage over the direct analysis in capturing the syntactic and semantic properties of
Mandarin bi-comparatives.
At first glance, it seems that Hsieh’s (2017) proposal is capable of solving the three
problems of the Reduction Analysis. However, with a closer examination, there are three
possible problems of Hsieh’s analysis.
First of all, in the analysis of the dou’s distribution in bi-comparative, Hsieh claims that
ellipsis only targets at maximal projections, the phrase level. However, the example in (99)
shows that ellipsis can also happen to a non-maximal projection.
(99) John ate a cake in the park, and Bill did at home.
He suggests that dou should be base-generated under AP when both the matrix subject and the
subject inside the bi-constituent are universal; however, if ellipsis can happen at the
intermediate level, Hsieh’s account would become problematic because in this way, dou can
also be base-generated under vP in the situation mentioned above, just like the dou in sentences
involving only one universal subject. By assuming that all the tokens of dou in his analysis are
base-generated under vP, it will make the analysis more consistent and elegant.
The second possible problem in Hsieh (2017) is that the solution which Hsieh uses to deal
with the lack of embedded standards seems to be unable to explain the grammatical sentence
in (100).
(100) Zhe-zhang zhuozi bi Lisi renwei de kuan this-CL table BI Lisi think DE wide ‘This table is wider than Lisi thinks it is.’
In Hsieh’s proposal, the bi-constituent is a PP, which is an adjunct island. It will block the
feature valuation between the T head and the verb renwei ‘think’. Hence, based on Hsieh’s
analysis, the sentence in (100) will be ungrammatical; however, this is contrary to the fact
because example (100) is grammatical.
Finally, the third problem in Hsieh’s proposal pertains to one of his assumptions. That is,
the post-bi constituent contains a vP and lacks all the higher functional clausal heads such as
Tense, Aspect, or Modal. However, the sentence below indicates that the post-bi constituent
can be larger than a vP, as illustrated in (101).
(101) Zhangsan mei-you xie gongke bi Lisi mei-you zuo jia-shi haiyao zaogao
Zhangsan not-have write homework BI Lisi not-have do chore more bad
‘That Zhangsan does not do his homework is worse than that Lisi does not do the chores.’
The post-bi constituent in (101) contains negation mei ‘not’ and the aspect marker you ‘have’,
both of which are higher than vP. Some may regard Lisi mei-you zuo jia-shi ‘Lisi does not do
the chores’ in (101) as a sentential subject since there is an elided predicate zaogao ‘bad’. In
this way, the whole post-bi constituent can still be a vP. However, in sentence (102) below,
which involves topicalization of the object keben ‘textbook’ within the bi-constituent, it is
difficult to claim that keben wang-le dai ‘forget to bring the textbook’ is a sentential subject.
Therefore, the assumption that the post-bi constituent contains a vP and lacks all the higher
functional clausal heads is still problematic.
(102) Zhangsan mei-you xie gongke bi keben wang-le dai hai zaogao
Zhangsan not-have write homework BI textbook forget-Asp bring more bad
‘That Zhangsan doesn’t do his homework is worse than that he forgets to bring his textbook.’
In sum, although Hsieh (2017) tries to argue for the Reduction Analysis, the three
problems faced by his analysis suggest that the Reduction Analysis alone still cannot account
for all the relevant data here.