Comparatives in English & Japanese
Emi Mukai (USC; )[*]
Draft of 2nd screening paper, version.10, 10/29/05
1. Introduction
I may probably delete the divisions by subsections.
1.1. Comparative constructions in English and Japanese
The sentences in (1a) and (1b) are instances of comparative constructions in English. The Japanese translations of them are given in (2). In the cases of (1a) and (2a), the predicates are adjectives. On the other hand, the adjectives modify the nouns in the case of (1b) and (2b). I call the former simply comparatives, while the latter NP comparatives.
(1) Comparative constructions in English
a. This book is longer [than XP] comparatives
b. Bill bought a longer book [than XP] NP comparatives
(2) Comparative constructions in Japanese
a. kono hon-wa [YORI XP-yori] nagai comparatives
this book-top XP-yori long
b. Bill-wa [YORI XP-yori] nagai hon-o katta NP comparatives
Bill-top XP-yori long book-acc bought
Examples which compare the cardinality of nouns such as in (3) are also an instance of NP comparatives in this paper.
(3) John bought more books than Bill did/bought.
This paper concerns whether the functional parallelism (i.e., the availability of translations) between English comparative constructions and Japanese ones lead us to assume that they are formally parallel. Some researchers indeed assumed/proposed that Japanese comparative constructions are the counterpart of English ones, while others claim that they are not exactly parallel though they contain some element(s) which share(s) some formal properties with each other. In this paper, however, I argue against those analyses and claim that the comparative meaning in Japanese comparative constructions comes in a way different from English comparative constructions.
Before going into the discussion, I introduce some terms which I use in this paper for the sake of exposition. I call the bracketed part in (1) and (2) the than constituent and the yori constituent, respectively, and the XP part is called the yori/than phrase/clause.
1.2. The difference between English and Japanese
Though (1) and (2) seem parallel to each other, there are some differences between English and Japanese comparative constructions.
1.2.1. The lack of direct measure phrases
English overtly differentiates an absolute adjective (e.g., tall) from a comparative one (e.g., taller), and a measure phrase (henceforth MP) can co-occur either of the forms. However, the interpretation of MPs depends on where they occur.
(4) a. John is 1m tall. 1m: direct MP
b. John is 1m taller. 1m: differential MP
On the other hand, Japanese lacks an overt morpheme which distinguishes an absolute adjective and a comparative one, and an adjective in Japanese can be translated into either of the forms in English. However, once an MP occurs right before an adjective, it should always be interpreted as a differential MP.
(5) John-wa 1m se-ga-takai
John-top 1m tall
= 'John is 1m taller (than someone else)'
=/= 'John is 1m tall'
1.2.2. The presupposition requirement
The second difference is related to a certain interpretation. It seems the case for both English and Japanese that neither the subject of a matrix clause (e.g., John) nor the noun in the yori/than clause/phrase (e.g., Bill) need to satisfy the property indicated by the predicate (e.g., tall), as shown in (6).
(6) a. John is taller than Bill (is). Neither John nor Bill has to be tall.
b. John-wa Bill-yori segatakai Neither John nor Bill has to be tall.
John-top Bill-yori tall
However, when the than/yori clause contains another adjective, English and Japanese come to differ from each other.
(7) a. #John is taller than Bill is tall. Neither John nor Bill has to be tall.
(at least for some speakers)[1]
b. John-wa [Bill-ga segatakai (no) ] yori segatakai.[2] John and Bill should be tall.
John-top Bill-nom tall no yori tall
For the ease of exposition I introduce (8).
(8) Presupposition Requirement (PR)
The argument of a predicate should satisfy the property expressed by the predicate.
The observation here can be recaptured in a way that PR can be irrelevant in English comparative constructions, while in the case of Japanese, PR becomes relevant if the yori constituent contains an adjectival predicate.
A similar contrast is obtained in how questions, which are traditionally analyzed on a par with comparative constructions in the way that how is grouped together with a comparative morpheme plus the than constituent (with putting details aside).
(9) a. How tall is John? No presupposition that John should be tall.
b. John-wa doredake segatakai no? It is presupposed that John is tall.
John-top how:much:degree tall Q
There are some English adjectives which seem to behave just like Japanese adjectives with respect to PR.
(10) a. How sad was he? It is presupposed that he was sad. (i.e., PR is relevant.)
b. John is more lonesome than Mary is sad.
It is presupposed that John is lonesome and Mary is sad. (i.e., PR is relevant.)
Kennedy & McNally (2005: sec. 4) introduce the distinction, relative and absolute adjectives as shown in (11), with citing Yoon (1996), Rotstein & Winter (2004), Unger (1975) among others. It seems safe to categorize sad into absolute adjectives.
(11) a. Relative adjectives: the standard of comparison is context dependent.
(e.g., tall, short, long, old, interesting, large, small, expensive etc.)
b. Absolute adjectives: the standard of comparison is not context dependent.
(e.g., , awake, full, flat, visible, open, closed, bent, straight etc.)
Based on the apparent parallelism between (7b) and (9b) on the one hand and (10) on the other, one might suspect that Japanese does not have the relative-absolute distinction and every adjective is parallel to absolute adjectives in English. However, it is not the case, since while the PR is not irrelevant for (6b) (repeated below), it is still relevant for (12).
(6) b. John-wa Bill-yori segatakai Neither John nor Bill has to be tall.
John-top Bill-yori tall
(12) John was sadder than Bill was. It is presupposed that John was sad.[3]
Therefore we need to investigate what differentiate (7b) and (9b) from (7a) and (9a) without recourse to the type of adjectives.
1.2.3. Subcomparatives (initial observation)
Sentences in which the than clause contains a different adjective from the matrix clause are called subcomparatives.
(13) a. The table is longer than the door is wide. (Bresnan 1973: (262))
b. #The table is longer than the door is old.
It seems the case that subcomparatives should satisfy the requirement in (14).
(14) Same Dimension Requirement[4]
A comparative construction is semantically well-formed only if the compared objects are compared along the same dimension.
Kennedy points out that even if the (deleted) predicate in the than clause is the same with that in the matrix clause as in (15), the sentence is still not good if it violates SDR.
(15) #The class was longer than this table is. (Kennedy 1997: chap 1, (15))
It has been observed that subcomparatives corresponding to (15) are not possible in Japanese (Snyder et al. 1995, Beck et al. 2004, Nakanishi 2004 among others).[5], [6], [7]
(16) *Kono tana-wa [ano doa-ga haba-ga-hiroi (no) yori] takai.
This shelf-top that door-nom wide no yori tall
'This shelf is taller than that door is wide.'
To sum up so far; an analysis for the comparative constructions in Japanese should be the one which can capture those differences between English and Japanese.
1.3. The summary of Main claims
---To be completed.
I claim in this paper that Japanese lacks a degree morpheme corresponding to English more/er, and that yori does is not the counterpart of than. Furthermore, though it has been widely assumed that gradable adjectives establish a relation between an object in their domain and a degree (i.e., a point on a scale) (cf. Cresswell 1976, von Stechow 1984 among many others[8]), I entertain the idea that every adjective, regardless of whether it is gradable or not, is universally of type <e,t>, based on the difference between English and Japanese as has been summarized in section 1.2. above. The summary of my main claims is in (17).
(17) Main claims
a. Every adjective, regardless of whether it is gradable or not, is universally of type <e,t>. The difference between gradable adjectives and non-gradable adjectives is whether a standard (which varies depending on a context) is supplied or not; it is in the case of the former, but not in the case of the latter.
b. The functional category #, which is of type <et, d,e,t, introduces a scale (i.e., a point in a totally ordered set).
c. A gradable adjective (<e,t>) in English has some feature, which should be checked off by # (<et,d,et). The mother node of a gradable adjective and # is therefore of type <d,et>.
d. Japanese does not have #. As a consequence, Japanese lacks a constituent of type <d,et>.
e. The yori constituent is an adjunct modifying an AP.
In short, while English comparatives have a structure in (18), the structure of a Japanese comparative is in (19).
(18) IP
3
Subj 6
...... #P
3 <d,et>
DegP #'
3 3
er 5 # AP <e,t>
than XP <et, <d,et>
(19) IP t
3
kono hon-wa e I' <e,t>
this book-top 3
AP <e,t> I
3
<et,et> PP nagai <e,t>
3 long
XP e yori <e,etet>
1.4. The organization of the paper
---To be added.
2. Previous analyses ---'yori' as being analyzed as parallel to than
2.1. The standard analysis of English comparatives (Bresnan 1973; von Stechow 1984; Heim 2000 among others)
Let us first briefly review the standard analysis of English comparatives. First of all, it has been widely agreed that an adjective is a two-place predicate and it is a function from a degree to a set of individuals (<d,et>).
(20) [[Adjective]] = ld. lx. x is Adjective to d
(e.g., [[tall]] = ld. lx. x is tall to d)
Adjectives then take DegPs as their first argument; i.e., the semantic composition of them is done by Functional Application.
(21) Functional Application (FA) (Heim & Kratzer 1998: 44: (3))
If a is a branching node, {b, g} is the set of a's daughters, and [[b]] is a function whose domain contains [[g]], then [[a]] = [[b]]([[g]]).
Adjectives are either be absolute (e.g., tall) or comparative (e.g., taller), as shown in (22).
(22) a. John is tall
b. John is 6 feet tall
c. That door is wider than this shelf is tall.
d. John is (1 inch) taller than Bill is.
Under the analysis, whether adjectives are realized as absolute or comparative depends on what the first argument of the adjectives, DegP, is like. Moreover, the DegP in the absolute tall in (22a) to (22c) is occupied by a different element from each other.
(23) 3
John e 3
is AP <e,t>
3
DegP tall <d,et>
In the case of (22a), the DegP is occupied by what is called the positive operator (henceforth the POS operator) in (24a). DegP in (22b), on the other hand, is an MP as in (24b), and it in (22c) is a simple null operator Op, which gets mapped to l in semantics (the motivation for assuming Op inside the than constituent will be addressed later). In the case of comparative, the DegP is headed by the degree morpheme (er/more/less etc.) as in (24d).
(24) a. POS <dt,t>
b. 6 feet d
c. Op
d. <dt,t>
3
DegP = (1 inch) 3
er 3
than 6
Bill is
The specification for the POS operator is in (25).[9]
(25) [[POS]] = lP<d,t>.id[P(d)=1 and d>ds] (Cf. von Stechow 1984: 60, R6)
(where ds is the standard degree which is contextually identified.)
The semantics of degree morpheme er is either (26a) or (26b) (the latter is used when comparatives contain a differential MP), which contains the maximality operator (see (27)).
(26) a. [[er]] = lD'<d,t>. lD<d,t>. max(D) > max(D') <dt,dtt>
b. [[er2]] = lD'<d,t>. lD"d. lD<d,t>. max(D) - max(D') = D" <dt,<d,dtt>
(27) von Stechow 1984: (117)
Max(P) is true of d iff P(d) and ~$d'[P(d') & d' > d]
Since er is assumed to take the than complement as its first argument, the than clause must undergo extraction at PF.[10]
It has been widely assumed since Chomsky (1977) that the than clause involves a movement, because, as shown in (29) and (30), it exhibits the properties listed in (28), which "serve[s] as a kind of "diagnostic" for wh-movement" (Chomsky 1977: 86).
(28) (Chomsky 1977: 86, (49))[11]
a. It leaves a gap.
b. Where there is a bridge [verb], there is an apparent violation of subjacency, PIC and SSC.
c. It observes CNPC
d. It observes wh-island constraints
(29) Comparatives 1 (=(C52))
a. Mary isn't the same as [she was five years ago]
b. Mary isn't the same as [John believes [that Bill claimed [that she was five years ago] ] ]
c. *Mary isn't the same as [John believes [the claim [that she was five years ago] ] ]
d. * Mary isn't the same as [I wonder [whether she was five years ago] ]
(30) Comparatives 2
a. Mary isn't taller than [she was five years ago]
b. Mary isn't taller than [John believes [that Bill claimed [that she was five years ago] ] ]
c. *Mary isn't taller than [John believes [the claim [that she was five years ago] ] ]
d. * Mary isn't taller than [I wonder [whether she was five years ago] ]
Though Chomsky (1977) regards what gets moved in the case of (30a), for example, as a whole AP as in (31a), it is DegP that undergoes movement at LF under the standard analysis as in (31b).[12] The AP must be deleted at PF under the latter analysis.
(31) a. ... than [Opi [she was [AP ti ] five years ago] ]
b. ... than [Opi [she was [AP [DegP ti] tall] five years ago] ]
The argument for (31b) is that comparatives can be captured by the same manner with so called subcomparatives as in (32a).
(32) a. The table is longer than the door is wide. (Bresnan 1973: (262))
b. (LF of the than clause under the standard analysis) ... than [Opi the door is [AP [DegP ti ] wide] ]
Under the move, the difference between comparatives and subcomparatives can be reduced to whether the deletion is involved or not.
In sum, the sentences in (33) are analyzed as in (34), (35), (36) and (37), under the standard analysis.