Journal of Language and Linguistics Vol. 3 No. 1 2004 ISSN 1475 - 8989

Scope Asymmetries between English and Japanese

Eriko Sato

Stony Brook University, New York, USA

Abstract

This paper examines scope interactions between quantificational NPs (QPs) in simple transitive sentences in English and Japanese, and provides an account of well-known English/ Japanese scope asymmetries: unlike the case in English, Japanese sentences (in unmarked word order) are always scopally unambiguous, systematically disallowing wide scope of object over subject. I argue that this difference is not due to a radical structural difference between the two languages, but is due to a lexical variation of a particular functional head, which I label ‘Dst’. The two languages differ in the selectional properties of Dst. Dist in Japanese is overtly expressed by mo and selects only NPs and PPs whereas Dst in English is covert and English speakers equate the selectional prosperities of Dst with those of conjunction.

1. Introduction

This paper gives an alternative account on well-known scope asymmetries between English and Japanese, found in sentences like (1a-b). In (1a), either the subject or the object takes scope over the other, and the choice of reader may differ depending on the choice of book, whereas in (1b), the subject asymmetrically takes scope over the object, and the choice of reader may not differ depending on the choice of book (Kuroda 1970, May 1977, 1985, Hoji 1985, 1986, Aoun and Li 1989, 1993a, Hornstein 1995):

(1)a.Some student read every book.

bdareka-ga dono hon-mo yonda

someone-NOM every book-MO read

‘Someone read every book.’

Two major proposals have been made in literature regarding English/Japanese scope asymmetry. One assumes these languages differ with respect to the applicability of the constraint that preserves surface command order at LF (the rigidity condition): it applies to Japanese but does not apply to English (Hoji 1985, 1986). The other proposal assumes that these languages share the same rigidity condition but have basically different surface structures (Huang 1982, Aoun and Li 1989, 1993a, Hornstein 1995). This paper offers a third proposal for the cause of scope asymmetry: English/Japanese scope asymmetry is due to the difference in the morphological properties of a particular functional head. It does not assume any difference between the two languages with respect to rigidity conditions nor with respect to basic surface structures.

In section 2, I show that scope asymmetry between English and Japanese is, in fact, very limited, because the wide scope of the object over the subject (the inverse scope) is, in fact, very limited in English. Recent literatures report that the inverse scope is available in English only when the object is a distributive universal (Beghelli 1993) and the subject is an existential like some N(s) (Sato 1996). I will present my basic assumptions and proposals that explain this fact and then, show that some lexical variation with respect to distributive universals is responsible for the scope asymmetry between the two languages. In section 3, I will summarize the descriptive analysis of distributive universals in Japanese in literature (Kuroda 1965, Nishigauchi 1986, Ohno 1989), and propose that the particle mo in Japanese corresponds to the distributive head (Dst) proposed by Beghelli (1993). In section 4, I show that Japanese scope facts can be explained on a par with English scope facts under my proposal. In section 5, I argue that the scope asymmetry between the two languages can be reduced to the difference in morphological properties of Dst.

2. English/Japanese scope asymmetries

2.1. The Suppressing Effects in English

The idea that not all the English sentences are ambiguous, contrary to what May claims, is first persuasively suggested by Liu (1990). Liu argues that many sentences that are claimed to be ambiguous by May are, in fact, not ambiguous if we exclude the so-called “independent (branching) reading,” where the values of QPs are determined independently of each other.[1] For example, a sentence like (2) is claimed to be ambiguous between a distributive reading and a non-distributive reading; however, in fact, it has three logically possible interpretations represented in (2’):

(2)Every man loves some woman.

(2’) /

(2’a) represents a wide scope of the subject over the object reading, where the value of the object (some woman) depends on the value of the subject (every man). (2’b) represents a wide scope of the object over the subject reading, where the value of the subject (every man) depends on the value of the object (some woman). Finally, (2’c) represents an independent reading, where there is no dependency between the two QPs (some woman and every man.) Importantly, since a QP like every man intrinsically cannot vary in its value depending on the value of another QP, (2’b) and (2’c) turn out to be logically equivalent. Because of this reason, it is not clear whether the non-distributive reading found in (2) is a wide scope of the object (2’b,) or an independent scope (2’c.)

Whether a wide scope of the object is available or not becomes clear once we choose a QP that potentially takes narrow scope (e.g. some student) as a subject, as in (3a-b):

(3)a.Some student read every book.

(=(1a))

b.Some student read more than two books.

In these sentences, a wide scope of the object and an independent reading are logically distinct. Surprisingly, the wide scope of the object is available only in (3a), but not in (3b): in (3b), the choice of the student is somewhat “fixed” and cannot vary. Liu notes that it does not mean that modified numeral NPs like more than two books are generically incapable of taking scope over other QPs. In a different construction as in (4), they can take scope over other QPs: in (4), the choice of book potentially varies depending on the choice of student:

(4)More than two students read some book.

By refining Liu’s descriptive observation and examining broader data, Beghelli (1993) concludes that the wide scope of the object is available only when the latter is a distributive universal like every N. Thus, unlike (3a), the wide scope of the object is unavailable in the sentences in (5) (Liu 1990, Ruys 1992, Beghelli 1993, Beghelli and Stowell 1997.):

(5)a.Some student read more than two books.

(=(3b))

  1. Some student read no book.[2]
  2. Some student read two books.[3]
  3. Some student read most books.[4]

As Beghelli correctly predicts, the following sentences all allow the wide scope of the object over the subject, since the object is a universal QP: the sentences in (6) mean that each of the books was read by a “possibly different” set of at least two / exactly two / more than two / two/ a few / many / severalstudents:

(6)a.At least two students read every book.

  1. Exactly two students read every book.
  2. More than two students read every book.
  3. Two students read every book.
  4. A few students read every book.
  5. Many students read every book.
  6. Several students read every book.

However, Beghelli’s prediction does not always hold. Sato (1996, 2003) shows that when certain QPs, namely, most N, no N, and few Ns appear in subject position, the wide scope of the universal in object position is systematically suppressed (the Suppressing Effect.) Unlike the sentences in (6), sentence (7) does not allow the wide scope of the object over the subject:

(7)Most students read every book.

In (7), the inverse scope reading would be paraphrased as “Each of the books was read by a possibly different set of students constituting a majority.” However, this reading is not available, and the choice of the students is somewhat fixed.[5] The presence of most Ns in subject position somehow suppresses the wide scope of the universal in object position. This can be further confirmed by the contrast between the two sentences in (8): the inverse reading is unavailable in (8a), but available in (8b):

(8)a.Most people approved every bill.

b.Many people approved every bill.

In (8a), the inverse scope reading would be paraphrased as “For each of the bills, there is a possibly different set of majority of people who approved it.” This reading would be compatible with a situation where every bill was passed, and thus it should be the preferred reading. However, this reading is unavailable. The only reading (8a) offers is the surface scope reading, which is paraphrased as “Each of the people that constitute a majority approved every bill,” and the choice of the people is somewhat fixed. By contrast, (8b) allows the inverse scope reading, which is paraphrased as “Each of the bills was approved by many people.” In this reading the choice of people may vary depending on the choice of bills.

Similarly, QPs such as no N and few Ns cause the Suppressing Effect. Consider (9a-b):

(9)a.No student read every book.

b.Few students read every book.

Sentence (9a) allows the surface scope reading, paraphrased as, “There was no student who read every book,” but it does not allow the inverse scope reading, which would be paraphrased as, “Each of the books had no student reader.” If (9a) had an inverse scope reading, the sentence would be judged false if some students read some books, but not all of them; however, the sentence is still judged to be true in such a context. Evidently, the sentence does not allow an inverse scope reading. Similarly, sentence (9b) allows the surface scope reading, paraphrased as, “There were few students who read every book,’ but it does not allow an inverse scope reading, which would be paraphrased as, “Each of the books was read by few students.” If there were an inverse scope reading, the sentence would be judged to be false when each book was read by many students although there were few students who read all of the books; however, the sentence is still judged to be true in such a situation. Apparently, the sentence does not have an inverse scope reading.

2.2. Relativized minimality

In Sato (2003), the account of the Suppressing Effect shown above is given under the assumptions and proposals summarized below:

(10)

a. Quantificational QPs bear their own quantificational force whereas non-quantificational QPs acquire it by being bound by an external unselective binder arbitrarily far away (Milsark 1974, 1977, Heim 1982, Reinhart 1995, Beck 1996, etc.).

b. The QPs that cause the Suppressing Effects (the suppressing QPs) (e.g. most Ns, no N, and few Ns) can only be quantificational but other QPs (e.g. some N(s), many Ns, a few Ns, two Ns) can be either quantificational or non-quantificational.[6]

c. QPs can basically remain in situ without undergoing Quantifier Raising (QR). QR applies only when it leads to a distinct interpretation (Fox 1995, Reinhart 1995), raising only quantificational features [quant], and adjoining them to some head (Chomsky 1995).[7] The scope of QPs is presumably the c-command[8] domain of [quant] or [quant]-bearing QPs.

d. Only quantificational QPs bear a [quant] feature (Abe 1993), which represents their quantificational force and forms a feature bundle with semantic features that represent the restriction of quantification.

e. Only distributive universals bear a distributive feature [Dst] that must be checked against the head of the Distributive Phrase (DstP), the latter of which occurs at sentence level (Beghelli 1993).[9]

f. A [quant] feature is raised only if it gets a free ride from other formal features such as [Dst].[10] Otherwise, [quant] remains in situ.

g. The movement of a [quant] feature is subject to the minimal Q-chain Constraint (MQCC), defined in (11).[11]

(11)The Minimal Q-Chain Constraint

Q-chains must be minimal where:

(i)Q-chain A is minimal if and only if there is no Q-chain B, such that the head of B c-commands the tail of but not the head of A and,

(ii)C=(i...n) is a Q-chain if and only if:[12]

(a)i is a [quant] feature or a [quant]-bearing element

(b)ilocally binds i+1

(c)n occupies its unique Case(-checking) position[13]

(d)C is maximal

Under these assumptions, we can give a straightforward analysis on the lack of the inverse scope in the following sentence:

(12)Most students read every book.

In order for the universal to take scope over the subject (most students), the [quant] feature of the former must be raised along with [Dst], beyond the latter. If this happens, (12) is mapped to the LF in (12’):

(12’) /

In (12’), a DstP projects above AgrP, and there are two Q-chains. One is the Q-chain whose head is [quant]i and whose tail is [t]i, and it is represented by a line ( ). The other is the Q-chain whose head and tail are [quant]j, and it is represented by a dot ( . Crucially, these chains are nested in (12’), and thus it is ruled out by the MQCC. Note that the MQCC basically disallows nesting and intersecting Q-chains but allows separate Q-chains.

The grammaticality of sentence (12) can be accounted for if [quant] of the universal can be reconstructed into Spec of AgrOP, after [Dst] is checked off, as in (12”):

(12”) /

In general, I allow [quant] features to be reconstructed into previous positions as long as it does not violate the MQCC. In this case, since [quant] is reconstructed into a Case-checking position, we can satisfy the requirement that the tail of a Q-chain must be in Case position.[14] In (12”), there are two chains. One is the Q-chain created by [quant]j which occurs in the Spec of AgrSP. It is marked by a dot (). The other is the Q-chain created by [,quant]i,, which is in the Spec of AgrOP. It is also represented by a dot (). The latter chain is a legitimate chain since its tail occurs in Case-checking position, and its head coincides with it. [t]i that is adjoined to Dst is not part of this Q-chain since it occurs higher than the head of the chain, and it is presumably deleted at the end of the derivation. Crucially, these Q-chains are separate, and thus the MQCC is satisfied, correctly predicting the grammaticality of the sentence. Since the head of the Q-chain created by most students c-commands the head of the Q-chain created by every book in (12”), we correctly predict the wide scope of most over every, where the choice of students does not vary depending on the choice of book.

Consider, next, sentence (13):

(13)Some student read every book.

Unlike in (12), the Suppressing Effect does not occur in (13), and the universal object may take scope over the existential subject. This can be also straightforwardly accounted for in the current analysis. Determiners like some may or may not bear [quant], unlike determiners like most. (13) is mapped to the LF in (13’a) if some bears [quant], but it is mapped to the LF in (13’b) if some does not bear [quant]:[15]

(13’) /

The representation in (13’a) contains two Q-chains that are nested, violating the MQCC. On the other hand, the representation in (13’b) contains only one Q-chain, which is the one created by every book. This is because some student in (13’b) does not bear [quant] and fails to form a Q-chain. Accordingly, (13’b) does not violate the MQCC. In (13’b), some student is unselectively bound via Existential Closure (Heim 1982). Following Reinhart (1995), I assume that some student in (13’b) introduces a (choice) function variable (f), which is bound by a default unselective existential operator () that may occur anywhere in the structure. If  occurs between AgrS and DstP as in (13”), we correctly predict the wide scope of every over some in sentence (13):

(13”) /

In (13”), two chains are nested, but crucially only one of them is the Q-chain, and thus the MQCC is not violated. (13”) correctly predicts the reading where the universal takes scope over the existential. The latter can be informally paraphrased as “for each of the books x, there is a (possibly different) function f that selects a set of “some student” Y such that x was read by (all the members of) the set Y.” This means that each of the books was read by a possibly different student.

2.3. The residue of the Japanese/English scope asymmetry

The analysis sketched above shows that the wide scope of the object over the subject is available in English only when the following two conditions are met:

(14)i.The subject is not a suppressing QP.

ii.The object is a distributive universal.

Consequently, English/Japanese scope asymmetries are also limited to these circumstances. In all other cases, English simply behaves like Japanese. Informally speaking, English comes close to resembling so-called “scope rigid” languages such as Chinese, German, Hungarian, and Japanese (Kuroda 1970, Huang 1982, Hoji 1985, 1986, Aoung and Li 1989, 1993a, Kiss 1991, Beck 1996). This suggests that the above conditions may also be relevant to the scope asymmetry between the two languages. Two candidate hypotheses are the following:

(15)

  1. There is an asymmetry between English and Japanese in terms of quantificational/non-quantificational status of QPs: Japanese has no determiner that can function non-quantificationally, as a cardinal adjective, whereas English has some such determiners.
  2. There are some differences between distributive universals in English and Japanese.

The hypothesis in (15i) is unlikely to hold.[16] The remaining hypothesis is (15ii), which states that there is some difference between distributive universals in English and those in Japanese. As stated above, I adopted Beghelli’s proposal that distributive universals in English are raised to check their distributive feature [Dst] against a functional head Dst. We must examine whether the same analysis applies to distributive universals in Japanese.

3. Distributive universals in Japanese

Japanese distributive universals consist of a Q-marker -mo and a pronominal such as dare ‘who,’ nani ‘what,’ dore ‘which,’ dono N ‘which N,’ itsu ‘when.’ The latter pronominal is called “indeterminate pronoun” (Ipro) by Kuroda (1965).

3.1. Indeterminate Pronouns

The function of an indeterminate pronoun is largely determined by the context in which it occurs. For example, when an indeterminate pronominal occurs with the Q-marker -mo, it functions as a universal quantifier (16a). When it occurs with the Q-marker -ka, it functions as an existential quantifier (16b). When it occurs with the interrogative C -ka, it functions as a question word (16c):

(16) a.dare-mo-ga ki-mashi-ta