Mark Newson a and Márton Kucsera b
4. A Different Approach 1 Outline of a Solution
The problems facing the WH-assumption are wiped out if we assume that the quantified multiple wh-structures do not actually contain multiple wh-phrases. Instead, we claim
MARK NEWSON AND MÁRTON KUCSERA
that those wh-phrases with quantifier interpretations are actually universal quantifiers which are realized as wh-phrases. We call this the Q-assumption. From this perspective there are two kinds of phrases with surface “wh” realizations: real wh-phrases which realize underlying wh-operators (represented below as WH) and those wh-phrases which realize underlying universal quantifiers (QWH). This contrasts with the WH-assumption, in which there is one type of wh-phrase, though sometimes this may be interpreted as a quantifier (WH∀).
To demonstrate the advantages of the Q-assumption, let us consider the distribu-tion of quantifiers and wh-elements under the two assumpdistribu-tions, side by side. First, let us consider the distribution of wh-phrases, as represented in (10) and (11):
(10) Ki látott mit?
who saw what
WH-assumption WH/*WH∀ V WH/*WH∀
Q-assumption WH V WH
(11) Ki mit látott?
who what saw
WH-assumption WH∀/*WH WH/*WH∀ V
Q-assumption QWH WH V
From the perspective of the WH-assumption, the distribution of wh-phrases is rather complex. At least one wh-phrase must immediately precede the verb; others may precede that wh-phrase or follow the verb. Of course, there are conditions on which of these get interrogative or quantifier interpretations and these add to the complexity of observations. Under the Q-assumption, once the quantifiers are factored out, we see that one wh-phrase must precede the verb and all others follow. Not only is this a comparatively simple distribution, but it is a pattern found in many of the world’s languages.
Now consider the distribution of universal quantifiers under the two assumptions, as represented in (12) to (14):
(12) (a) Mindenkit János hívott meg
everyone-ACC John-NOM invited-3.S PERF
(b) János hívott meg mindenkit
“It was John who invited everyone.”
MULTIPLE WH-STRUCTURES IN HUNGARIAN: A LATE INSERTION APPROACH
(13) Mit látott mindenki
what saw everyone
WH-assumption WH V Q∀
Q-assumption WH V Q∀
“What did everyone see?”
(14) Ki/*mindenki mit látott?
who/everyone what saw
WH-assumption WH∀/*Q WH V
Q-assumption QWH/*Q WH V
“For everyone, what did they see?”
First of all, as (12) shows, universal quantifiers can generally appear before or after the verb. When they precede the verb, they also precede the focus which sits in the imme-diate preverbal position (12a). As these sentences contain no wh-phrase (or quantifier realized as such) there is no distinction between their treatment under either the WH- or the Q-assumption.
From the WH-assumption perspective, as shown in (13) and (14), when quantifiers appear in interrogative clauses, they must follow the verb, even though they can precede other kinds of foci. Clearly this is a complication that requires an explanation. Both É. Kiss (1993) and Lipták (2000) assume that this follows from the fact that a wh-phrase preceding a wh-phrase is interpreted as a universal quantifier and this somehow blocks the appearance of the quantifier, which would yield a structure with the same interpretation.
However, as we have seen, the interpretive rule they assume has very little explanatory power and likewise it provides a poor basis for explaining why quantifiers cannot precede wh-phrases in focus position.
Once more, the Q-assumption provides a more straightforward picture. Quantifiers can appear before foci and postverbally in both interrogative and declarative contexts.
The only issue to be addressed is that when they come before a wh-phrase they are realized as wh-phrases. Accounting for this turns out to be a much easier task than any account adopting the WH-assumption, as we will demonstrate in the following sections.
4.2 Late Insertion
One of our criticisms of the WH-assumption is the lack of a theory that allows one lexical item to be reinterpreted as another. It is another advantage of the Q-assumption that there already exists a theory which allows for the situation in which an element is realized differently in different contexts.
A number of current frameworks have adopted a late (vocabulary/lexical) insertion approach, such as Distributed Morphology (Halle and Marantz 1993) and Nanosyntax (Starke 2009). This moves away from the lexicalist tradition in assuming that morphemes
MARK NEWSON AND MÁRTON KUCSERA
do not come preformed, with all properties intact, ready to be manipulated by syntactic operations. Instead such frameworks assume that the syntactic system operates on abstract sub-morphemic elements, which lack phonological properties, and builds them into larger constructs. Only once the syntax is finished are these syntactically constructed morphemes realized by associated exponents.
The main advantage of this approach is that it is not constrained by the idea that exponents and underlying syntactic/semantic elements are in a fixed relationship and therefore a given exponent can be used to realize a number of morphemes which differ in their sub-morphemic composition. This idea has proved useful in accounting for various phenomena, such as syncretism, and in simplifying the description of morphological distributions. The theory is also compatible with the idea that the same underlying construct can be realized by different exponents in different contexts, i.e., the idea behind the Q-assumption.
Typically late insertion approaches work with a simplified vocabulary/lexicon and the more limited lexical resources are made to work harder with exponents competing against each other for selection. The selected exponent is the “best fitting” one for any instance of morpheme realization. This relies on the assumption that exponents do not have to be associated with exactly the set of features that they are used to spell out. For example, under some assumptions as to what counts as the “best fit,” it may be that a morpheme constructed of sub-morphemic elements [a, b] is realized by an exponent X that is associated with [a] in its lexical specification, if there is no better fitting expo-nent. Thus, the selected exponent does not have to be associated with exactly the set of features that it is used to spell out.
Turning to the case in hand, the situation must be as follows. The Hungarian syntactic system constructs a universal quantifier out of a set of sub-morphemic elements (e.g., [Op+∀+non-human] “everything,” [Op+∀+human] “everyone,”
[Op+∀+place] “everywhere,” etc.).6 It also constructs a set of interrogative morphemes (e.g., [Op+WH+non-human] “what,” [Op+WH+human] “who,” [Op+WH+place]
“where,” etc.). Each of these has an associated exponent which realizes them under normal circumstances:
(15) (a) [Op+∀+non-human] ↔ minden [Op+∀+human] ↔ mindenki [Op+∀+place] ↔ mindenhol
6 It is not our concern in this paper to give the details of exactly how these morphemes are constructed. Our proposals are compatible with a number of different frameworks and so it is not important to select any particular one here.
MULTIPLE WH-STRUCTURES IN HUNGARIAN: A LATE INSERTION APPROACH
(b) [Op+WH+non-human] ↔ mit [Op+WH+ human] ↔ ki [Op+WH+place] ↔ hol
However, for some reason to be identified, the exponents in (15a) are not the best spell out possibility for underlying universal quantifiers when they precede wh-phrases and those in (15b) turn out to be better. All that remains is to identify the reasons behind this.
4.3 Clausal Typing
Our proposal will be based on the notion of clausal typing, introduced by Cheng (1991).
Clausal typing is the overt marking of a clause’s interrogative status in one of a number of ways.
Generally Cheng considers there to be two types of languages with respect to clausal typing: those that mark interrogative clauses with special particles and those which do not. The latter utilize wh-fronting instead. The core of Cheng’s theory is the Clausal Typing Hypothesis:
(16) Every clause needs to be typed.
For now, all that needs to be noted is that Hungarian is the kind of language which types wh-interrogative clauses via wh-fronting.7 We have already noted that in Hungarian inter-rogative clauses one and only one true wh-phrase is fronted to a position immediately before the verb. This can be seen as a direct result of the application of (16).
However, another very well known aspect of the syntax of Hungarian is the fact that scope relations are overtly marked by placing operators in front of those they scope over. We claim that these two properties, clausal typing by wh-fronting and the leftmost condition on wide scope operators (henceforth LWO), are in direct conflict with each other and that it is precisely this conflict which gives rise to the realization of underlying universal quantifiers as wh-phrases. Before we can build on these claims, we will need to further investigate the properties of clausal typing.
4.4 The Semantic Basis of the Typing Requirement
One way to actualize the claim that there is a conflict between the Clausal Typing Hypothesis and the LWO would be through the idea that both are left edge conditions.
This seems quite natural given that fronting obviously involves the left periphery. Thus the two conditions require different elements to be leftmost and the conflict arises in an interrogative clause with a wide scope non-interrogative operator: the typing condition 7 Hungarian uses the particle strategy for yes-no questions, though we will have nothing to say about this in the paper.
MARK NEWSON AND MÁRTON KUCSERA
would require the wh-operator to be leftmost and the LWO would require the non-interrogative to be leftmost.
This seems to be borne out by the data we have reviewed so far, as a narrow scope quantifier coming to the right of a clausal typing wh-phrase is perfectly grammatical ([17a] with interpretation [i]), but a wide scope quantifier coming either to the left (17b) or the right ([17a] with interpretation [ii]) of a typing wh-phrase is ungrammatical:
(17) (a) Mit látott mindenki?
what-ACC saw everyone-NOM
(i) Qy ∀x [x saw y]
(ii) * ∀x Qy [x saw y]
(b) * Mindenki mit látott?
These observations might suggest that this conflict is located entirely in the syntax and has to do with a competition for the relevant leftmost position.
However, there are at least two reasons to believe that this is not the best charac-terization of the situation. The first is that it isn’t true that universal quantifiers never precede wh-phrases. This is possible if the quantifier is interpreted as a contrastive topic:
(18) /Mindenki \mit látott?
everyone-NOM what-ACC saw
“What did /everyone \see?”
A question such as (18) can be understood in the following context. Suppose a small group of tourists visit an art gallery but decide to explore it separately, each member going to see the pieces that they are personally interested in. The question asks for the identity of the artwork that all members of the group ended up seeing, in contrast to those which only subsets of the group may have seen. Note that the question has a special intona-tion pattern, indicated by the rise on “Mindenki” and the fall on “mit.” We see the same rise-fall pattern in the English translation too and it is a well known phenomenon in a number of other languages (for example, see Gyuris [2009] and references cited therein).
The semantic effect of this intonation pattern seems universally to force a narrow scope interpretation on the quantifier. Hence, unusually for Hungarian, (18) has an inverse scope interpretation, equivalent to that in (17a), interpretation (i).
We might try to salvage the left edge character of clausal typing by defining the domain that the typing wh-element must precede as being smaller than the domain that contains the contrastive topic. However, this would not address the second reason to doubt the syntactic characterization of the conflict between clausal typing and scope marking.
As the following data show, it is not only the leftmost universal quantifier preceding a
MULTIPLE WH-STRUCTURES IN HUNGARIAN: A LATE INSERTION APPROACH
wh-phrase that must be realized as a wh-phrase, but all non-contrastive topic universal quantifiers that precede a wh-phrase:
(19) (a) Ki kinek mit adott?
who-NOM. who-DAT what-ACC gave
“Who gave what to who?”
∀x ∀y [what did x give to y]
(b) Mit adott mindenki mindenki-nek
what-ACC gave everyone-NOM everyone-DAT Qy [everyone gave y to everyone]
(c) Ki mit adott mindenki-nek
who-NOM what-ACC gave everyone-DAT
∀x [what did x give to everyone]
In (19a) there are two universal quantifiers realized as wh-phrases (ki and kinek) and only one actual wh-phrase (mit). Both of the quantifiers precede the wh-phrase, but obviously only one of them is at the left edge of whatever domain is relevant for clausal typing. If clausal typing were simply a left edge phenomenon we would expect only the leftmost quantifier to be realized as a wh-phrase and the second one to be realized as a quantifier. This, however, is ungrammatical:
(20) *Ki mindenki-nek mit adott?
who-NOM everyone-DAT what-ACC gave
The rest of the data in (19) suggest a different account as to what conditions lead to a universal quantifier being realized as a wh-phrase. In (19b) both quantifiers follow the verb, and subsequently the wh-phrase. True to the nature of Hungarian, the scope inter-pretation of the quantifiers is narrow with respect to the wh-phrase. In (19c) however, one quantifier precedes the wh-phrase and one follows. The preceding one is realized as a wh-phrase and the following as a quantifier. The preceding quantifier has a wide scope interpretation with respect to both the wh-phrase and the other quantifier.
Thus, quantifiers that are realized as wh-phrases not only precede wh-phrases, but have wide scope interpretations with respect to them as well. In fact, associating clausal typing with a wide scope interpretation, rather than a leftmost position, handles the data more straightforwardly. Note that in (19a), where both quantifiers are real-ized as wh-phrases, while only one is at the left edge, both have scope over the real wh-phrase. In this situation, the quantifiers’ scopes do not interact. Consequently, both are equally interpreted as having wide scope. It is only in (19c) that the two quantifiers’
MARK NEWSON AND MÁRTON KUCSERA
scopes interact, and in this the one with wide scope is realized as a wh-phrase and the one with narrow scope is not. We therefore propose the following version of clausal typing theory:
(21) The Clausal Typing Hypothesis Every clause needs to be typed.
(22) Clausal typing comes in two forms:
(i) typing with particles
(ii) typing by realizing wide scope operators as wh-phrases.
Obviously, for a language making use of strategy (22ii), such as Hungarian, when the wide scope operator is a wh-phrase, its realization as such is straightforward. It is only cases where interrogative clauses have wide scope non-wh-operators that we get the special realization of this operator as a wh-phrase.
In the last section of this paper we will discuss how clausal typing interacts with the process of vocabulary selection.
5. Vocabulary Selection