Data

2.1 Patterns of Intervention

In this section, I present new data showing that intervention effects are sensitive to the types of quantifiers. To start with, when c-commanded by monotone decreasing quantificational DPs, the A-not-A form induces intervention. In (3), we already see that no one induces intervention in an A-not-A question. (4) demonstrates that other monotone decreasing quantificational DPs also induce intervention.⁴

(4) *{Henshao ren/ budao wu-ge ren} qu-bu-qu?

{few person/ less than five-CLF person} go-NEG-go “For {few people/ less than five people}, do they go or not?”

Non-decreasing monotone DPs display some gradability in terms of intervention;

simplex quantificational DPs that do not bear a numeral determiner, such as duoshu ren

‘most people’, are the most acceptable when preceding the A-not-A form, as shown in (5).

(5) Daduoshu ren qu-bu-qu?

most person go-NEG-go “For most people, do they go or not?”

Meanwhile, judgments are more degraded when a monotone increasing DP with a modified numeral determiner c-commands the A-not-A form. Similar decrease in acceptability(decrease in acceptability) is witnessed in the presence of a non-monotonic bare numeral DP. This is illustrated by (6).

(6) (a) ??{Zhishao wu-ge ren/ chaoguo wu-ge ren} qu-bu-qu?

{at least five-CLF person/ more.than five-CLF person go-NEG-go

“For {at least five people/more than five people}, do they go or not?”

(b) ??Wu-ge ren qu-bu-qu?

five-CLF person go-NEG-go

“For five people, do they go or not?”

4 A monotone increasing quantificational determiner, such as most, is ‘monotone increasing’

because when the predicate in the body of the quantified expression is made less restrictive, the truth value is preserved (Westerstahl 2015). Thus, Most men work hard entails Most men work.

Alternatively, this is called ‘right upward monotone’ in the literature.

By contrast, for monotone decreasing quantifiers, when the predicate in the body of the quantified expression is made less restrictive, the truth value is not necessarily preserved. Quite the opposite, it is preserved when the body is made more restrictive: Few men work entails Few men work hard.

Finally, for the same set of monotone increasing and non-monotonic quantifiers with numeral components, intervention effects may be ameliorated under embedded contexts.

By contrast, no amelioration is witnessed for monotone decreasing quantifiers:

In sum, intervention effects in A-not-A questions exhibit a complex overall pattern as follows:

(9) (a) Monotone decreasing quantifiers induce intervention effects in both matrix and embedded contexts;

(b) Monotone increasing, non-numeral quantifiers don't induce intervention effects in all environments;

(c) (Monotone increasing) modified numerals and (non-monotonic) bare numerals induce weak intervention in matrix A-not-A questions, which is ameliorated under embedded contexts.

2.2 Previous Accounts

In his classic account, Huang proposes (Huang 1991) that, in an A-not-A question, a null interrogative operator (termed NQ by Huang) initially merges with the regular predicative element in surface syntax and forms a phrasal unit. After the initial merge, NQ undergoes covert movement at LF to check off the interrogative feature at [Spec, CP].

When the LF component feeds into the semantic representation module, NQ is interpreted as taking the predicative element as its argument and returns a disjunctive set out of it. The LF structure and the semantics of an A-not-A question are represented as follows (Huang et al. 2009):⁵

5 Furthermore, Huang assumes that the level of surface syntax also feeds input to the PF component, and he proposes that the reduplicated form of the A-not-A part is a matter of Spellout.

That is, the instruction PF receives requires that the [Predicate + NQ] chunk to be spelled out phonologically as a positive and a negative predicate copy. As a consequence, this account claims that both the reduplication and the negative morpheme are inserted as something completely arbitrary, and does not bear on the interpretation process.

(7) Wo yijing zhidao {zhishao wu-ge ren/ chaoguo wu-ge I already know {at.least five person/ more.than five ren/ wu-ge ren} qu-bu-qu.

person/ five-CLF person go-NEG-go

“I already knew whether {at least five people/more than five people will go or not.”

(8) *Wo yijing zhidao {meiyou ren/ henshao ren/ budao wu-ge I already know {no person/ few person/ less.than five-CLF ren/ zuiduo wu-ge ren} qu-bu-qu.

person/ at.most five-CLF person} go-NEG-go

“I already knew whether {nobody/ few people/ less than five people/

at most five people} will go or not.”

Wide-Scope Indefinites and Topicality 99

(10) (a) Ni qu-bu-qu?

you go-NEG-go

(b) LF: [_CPNQi[+A-not-A] [_IP Ni t_i[+A-not-A] qu]]

(c) Semantics: “For x, x ∈{go, not go}, you x?” = {you go, you not go}

Intervention effects can be accounted for if relativized minimality is incorporated into Huang’s theory of covert movement. For example, Yang (2011) assumes a recent formulation of the relativized minimality framework (Starke 2001; Rizzi 2004), in which only features undergo movement and movement is subject to the following constraint:

(11) MAXIMAL MATCHING FILTER (Yang2011,63)

Let X and Y be bundles of features in a sequence of […X…Y…]; Y cannot cross X when Y is maximally matched by X.

In other words, if another scopal element is closer to the landing site than the interrogative operator and bears at least as much featural information as the interrogative operator, that scopal element would be an intervener. A quantifier is assumed to carry a superset of the set of features needed to trigger covert movement of the interrogative operator (Rizzi 2004). As such, covert movement is blocked when a quantifier c-commands the interrogative phrase, since the quantifier would then be closer to the landing site.

The above theories face both empirical and conceptual problems. Empirically, the formulation of relativized minimality in terms of feature matching fails to predict how a fine-grained distinction within quantifier types would make any differences during intervention. If quantifiers in general possess enough features to maximally match the interrogative operator, then by including monotonicity as a further dimension in the feature geometry, we only increase the inventory of the feature set for the quantifiers.

Therefore, both monotonic increasing and decreasing quantifiers are supposed to maximally match the interrogative operator and block its covert movement. Furthermore, it is rather unclear whether we should bring monotonicity, a semantic primitive, into our feature geometry, especially since we find no independent evidence that monotonicity plays a role in creating intervention for environments other than A-not-A questions.

Furthermore, assuming that NQ initially merges in situ and then moves to take scope, we would expect that, in embedded questions, NQ still moves to take scope at the embedded clause’s [Spec, CP]. As such, it will have to cross the quantificational intervener along the way, therefore falsely predicting that the intervention in embedded

contexts does not differ from that in matrix contexts. In other words, even assuming that quantifier types can be fine-tuned to accommodate the monotonicity issue, it is unclear how a relativized minimality account handles the selective amelioration phenomenon (as in 7–8) in a principled manner.

Conceptually, the current version of NQ movement does not derive the right interpretation for A-not-A questions. According to Huang’s analysis in (10), NQ in (12) ranges over two opposing predicates. As NQ moves to take scope, an operator-variable pair is formed where NQ associates with its gap position, yielding the semantics in (12b):

(12) (a) Daduoshu ren hui-bu-hui qu?

most people will-NEG-will go

(b) “x∈{hui qu, bu-hui qu}, daduoshu ren x? (= “x∈{will go, will not go}, most people x?)”

This in turn derives two alternative propositions:

(13) {p₁= Most people will go, p₂= Most people won’t go.}

In p₁, the positive predicate is within the nuclear scope of the quantifier most. In p₂, the negative predicate is also within most’s nuclear scope, which means negation scopes under most. The problem with this interpretation is that the two alternative propositions are not exhaustively carving up the logical space. In addition to the situations expressed by p₁ and p₂, there is a third situation that belongs to neither of the two, represented by p₃:

(14) p₃ = Neither most people will go, nor most people won’t go.

For example, imagine a situation where exactly half of the people will go. This situation instantiates p₃, and does not instantiate p₁ or p₂. This result is undesirable, because various studies have shown that speakers use A-not-A questions when presented with two alternatives that exhaustively carve up the logical space (McCawley 1994; Wu 1997;

Schaffar and Chen 2001). A natural way to address this problem is to change p₂ by making sentential negation scope above most, yielding the proposition:

“It is not the case that most people will go.” This guarantees exhaustivity. However, the semantics still does not quite fit with our intuition of what this quantified A-not-A question is about. This problem becomes evident when we consider the meaning of a negative answer to the question in (12a), given below:

(15) A: Daduoshu ren hui-bu-hui qu?

“The majority of people, will they go or not?”

B: Bu. ‘No.’

= “The majority of people, as a group, they will not go.”

≠ “It is not the case that most people will go.”

That is, in the negative answer, the quantifier does not fall within the scope of sentential negation. In fact, the question is interpreted in a context where there exists a plurality of

Wide-Scope Indefinites and Topicality 101

individuals. They constitute the majority of all the contextually relevant individuals, and they either will collectively go as a group, or will collectively not go.

In document Proceedings of the 5th Central European Conference in Linguistics for Postgraduate Students organized by Department of English and American Studies (Pldal 98-102)