NEG lowering into quantifiers

ABSTRACT

This paper is concerned with the derivation of English sentences in which negation (NEG) surfaces in a matrix clause but is interpreted as if it were in an embedded clause. The cases we are interested in allow NEG to be interpreted as having scope over a quantified subject NP in the embedded clause or over the verbal predicate in the same embedded clause. Syntactic approaches to this phenomenon have proposed a rule of NEG-raising which proceeds upwards in multi-clause structures, very much like a garden-variety movement transformation. Pragma-semantic approaches, in contrast, appeal in general to either a combination of the excluded middle law and a pre-suppositional analysis or scalar implicatures. Here we will argue that while the syntactic treatment of NEG seems to be generally correct, a leftwards/upwards approach to NEG movement does not yield the appropriate semantic representations for the sentences under consideration; rather, we propose a syntactic rule of NEG-lowering to account for the data we examine.

KEYWORDS:

• Negation
• English
• NEG-raising
• quantifier raising
• quantifier lowering
• operator scope

Acknowledgements

I would like to express my gratitude to Susan Schmerling for her comments, corrections, and patience with my never-ending onslaught of questions. My thanks are also extensive to two anonymous reviewers whose suggestions have greatly helped improve the paper. The usual disclaimers apply.

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

¹ An anonymous reviewer has suggested to us that only the NEG-most interpretation is available (such that there is no NEG-raising interpretation available for (5)), but all of our informants (native speakers from the Southern and Midwestern US, the UK, and Canada) claim both interpretations are possible; some have even said they are ‘equally available’.

² (7) is in principle ambiguous between a de re reading and a de dicto reading (hence, also (7a) and (7b)). Consider the following tests, which disambiguate in both directions (towards de re and towards de dicto):

(i) I think a specific Japanese doesn’t like sushi/I think a Japanese, namely, Akira Toriyama, doesn’t likes sushi (de re)

(ii) I think a Japanese doesn’t like sushi, but I don’t know who it is (de dicto).

³ In this respect, Horn (1978, 153) states that “If NR [NEG-raising] is a rule of grammar, it is not surprising that it shares with other extraction rules (as well as other syntactic processes) the property of sensitivity to weak and strong island constraints.” Also, Collins and Postal (2014, 103) claim that “Classical NR [NEG-raising] is subject to island constraints.”

⁴ Note that (13a) does not ascribe to Peter a belief with respect to any person other than John; this is what makes (13b) unsuitable as a possible derivational step generated by means of NEG-raising from the embedded VP.

⁵ We assume here a version of scope theory in the sense of Keshet (2010), Keshet and Schwarz (2019), in which an expression receives a de re interpretation if it appears structurally above an intensional operator. This theory is a modification of Montague’s analysis that solves some issues with overgeneration, but retains the core idea that de re NPs are structurally higher than de dicto ones.

⁶ Note that only NEG > V readings can be said to ‘preserve meaning’ in a Katz and Postal (1964) way; this seems to be related to the fact that NEG-raising applies to a Structural Description NEG-V₂ and produces a Structural Change NEG-V₁: the process is structure preserving.

⁷ A reviewer has pointed out that the status of much and many as NEG-raisers is problematic, and that (following Löbner 2000) these elements are ‘tolerant’ operators, which exclude NEG-raising readings for independent reasons. This is consistent with our characterisation of the behaviour of many people in Section 4.

⁸ A reviewer raises the question of whether not any and not some are indeed ill-formed, in the light of examples like the following (provided by the reviewer, we do need to say that not all of our informants accept (i)):

(i) I know the pain of losing a child to gun violence. And not anyone in this room, anyone in this country, should ever be faced with that pain.

– Rep. Lucy McBath (D-GA), 9/11/2019, proposing legislation to deal with gun violence https://tinyurl.com/toltdnb

It is legitimate to ask whether the interpretation of not any in Hintikka’s sense is the one holding in (i). What not anyone means in (i) is simply nobody. If not any was interpreted compositionally, then the reading ‘not just anybody, but a specific person’ should be available in (i). We contend it is not, and that the use of not anybody here is thus not a counterexample to Hintikka’s approach. If we accept Baker’s claim that the logical negation of a some-sentence is an any-sentence, then not anyone should be interpreted as not (not someone), contrary to fact. But the question is deeper: in not any N, is not a sister of any or of the NP any N? Note that only in the former case would Hintikka’s claim be incorrect. For the case of ‘free choice any’, as in Not anybody can do that (as suggested by a reviewer), it seems to be the case that NEG has scope over the whole NP, not just over any.

⁹ In this approach, “not is generated on NPs and adverbials and, in some circumstances, transformationally located into the Aux” (Lasnik 1972, 9). Lasnik argues (correctly, in our opinion) that this position is untenable, for reasons that differ from ours but which can be added to the bunch of arguments against base-generating NEG in the domain of Q.

¹⁰ The Left Branch Condition: No NP which is the leftmost constituent of a larger NP can be reordered out of this NP by a transformational rule (Ross 1967, 207).It is important to note that the LBC has been generalised. For example, as pointed out above, in the framework of Multiple Spell-Out (Uriagereka 2002), where a maximal projection in sisterhood relation to another maximal projection triggers the Spell-Out of both structures separately so that each complies with the linearization requirements imposed by the Linear Correspondence Axiom (Kayne 1994). See also Gazdar (1981) for a non-transformational approach and Müller (2008) for discussion. Here we will refer to the LBC to remain faithful to Lasnik’s argument, but it is crucial to point out that subextraction from specifiers is a general ban in transformational generative grammar.

¹¹ We adopt here the proposal made in Ladusaw:

Q_i will be in the scope of Q_j iff Q_j c-commands Q_i

The syntactic definition is, as Ladusaw notes, equivalent to the following semantic definition:

… the scope of an expression α [is] the constituent whose meaning is the argument of the meaning of α in the interpretation of the formula [where α appears] (Ladusaw 1980, 37)

In a structural configuration where NEG is adjoined to α, α will be in the scope of NEG.

¹² Sauerland and Elbourne (2002) follow Aoun and Benmamoun (1998) in arguing that in the relevant cases, where total reconstruction ensues, leftwards movement must be purely phonological (i.e., not syntactic). Movement at LF or at the overt component, syntax before Spell-Out, cannot thus totally reconstruct.

¹³ Possibly, our argument for operator lowering can be applied to the problem of tense and aspect in epistemic modals in Spanish, which also seem to require a lowering rule so that they affect the lexical V and not the modal, as in (i) and (ii); in both cases we are talking about a possibility in the present about a past event of raining (despite the fact that haber + participle should receive an anteriority interpretation; Luis García Fernández, p.c.):

(i)   Ha       podido llover

    pfv.prs.3sg  mod:ptcp  rain:inf

(ii)  Puede          haber   llovido

    mod.pst.ipfv.3sg  pfv:inf  rain:ptcp

    ‘It could have rained.’

or structures like raised passives as analysed in Krivochen and Bravo (2019):

(iii)  El  palacio  fue        terminado de construir en el  siglo   XVII

    the palace  be.pst.pfv.3sg finished  of build:inf in the century XVII

    ‘The palace was finished being built in the 17^th century.’

Note that in (iii) the passive auxiliary ser appears before the phasal aspectual auxiliary terminar de, and it is this element that manifests the participial morphology required by the passive. However, the diathesis-changing contribution of the passive does not affect the auxiliary (which does not denote an event and has no argument structure), but rather the lexical verb construir. Also, Krivochen (2019) proposes a rule of SELF-lowering that applies to intensive SELF forms and generates sentences like (v) from (iv) (note that if who moved leftwards out of a configuration [_XP who himself], then a Left Branch Condition violation would ensue and the resulting sentence should be ungrammatical, contrary to fact):

(i) Who_i himself_i executed Bill_j?

(ii) Who_i executed Bill_j himself_i?

The analysis presented in that paper provides further evidence for the existence of lowering rules in the grammar; if that is the case, then NL is not a syntactic hapax, but rather a member of a class of rules that are independently motivated.

¹⁴ It is also relevant to note that e.g. Martins (2014) proposes that metalinguistic negation merges in Spec-CP, thus effectively making it a higher operator.

¹⁵ This position contrasts strongly with that expressed by many a Minimalist, including Hornstein (1999, 45):

[S]emantic structure is a by-product of grammatical operations driven by formal concerns. Grammars seek morphological rectitude, not meaning. What meaning there is, is the unintended consequence of this mundane quest.

‘Formal concerns’, in our view, are vacuous unless put at the service of structuring meaning. Otherwise, saying ‘there is Movement, thus, there is NEG-raising’ would be all there is to say; the different readings that obtain with different QPs would be dismissible, rhetorically, as ‘unintended consequences’ or ‘by-products’.

¹⁶ This case, where NEG has scope over the root (i.e., ‘it is not the case that S’), corresponds to what an anonymous reviewer has called ‘higher-neg readings of main clause negation’, where ‘no neg-raised or neg-lowered reading is available’. Such a reading is not always available, however. The specification of the contexts in which it does or does not arise is beyond the scope of this paper.

¹⁷ Some of our informants report that, against Epstein (1976) and an anonymous reviewer, several also falls into this category. In that case, to (46) and (47) we should add (i):

(i) I don’t think several linguists read Montague

  a. I think NEG several linguists read Montague (few linguists did)

  b. I think several linguists NEG read Montague (several linguists did not read Montague; de re)

¹⁸ It would be possible (and there is nothing in the current meta-theory that would prevent us from doing it) to assign arbitrary diacritic features to each specific quantifier, or leave the paradigm to the hands of ‘nanoparameters’ (e.g., Biberauer and Roberts 2015, 9), which are associated to specific individual lexical items. Neither option seems satisfactory from a descriptive or explanatory point of view, though. We thus prefer to leave the question open.

Additional information

Notes on contributors

Diego Krivochen

Diego Krivochen obtained his PhD at the University of Reading in 2018. His work focuses on formal language theory and its relevance for grammatical description, theoretical syntax and the meta-theory of syntax, Spanish grammar, and English grammar.