Mayfest 2018: NayFest

9:00-9:20    —Breakfast (provided)—
9:20-9:30    —Welcoming Remarks—
9:30-10:40    Emmanuel Chemla (Ecole Normale Supérieure - Laboratoire de Sciences Cognitives et Psycholinguistique) - Cognitive constraints & high and low level semantics
10:40-11:00    —Coffee Break—
11:00-12:10    Laurence R. Horn (Yale University) - Negation and word order in the footsteps of Neg-first
12:10-1:30    —Lunch—
1:30-2:40    Jon Gajewski (University of Connecticut) - It's not syntax, I don't think: Neg-raising and parentheticals
2:40-2:50    —Break—
2:50-4:00    Ivy Sichel (University of California, Santa Cruz) - Neg-words in fragments
4:00-4:20    —Coffee Break—
4:20-5:30    Elena Herburger (Georgetown University) - NPI-hood: What is it?
6:00-late    —Party with Dinner (provided)—
Language Science Center (2114/2118 H. J. Patterson Hall)

9:00-9:30    —Breakfast (provided)—
9:30-10:40    Hedde Zeijlstra (University of Göttingen) - Mirroring strong and weak NPIs and PPIs
via Skype
10:40-11:00    —Coffee Break—
11:00-12:10    Theresa Biberauer (University of Cambridge) - Marked negation: Core insights from the periphery
12:10-1:30    —Lunch—
1:30-2:40    Chung-hye Han (Simon Fraser University) - Using negation as a probe for structure
2:40-3:00    —Coffee Break—
3:00-4:10    Ming Xiang (University of Chicago) - On the processing complexity of negation
4:15    —Closing Remarks—

Much of what generative inquiry has shown us about the distribution and formal make-up of negative elements, how they are acquired, and how they may vary and change over time centres on declarative clausal negation. The purpose of this talk is to argue for the importance of developing a more fine-grained understanding of non-declarative and, more generally, marked negation-containing structures. Much - though not all - of the exemplification will be drawn from Afrikaans, a language with a highly distinctive negation system which, I hope to show, offers not just the language-acquiring child, but also the linguist unique insights into the various ways in which marked negation contributes to core syntax.

I will focus on three distinct marked negation types:

(a) negative imperatives
(b) negative exclamatives
(c) emphatic negation

Negative imperatives will be our starting point. The argument here will be that, in the current impoverished UG era (Chomsky 2005), these structures need to be understood, alongside their positive counterparts, as a key component of the Primary Linguistic Data (PLD). UG itself can no longer steer the acquirer to specific, grammar-relevant input in the manner assumed in the classic Principles & Parameters framework (Chomsky 1981), raising the question of how the generally agreed-upon need for appropriately ordered grammar building (see Fodor & Sakas 2017 for an overview) can be modelled in a UG-poor Three Factors model. In the context of Biberauer’s (2017) Maximise Minimal Means/MMM model, a partial answer suggests itself, with positive and negative imperatives emerging as particularly important. What I will demonstrate here, on the basis of a comparison between spoken and dialectal Afrikaans, Dutch, and German, is that a wide range of these language’s core syntactic properties are clearly signalled in imperative, and, particularly, negative imperative structures. The idea that this negative-imperative “signalling” role really seems to matter will be further reinforced by consideration of to date only minimally discussed negative imperative-related contact phenomena in Namibian German, a variety of German that has been in contact with Afrikaans over roughly 100 years now (Biberauer, Bockmühl & Shah 2017).

The remainder of the talk will focus on negative structures which facilitate insight into the way in which the elements making up the core/unmarked negation system are harnessed for “extended” (non-core) functions, just as one would expect on an MMM perspective. Of particular interest here will be, on the one hand, “expletive” uses of negation such as those seen in negative exclamatives, and, on the other, the devices that modern spoken Afrikaans employs to signal emphatic negation. In the former case, we will again be able to compare negative exclamatives with their positive counterparts, which, at first sight, seem to be in a free variation (“true optionality”) relationship with each other; closer investigation, however, reveals this not to be so. Both negative exclamatives and emphatic negation structures will further allow us to see patterns that are not unique to Afrikaans, but that recur crosslinguistically, calling for a unified explanation. Appealing to by now well established generative insight into how formal structure is “recycled” (e.g. in the context of grammaticalisation; Roberts & Roussou 2003, van Gelderen 2003, Roberts 2010), and also to emerging insights from the much younger generative study of the formalisation of speaker-perspective and of phasal make-up, I will conclude by suggesting that (b)- and (c)-type negative “extensions” reflect the more general mechanism via which core grammar is recycled for expressive purposes.

Content words — e.g., nouns and adjectives — are generally connected: there are no gaps in their denotations; no noun means ‘table or shoe’ or ‘animal or house’. We explore a formulation of connectedness which is applicable to content and logical words alike. Interestingly, the classic notion of monotonicity of a quantifier Q is found to be equivalent to Q and its negation being connected. On a first inspection, logical words satisfy this generalized version of the connectedness property at least as well as content words do — that is, both in terms of lexicons and in terms of acquisition biases in artificial settings.

English allows a construction in which a sentence contains a parenthetical with a clausal gap, as in (i). I will refer to phrases such asI think in (i) as clausal parentheticals. Typically, clausal parentheticals cannot be negative, cf. (ii).

(i) There is beer in the fridge, I think.
(ii) *There is beer in the fridge, I don't think.
It has been noted that when the clausal parenthetical contains a neg-raising predicate, an apparent doubling of a negation in the main clause is allowed, as in (ii).

(ii) There is no beer in the fridge, I (don't) think.

This doubling has been taken to be an argument in favor of syntactic approaches to neg-raising, as in Ross (1973) and Collins & Postal (2014). I will defend an analysis of the doubling in (ii) that is compatible with a semantic/pragmatic approach to neg-raising, as in Horn 1989, Gajewski 2007, Romoli 2013.

In this talk, I present how negation scope interacting with object and subject quantifiers can inform us as to the position of the verb in the clause structure of Korean and Japanese, which are head-final. Data from several Truth-Value Judgement Task (TVJT) studies with both adult and child participants show that while wide scope reading of negation is uniformly unavailable to all speakers with respect to the subject, it is unavailable to only half the speakers with respect to the object. I use this experimental data to support the proposal that there is a split in the speech community: one group that places the verb high in the clause structure and another that places the verb low. This variation within the speech community is argued to be the consequence of a paucity of evidence from the input to the learners of Korean and Japanese to support an analysis of verb-raising. I will then discuss negative null object constructions in Korean and Japanese as a potential domain that can diagnose the position of the verb. In two TVJT studies, I found that while none of the Korean participants could recover an adjunct from an antecedent clause in sentences with null objects, half of the Japanese participants were able to recover an antecedent adjunct. This inter-speaker variation detected in the Japanese null object data is consistent with the proposal that verb-raising is available to some speakers, but not all. The speakers with verb-raising can elide the entire VP resulting in null object, in which case an adjunct from the antecedent clause can be recovered. However, speakers without verb-raising must elide the object DP to obtain null object, and thus antecedent adjunct cannot be recovered.

Why should there be expressions (NPIs) whose distribution is limited to downward entailing environments of various sorts? One intriguing kind of answer that has been explored is that this follows from their meaning (e.g. Kadman and Landman 1993, Lahiri 1998, Krifka 1995, Chierchia 2004). Building on Herburger and Mauck (2009, 2013), I argue that while these kinds of analyses offer many important insights, they may ultimately, and for a number of different reasons, not fully answer the question. In the end, it may in fact be more adequate (and more economical) to say that an NPI is something that carries a feature [+NPI]. I discuss that [+NPI] cannot amount quite the same as Klima’s [+ affective], for the contrast between every and no discussed in Ladusaw (1979). The syntactic licensing that is involved requires a certain amount of decomposition and the notion of Local Polarity. The analysis I endorse relies heavily on previous work by Sánchez Valencia (1991), Dowty (1994) and Ludlow (2002).

Regarding the question why so many NPIs have an existential-like semantics—a curious fact if NPI-hood is but a feature as I am suggesting—I suggest that a certain kind of semantics pragmatically predisposes certain expressions to becoming NPIs without it fully determining it. I discuss how this is true of low-scalar NPIs, which allow for emphasis, but also of high scalar NPIs, which allow for understatement and equivocation (cf. Israel 2011).

If NPI-hood is but a feature we can explain why of various synonyous expressions one can have the distribution of an NPI without that being also the case for the other(s) (e.g. some vs. a vs. any; need + bare infinitive vs. need + to infinitive). NPI-hood being nothing ‘deep’, we also predict that it should not necessarily remain stable over time. In this context I discuss how NPI features may come into existence historically, how they can strengthen from being ‘weak’ to becoming ‘strong’ (e.g. Martins 2000, Jäger 2010, Eckardt 2006) and how they can be in some contexts be reinterpreted as having semantic contexts [+NEG]. The coexistence of [+NPI] and [+NEG] on two homophones (in virtually complementary distribution) is argued to be the source of Negative Concord (Herburger 2001). When [+NPI] is lost, leaving the [+NEG] option as the only one, we obtain a genuinely negative paradigm (e.g. standard English nobody, nowhere, etc.) Finally, special attention is given to the little discussed fact that [+NPI] features can simply be lost, at which point the expression, without changing its meaning, loses its distributional restriction to downward entailing contexts. Relevant examples include Dutch ooit ‘ever’ > ‘sometime’ (Hoeksema 1998) and German einig ‘any’ > ‘some’ (Jäger 2010).

The most influential monograph on negation in natural language, Otto Jespersen’s Negation in English and other languages, has just celebrated its hundredth anniversary in print. I will focus on a conspiracy Jespersen identifies at the beginning of this work that I have dubbed NEG-FIRST (Horn 1989): the functionally motivated tendency for negation to precede its focus and to occur as early in the sentence as possible. While the original evidence for this conspiracy came from the diachronic shifts in the expression of sentential negation now known as “Jespersen’s cycle” (Dahl 1979), recent work since Kiparsky & Condoravdi 2006 has challenged the role of Neg-first as a prime motivator of the cycle. It is thus worth exploring the role Neg-first plays in a range of other phenomena affecting the formal distribution and scope of negative elements. Among the constructions to be examined in this light are:

• Neg-Raising: I don’t want to miss Mayfest [≈ I want not to miss Mayfest ]
• V2 Negative Inversion: Never have I been to such an exciting conference.
• V1 Negative Inversion (dialectal U.S. Eng.): %Can’t {anybody/nobody} tell me what to do.
• Pleonastic Negative Parentheticals: They shouldn’t, I don’t think, tell you what to do.
• Asymmetries in the distribution of negative concord in pre- vs. post-verbal indefinites
• Asymmetries in the distribution of prefixal vs. suffixal negation

While the preference for the early positioning of negation is a powerful tendency, it is still defeasible given competing universal and parochial linguistic constraints. Of particular note in this connection is the interaction of Neg-first with two other conspiracies prefigured by Jespersen (1917): MAXCONTRARY (the tendency for formally contradictory negation to strengthen to contrary readings) and NEXAL NOT (the preference for expressing overt negation in canonical, syntactically unmarked positions).

Neg-words in Neg-concord (NC) languages may constitute a fragment answer to a positive question, in (1b) (Watanabe 2004), similar to negative indefinites in non-NC systems, such as English (1c). This poses a challenge for theories of NC, and especially for those theories in which do not view neg-words as inherently negative (e.g., Ladusaw 1992, Brown 1999, Giannakidou 2000, Weiss 2002, Zeijlstra 2004, Penka 2010): where does the negativity of the fragment reside and how is the neg-word licensed? Theories of fragment answers which involve neg-word fronting combined with TP-sluicing, as in (2), may be on the right track, but because the antecedent is positive they do not provide an immediate solution to this problem. In this talk I present cross-linguistic evidence from long-distance NC, in (3), to support the view that sluicing is necessarily involved, and there must be an instance of covert high negation (Schwartz & Bhatt 2006, Zeijlstra 2011, among others) above the position of the fragment. The grammaticality of LD NC, as in (3) makes it possible to compare, in LD contexts, NC fragments in Hebrew with Neg-Indef fragments in English, in (4), and here we observe a difference. Whereas a neg-indef yields ambiguity, a neg-word yields only the interpretation in which Neg scopes in the matrix clause. This suggests that the operation that licenses the N-word fragment is not unconstrained; the negation has to take wide scope; and this suggests that sluicing is involved. A high negation above the fragment will be shown to be supported by LD NC across factive CP (in the complement of surprise, regret). The first part of the talk will set the stage for the contrast in (4) by examining the surprisingly liberal locality constraints on Hebrew Neg-words. Hebrew Neg-words can be separated from the negation which licenses them by some kinds of finite clauses, such as negated verbs of saying, in (2); but when the matrix verb imposes truth or factivity on the finite CP complement, grammaticality diminishes incrementally: the CP complement of negated veridical verbs (demonstrate, prove; Anand and Hacquard 2014) allows a Neg-word only in object position, and negated verbs which select factive CPs disallow Neg-words in all embedded positions.

(1)   a. hu *(lo) ra’a af talmid / af seret             Hebrew
      he NEG saw N-student / N-film / N-student / N-film
      ‘He didn’t see any student / any film.’
    b. A: Et mi hi ra’ata?
        ACC who she saw
        ‘Who did she see?’
      B: Af exad
        Neg-word
        =hi lo ra’ata af exad
        she didn’t see Neg-word
    c. A: Who did she see?
      B: Nobody
(2)   Nobody [TP she saw nobody]
(3)   lo amarti Se-hi ra’ata af exad
    neg said.I that-she saw N-word
    ‘I didn’t say that she saw anyone.’
(4)   a. A: Who did you say she saw?
      B: Nobody
      =Nobody did I say that she saw / I didn’t say she saw anybody
      =I said that nobody did she see / I said that she didn’t see anybody.
    b. A: Et mi amart Se-hi ra’ata?
        ACC who you.said that-she saw
       B: Af exad
        N-word
        =lo amarti Se-hi ra’ata af exad
        I didn’t say she saw N-word
        ≠amarti Se-hi lo ra’ata af exad
        I said that she didn’t see anybody

Negative sentences are often reported to be more difficult to comprehend than their positive counterparts. Traditional findings from sentence-picture verification tasks have often suggested that negation is more complex because it is another stage of operation to perform in addition to the initial stage of processing an affirmative proposition (e.g. Carpenter and Just 1975; Clark and Chase 1972). Pragmatics-based accounts of negation comprehension, however, argue that negation in pragmatically felicitous if it is used to deny or contradict a positive proposition that is made salient or “expected” in some way (e.g. Wason, 1965), and when negation is “pragmatically licensed”, the processing complexity associated with them may disappear (Nieuwland and Kuperberg, 2008). In this talk, I will present two sets of results to assess the processing mechanisms for negation. First, we will look at a set of ERP studies that examined how “negativity” is processed to license NPI items. These studies show that negation, including implicit negation, can be computed very quickly to license NPIs. They provide evidence that the process of computing a negated representation is not necessarily difficult. It also challenges the view that processing negation requires a two-stage model, since it is not obvious that an affirmative proposition needs to be constructed when people process sentences containing NPIs. In the second set of studies, we focus on the effect of contextual facilitation of negation. Based on the pragmatics account, one would predict that the stronger expectation one may form, based on the context, for an affirmative proposition, the less difficult it would be to process the related negative proposition. This effect was reported in Nordmeyer and Frank (2014). However, preliminary results from a set of Mechanical Turk studies (with modified designs) did not completely replicate the main findings. This raises questions as to whether contextual expectancy of the affirmative proposition should be the most relevant factor for comprehending negation. An alternative QUD-based account will be explored.

Strong Negative Polarity Items (NPIs) are NPIs that are only fine in anti-additive contexts, whereas weak NPIs are generally licensed in all downward-entailing contexts. Similarly, strong Positive Polarity Items (PPIs) are anti-licensed in all downward-entailing contexts; weak PPIs are only bad in anti-additive contexts

According to Chierchia (2006, 2013), NPIs that denote lower scalar endpoints, such as existentials like any or ever, are fine in downward entailing contexts, since outside such contexts their semantics would give rise to a contradiction. This contradiction arises as such NPIs both obligatorily introduce domain alternatives and trigger the presence of a covert exhaustifier. As argued for in earlier work, elements with the same properties that denote the highest endpoint of a scale, such as universal quantifiers, must be PPIs.

For Chierchia (2013), following Gajewski (2011), the weak-strong NPI distinction results from whether the exhaustifier only looks at the semantics of the NPI licenser, or also at its pragmatics (both the presupposition and the implicatures). Weak NPIs trigger the covertly present exhaustifier (EXH) to look at the semantic contribution of the licenser only; strong NPIs also trigger EXH to look the licenser’s pragmatic contribution. Since the joint semantic-pragmatic contribution of all non-anti- additive licensers is no longer downward entailing (e.g., enriched ‘few N’ means ‘few but at least one N’), only anti-additive licensers can license strong NPIs.

This analysis ignores a recent observation made by Collins & Postal (2014), who observe that strong NPIs are generally strict NPIs (i.e. NPIs that must be licensed within a local syntactic domain, such as a finite clause or a syntactic island) and vice versa. Licensing such strong/strict NPIs across such locality boundaries, like adjunct islands, is not possible. Weak NPI licensing, by contrast, is not subject to such syntactic locality constraints and may apply across locality boundaries. This is a problem for the Chierchia-Gajewski approach, which presupposes feature checking to underlie both strong and weak NPI licensing.

I show in this talk that the distinction between strong/strict NPIs and weak/non- strict is best captured by modifying Chierchia’s (2013) approach to NPI-hood. I argue that there are two ways for establishing a relation between an NPI and a (covert) exhaustifier: one established in the syntax and one where NPIs are exhaustified in purely pragmatic way: if an element introduces sub-domain alternatives, at some pragmatic level, these domain alternatives must be exhaustified. This means that there are two different types of exhaustifiers, a syntactic and a pragmatic type. The next step is two assign these two types of exhaustifiers different semantic/pragmatic properties. The syntactic exhaustifier takes the enriched meaning contribution of an NPI licenser into consideration; the pragmatic exhaustifier only applies to the assertion. Then, NPIs exhaustified by the latter type are weak NPIs, the ones exhaustified by the former type are strong NPIs.

In the final part of the talk, I apply this analysis to PPI-hood and I show that a number of previously ununderstood properties of PPI-anti- licensing now follow naturally, including the fact that strong PPIs can never take scope below negation, but weak PPIs can, provided they appear below negation at surface structure.