Negative concord in Washo as negative agreement

Washo, a Native American isolate, displays negative concord morphology in the context of negation. Negative concord in Washo comes in the form of the morpheme -Na, which may be suffixed onto optionally many sentential elements in a single clause. Given the apparent lack of semantic contribution by this morpheme, I argue building on accounts along the lines of Zeijlstra (2004) that negative concord in Washo is best treated as the result of multiple agreement between an interpretable Neg feature high in the clause and optionally many uninterpretable Neg features present on the items it c-commands. Importantly, the Washo data shed light on an unexplored system of negative concord and the possible range of agreement phenomena that are sensitive to negation.

1. Introduction.This paper concerns itself with an unfamiliar type of negative concord found in Washo, in which the suffix -Na occurs on optionally many sentential elements within the scope of negation.The examples in (1-2) illustrate this distribution by showing that -Na may appear on the object t'ág1m 'pinenuts' when the sentence is negated (1), and may not appear otherwise (2). 1 Negation in Washo is expressed by the suffix -é:s.
(1) mé:hu boy t'á:g1m -Na pinenuts-NC P-íPiw -é:s -i 3/3-eat-NEG-IND 'The boy didn't eat pinenuts.' Washo Archive (2) *mé:hu boy t'á:g1m -Na pinenuts-NC P-íPiw-i 3/3-eat-IND 'The boy ate pinenuts.'While -Na has been glossed informally as a "negative concord" morpheme in various works (i.a.Bochnak et al. 2011;Bochnak 2013;Hanink 2016Hanink , 2018)), a dedicated investigation of the factors contributing to its licensing has not before been carried out.The aim of this paper is therefore twofold: to show that the suffix -Na i) is in fact a negative concord morpheme, but does not behave like one in the familiar sense of the term, and ii) is best treated as the realization of negative agreement.The Washo data therefore provide novel support for syntactic accounts of negative concord in which dependent negative morphology is treated as a reflex of agreement (i.a. Zeiljstra 2004;Penka 2007;cf. Haegeman & Londahl 2010).
The outline of this paper is as follows.In §2, I provide background on Washo and describe the wide distribution of -Na in the language.In §3, I discuss previous approaches to the analysis of negative concord, which I then show to be problematic for the Washo facts.In §4, I present a preliminary, agreement-based analysis based on current understanding of the data, and briefly discuss a potential confound in §5.§6 concludes.
2. Negative concord in Washo.Washo is a highly endangered Native American language spoken around Lake Tahoe in the United States.While largely considered to be an isolate, it has also been linked to the proposed Hokan group (Campbell, 1997;Mithun, 1999).The neutral word order of the language is SOV, with phrases being largely head-final (aside from nominal phrases and their modifiers, which are neutrally word-initial).The uncited data in this paper come largely from my fieldwork with speakers from the Washo community in Nevada.
The distribution of -Na in Washo is highly flexible.For example, it can occur on any argument of the verb, including subjects (3), direct objects (4) (repeated from (1)), and indirect objects (5), as long as that argument is under the scope of negation: (3) subject: béverli-Na Beverly-NC This suffix can also occur on non-arguments within the clause, for example on adverbs, as in (6), as well as on postpositional phrases, as in ( 7):

Negative concord and agreement.
3.1 NEGATIVE CONCORD.Canonically, negative concord describes the co-occurence of a negative dependent with an independent expression of negation (see Giannakidou & Zeijlstra 2017 for a recent overview).Crucially, the joint appearance of more than one instance of morphological negation in negative concord languages does not give rise to multiple negation in meaning.This phenomenon is exhibited in many European languages, for example in Italian (12) and Polish (13), in which sentential negation co-occurs with n-marked negative elements: (12) Italian Gianni Gianni *(non) not ha has visto seen niente n-thing 'Gianni hasn't seen anything.'Not: 'Gianni has not seen nothing.' (13) Polish Janek Janek *(nie) not pomaga help nikomu n-person 'Janek doesn't help anybody.'Not: 'Janek doesn't help nobody.' Giannakidou & Zeijlstra (2017: 8) In ( 12) and ( 13), the presence of the negative dependents niente and nikomu, respectively, does not result in a double negative reading.This differs crucially from the behavior of a language like Dutch, where each negative element contributes its own negation: (14) Jan Jan belt calls niet neg niemand n-body 'Jan doesn't call nobody = Jan calls somebody ' Zeijlstra (2008: 2) Negative dependents such as niente and nikomu in Italian ( 12) and Polish ( 13) are canonically referred to as n-words (Laka 1990;Giannakidou 2000), referring to the morphological shape in which they appear.Giannakidou (2006) offers the follow definition for such dependents: (15) n-words An expression α is an n-word iff: a. α can be used in structures containing sentential negation or another α-expression yielding a reading equivalent to one logical negation; b. α can provide a negative fragment answer.Giannakidou (2006: 2)2 Based on this definition, -Na is n-word-like with respect to condition (a): the dependent morphology is negative, yet only a single logical negation results.In fact, given the type of data we find in Washo, it is difficult to imagine what it would mean were -Na contributing an additional negative meaning in examples such as (16) (repeated from (9)): (16) Adél-Na Adele-NC wáP-Na here-NC P-áNal-é:s-i 3-reside-NEG-IND 'Adele doesn't live here.'I note here that condition (b) does not apply to Washo, which seems to lack fragment answers entirely -even the polar answers 'yes' and 'no' -and so I do not discuss this condition.
Neither of these treatments extend to Washo, for several reasons.First, n-words in Washo are instead n-morphemes.As such, the negative dependents observed in the language are not lexical items in their own right; rather, they are functional morphemes that occur on independent lexical items whose forms are otherwise invariant.Second, the distribution of -Na is much more flexible in Washo than is generally found for n-words in more familiar negative concord languages, most crucially in that they are not limited to existential or universal meanings.The latter point can be seen, for example, from the fact that -Na does not have an effect on scope relations.The examples in ( 17)-( 18) show that the -Na does not affect scope when it surfaces on an indefinite, either in subject or object position: Nevertheless, the similarities that do exist between -Na and n-words suggest that this morpheme should be thought of as a type of negative concord, albeit one that is unfamiliar from the view of better-studied languages.

NEGATIVE CONCORD AS AGREEMENT.
In a different type of approach that is purely syntactic, Zeijlstra (2004) argues that negative concord is not a semantic phenomenon, but is instead the result of agreement (see also Penka 2007;Zeijlstra 2008).On this view, n-words are simply non-negative probes for Agree.According to Zeijlstra (2004), n-words are then a sub-set of strong NPIs that must be licensed by negation in the syntax, and that enter the derivation as semantically non-negative indefinites.Adopting a Minimalist approach of feature checking (along the lines of Chomsky (1995Chomsky ( , 2000Chomsky ( , 2001))), Zeijlstra proposes that these indefinites carry an uninterpretable uNEG feature that must be checked against a higher, semantically negative element bearing an iNEG feature, which for Zeijlstra is a non-pronounced negative operator high in the clause (in Spec, NegP), rather than any overt realization of negation.
A crucial part of his analysis is moreover the adoption of Multiple Agree (Hiraiwa 2001), which accounts for the availability of multiple n-words in a given agreement domain (cf.Haegeman & Londahl 2010, who argue against the use of Multiple Agree in their account of West Flemish).In a nutshell, on this view, negative concord is reduced to multiple agreement between a negative operator and the negative items it c-commands (via Upward Agree; i.a.Merchant 2006Merchant , 2011;;Baker 2008;Zeijlstra 2012).
3.4 SYNTACTIC LICENSING IN WASHO.It was shown above that the suffix -Na in Washo displays behaviors consistent with better known instances of negative concord in European languages.For example, it also behaves like a strong NPI, in that it is licensed only by negation and not by, e.g., questions (21) or conditionals ( 22 ∅-wagayáNa-hel-i-gi 3-talk.to-SUBJ-IND-NMLZk'-éP-i 3-be-IND 'If the boy were here, he would talk to the girl.'Secondly, -Na neither contributes any additional meaning of its own, nor does it have an effect on scope relations, ruling out an interpretation-based account.Taken together, these behaviors suggest that the most promising analysis on the market for Washo is one involving agreement. Further evidence for an agreement analysis comes from the syntactic licensing conditions of -Na.First, -Na is licensed only when in the scope of syntactic negation, which may come either in the form of sentential negation (23) (repeated from (2)), or constituent negation (24), both of which are realized by the suffix -é:s: Additionally, -Na is licensed only locally.This can be seen in the context of embedded clauses, which may not contain -Na-marked elements in the absence of clausemate negation.In [ CP Adél [uNEG] Note that this implementation of the analysis differs crucially from Zeijlstra (2004), according to which the sentential elements bearing uNEG are necessarily indefinite: In Washo, uNEG is free to occur on anything.
Unlike Zeijlstra moreover, the negative element I assume to bear iNeg is (sentential) negation itself.This is sufficient in the general case, as Washo is an SOV language, and the negative suffix is high enough in the clausal periphery such that the c-command condition for Agree will be met.The structure of a clause such as (30a), for example, is schematized in (30b).The negative suffix is hosted by its own projection, NegP, which intervenes between TP and MoodP, the clausal anchor in Washo (see Bochnak (2016); Hanink & Bochnak (2018) for more on MoodP in Washo). 3This ordering is consistent with the morphology of the language, in accordance with the Mirror Principle (Baker 1985).Note that there is no evidence for the subject moving higher than Spec, TP in Washo (cf.Arregi & Hanink 2018), though one potential issue for this assumption comes from whmovement.Washo is an optional wh-movement language, and so we might expect wh-words that have undergone movement to be outside the scope of negation.The behavior of -Na with respect to Ā-movement requires further investigation. 4.Ruling out other licensing factors.In the remainder of this paper, I offer a brief discussion regarding the issue of ruling out other factors that might contribute to the licensing of the negative concord suffix in Washo.The reason for this discussion is that the interaction of nega-tion with other (semantic/pragmatic) factors can be notoriously difficult to understand, e.g., the seemingly superfluous VP-final nie in Afrikaans, which Merchant (2017) treats as a negativeisotope VP-level clitic; the augment morpheme across Bantu, which may or not be sensitive to negation (i.a. Mzolo 1968;Halpert 2012); the genitive of negation in languages such as Russian (see i.a. Brown 2010;Partee & Borschev 2002), where the optional genitive case marking of sentential elements under the scope of negation has been argued to be sensitive to semantic or pragmatic factors.
Based on such complications, one factor that remains to be understood in Washo is whether -Na contributes some semantic meaning that is not immediately obvious in a fieldwork setting.A good contender for such a contribution might come, for example, from focus, a domain with which negative concord has been closely tied before (e.g., Watanabe (2004) on Japanese).Preliminary work suggests that focus does not play a role, however.First, the negative concord suffix is not obligatory in clear focus contexts involving negation, such as in ( 31 Second, there are specific contexts in elicitation that do not lend themselves to a focus interpretation.One such context is given in (32), where the context involves a first mention of a particular 'big rock', backgrounding it.For this reason, focus-marking should not be expected on 'big rock' in the follow-up utterance.Nevertheless, -Na appears: (32) Context: you go walking, and someone asks if you saw the big rock by the lake.déPek rock t'íyeliP-Na big-NC l-í:gi-yé:s-i 1-see-NEG-IND 'I didn't see the big rock.'Third, -Na is never licensed without negation in non-negative focus contexts, as in (33), which contains the focus-sensitive restrictive suffix -k'eN, which means something like 'only': The upshot of this is that focus does not seem to be at play, though that does not necessarily rule out the contribution of other semantic or pragmatic factors to the licensing of -Na.More fieldwork is needed to determine whether such other factors exist.Importantly however, the syntactic licensing conditions for negative concord remain.If there are in fact other factors governing the appearance of -Na, the only aspect of the analysis presented in §4 that would change would be the optionality of this morpheme.

Conclusion.
The negative concord suffix -Na in Washo presents a novel type of negative agreement marking that is best captured by an agreement analysis invoking Multiple Agree, along the lines of the proposal put forward by Zeijlstra (2004).The Washo data presented in this paper are important in two ways.First, they expand our understanding of the possible range of agreement phenomena, as they present a hitherto unfamiliar system of negative agreement.Second, they show with direct evidence that agreement analyses of negative concord are on the right track, at least for certain languages.