GIANINA IORDĂCHIOAIA
Abstract. Despite the large literature that has been addressing Negative Polarity Items (NPIs) (see Baker (1970), Ladusaw (1980), Linebarger (1980), van der Wouden (1997) a.o.), they still remain an important topic for linguistic research. This is motivated, on the one hand, by the productivity of the phenomenon in natural language, and, on the other hand, to the recent computational developments in linguistics, which look for ways to automatically identify NPIs in large electronic linguistic corpora (see Hoeksema 2002, Sailer and Trawinski 2006). In a negative concord language like Romanian, NPIs are in competition with n-words, the typical concord items usually appearing with sentential negation. This paper proposes an investigation of the conditions under which the NPI vreun appears in Romanian negative contexts. It argues for a distinction between two semantic roles that negation plays – ‘predicate negation’
and ‘denial’ - of which the latter is responsible for licensing vreun.
0. INTRODUCTION
The present paper aims at characterizing vreun, a Romanian negative polarity item1 (NPI) which has a very particular behaviour, since in the presence of the negative marker (NM) ‘nu’ it is in competition with n-words. The kind of contexts that will be taken into account are exemplified below:
(1) a. Nu cunosc vreun / niciun medicament care să-l ajute.
NM know v-any2/ no medicine that Subj-him help
‘I don’t know of any medicine that can help him.’
b. Maria nu a citit *vreo/ nicio carte.
Maria NM has read v-any/ no book
‘Maria didn’t read any book.’
c. Ion vrea să împrumute bani nu de la vreo/ *nicio rudă, ci de la asociaţii lui.
1 I will exclusively deal with negative environments regarding the distribution of vreun. For a more complex description of vreun as an indefinite, that takes into account also its occurrence in non- negative contexts, the reader is referred to Farkas (2002) and Farkas (2005).
2 The translation of vreun is more or less equivalent to the English any. That is why I will gloss it as `v-any'.
RRL, LII, 1–2, p. 195–209, Bucureşti, 2007
John wants Subj borrow money not3 from v-any/ no relative, but from partners his.
‘John wants to borrow money not from any of his relatives, but from his business partners.’
I will argue that the condition for the licensing of vreun when it appears with clausemate negation is that the latter bears the semantic role of denial, as opposed to predicate negation which usually licenses n-words.
First I will present some general observations on the distribution of vreun in negative contexts. Further on, I will show why one cannot fully describe it by the usual means of describing other polarity items. In sections 3. and 4., the phenomenon of denial is presented and argued to be the right context for licensing vreun. The last section offers some general lines for an HPSG4 analysis.
1. THE DISTRIBUTION OF VREUN
First of all, a few remarks on the grammatical category of vre- indefinites should be made. Vre- usually appears as a determiner: vreun (masculine), and vreo (feminine). The only bare form available is the pronominal vreunul / vreuna (masculine / feminine ‘anyone’). Otherwise, the corresponding bare indefinite is cineva / ceva which stands both for ‘anybody’ / ‘anything’, and ‘somebody’ /
‘something’. Another quite frequent item containing vre- is the time adverbial vreodată ‘ever’. Here, I will use the masculine determiner vreun to refer to the class in general.
1.1. Downward entailing contexts
Vreun is the Romanian typical NPI for downward entailing contexts, as the examples5 in (2) clearly show:
(2) a. Interrogatives:
Ai văzut vreun ţigan fericit?
have seen v-any gypsy happy
‘Have you seen a / any happy gypsy?’
b. Conditional clauses:
Dacă găseşti vreo carte despre asta, cumpără-mi-o.
if find v-any book about this, buy-me-it
3 I will gloss nu as NM (“negative marker”) when placed on the verb, and as not when it bears constituent negation. See Barbu (2004) for the reasons why one should distinguish between the two syntactic roles of nu.
4 HPSG stands for Head-driven Phrase Structure Grammar.
5 The examples are taken from Farkas (2002).
‘If you find a book about this, buy it for me.’
c. Restriction of quantifiers:
De cîte ori s-a plâns vreun copil, a ieşit scandal.
of each time Refl-has complained v-any child has come out scandal
‘Each time a child complained, there was trouble.’
1.2. Clausemate negation
Romanian is a negative concord (NC) language, which means that in places where double negation languages like English use NPIs, Romanian employs n- words for the same purposes:
(3) Ion nu a spus nimic nimănui.
John NM has said nothing nobody
‘John didn’t say anything to anybody.’
Under these circumstances, ‘vreun’ is usually infelicitous with clausemate negation (see (4) below):
(4) *Maria nu a citit vreo carte.
Maria NM has read v-any book
‘Mary didn’t read any book.’
But, as Farkas (2002) points out, there are cases like (1a) above, when the presence of ‘vreun’ is optional to that of the n-word, the effect being that the former yields a ‘less categorical statement’ than the latter.
2. NEGATIVE CONTEXTS AND POLARITY
This kind of behaviour is quite unusual for an NPI, since clausemate negation is the negation per se. That is why the first thing one might think of when looking for an explanation for ‘vreun’ is to investigate the properties of negation and their influence on NPIs. Van der Wouden (1994) and van der Wouden (1997) offer a very detailed description of negative contexts and in relation with that, they define very refined conditions for the polarity phenomena. According to the strength of the negative operator, determined by making use of De Morgan’s laws, three types of negative expressions result: downward entailing, anti-additive and antimorphic.
They are characterized as follows:
(5) De Morgan’s Laws:
a. ¬(X intersection Y) = ¬(X) union ¬(Y) b. ¬(X union Y) = ¬(X) intersection ¬(Y)
(6) Negative Expressions:
a. Downward entailing: ‘few’, ‘at most three’, ‘hardly’
(X subset of Y) entails f(Y) subset of f(X) b. Anti-additive: ‘nobody’, ‘never’, ‘nothing’
f(X union Y) = f(X) intersection f(Y) c. Antimorphic: ‘not’, ‘not the teacher’
f(X intersection Y) = f(X) union f(Y) f(X union Y) = f(X) intersection f(Y)
Following the classification in (6), van der Wouden delineates three classes of negative/ positive polarity items, in terms of the strength that they display in their (in)compatibility with negative contexts. The table below summarizes all these situations:
(7)
Negation NPI PPI
Downward entailing Anti-additive Antimorphic
strong -- -- +
medium -- + +
weak + + +
strong -- -- --
medium + -- --
weak + + --
For English, one would characterize ‘a bit’ as a strong NPI, since it is compatible only with antimorphic negation (cf. 8a) , ‘yet’ as of medium strength, due to its ungrammaticality with DE contexts (cf. 8b), while ‘any’, which is felicitous with all negative expressions, qualifies as a weak NPI (see 8c):
(8) a. [Chomsky wasn't/ *No one was/ *At most three linguists were] a bit happy about these facts.
b. [Chomsky didn’t talk/ No one talked/ *At most three linguists talked] about these facts yet.
c. [Chomsky didn’t talk/ No one talked/ At most three linguists talked] about any of these facts.
2.1. “Vreun”
If we introduce the Romanian vreun into this picture, we tend to classify it as a weak NPI, since it is compatible with downward entailing environments (cf. 2), and incompatible with neutral6 affirmative contexts (cf. 9):
6 I will ignore in the present discussion hypothetical contexts (cf. Farkas 2002), where vreun can also appear.
(9) * Ioana a citit vreo carte de Chomsky.
Joanna has read v-any book by Chomsky
‘Joanna read any book by Chomsky.’
But although vreun can successfully appear with anti-additive nimeni in (10), with antimorphic nu it is usually infelicitous, as we have already seen in (4) above:
(10) Nimeni n-a citit vreo carte de Chomsky.
nobody NM-has read v-any book by Chomsky
‘Nobody read any book by Chomsky.’
The ungrammaticality of (4) is very atypical for an NPI, and it draws one’s attention towards weak PPIs, which are characterized as compatible with downward monotonicity and anti-additivity, but incompatible with antimorphicity.
However, the ill-formedness of affirmative sentences containing vreun, as (9), prevents us from concluding that it is a PPI. It makes no sense for a PPI to be disallowed in positive contexts.
2.2. Bipolarity?
Van der Wouden points out a more or less similar case in Dutch. The item ooit' (‘ever’) is usually ungrammatical in contexts that are not monotone decreasing and under antimorphic negation. It is fine with anti-additive ‘geen van de kinderen’ (‘none of the children’), and with the downward monotone weinig kinderen (‘few children’):
(11) a. *Een van de kinderen gaat ooit bij oma op bezoek.
one of the children goes ever with granny on visit
‘One of the children ever visits granny.’
b. Weinig kinderen gaan ooit bij oma op bezoek.
few children go ever with granny on visit
‘Few children ever visit granny.’
c. Geen van de kinderen gaat ooit bij oma op bezoek.
none of the children goes with granny on visit
‘None of the children ever visits granny.’
d. *Een van de kinderen gaat niet ooit bij oma op bezoek.
one of the children goes not ever with granny on visit one of the children never visits granny.
Van der Wouden calls ooit a bipolar element, since it displays properties that go with both negative and positive polarity, so it is somewhere between weak NPIs and weak PPIs. I will not argue with the notion of bipolarity, although a singular
case that would belong to this class out of the impressive number of polarity items that have been so far identified in Dutch7 makes one skeptical about its empirical motivation. Besides that, van der Wouden (1994) mentions a remark of Jack Hoeksema, who points out that ooit is losing its NPI character, since it appears in sentences like (12), which were completely ill-formed a century ago:
(12) Ooit kende Groningen meer dan duizend molens.
ever knew Groningen more than thousand mills ‘Groningen used to have more than a thousand mills.’
Even if one accepts the existence of bipolar elements, it is doubtful that this could explain the behaviour of vreun with respect to negation. First of all, for Romanian, assuming another class of polarity items doesn’t throw more light on the interaction between the negative operator, on the one hand, and polarity items and n-words, on the other.
Secondly, in negative contexts vreun is synonymous with the typical weak PPI ‘ceva’ / ‘cineva’ (‘something’ / ‘somebody’), although it differentiates from the latter, by incompatibility with affirmative environments. Ceva qualifies as a weak PPI, since it fits perfectly the profile in (7): grammaticality with monotone decreasing and anti-additive8 expressions, and ungrammaticality with the antimorphic ones:
(13) a. Ioana a citit ceva de Chomsky.
Joanna has read something by Chomsky ‘Joanna read something by Chomsky.’
b. Puţini studenţi au citit ceva de Chomsky.
few students have read something by Chomsky
‘Few students read something by Chomsky.’
c. ?Nimeni n-a citit ceva de Chomsky.
nobody NM-has read something by Chomsky
‘Nobody read something by Chomsky.’
d. *Ioana n-a citit ceva de Chomsky.
Joanna NM-has read something by Chomsky
‘Joanna didn’t read something by Chomsky.’
7 Hoeksema (2002) speaks of about 700 Dutch polarity items.
8 The example with nimeni sounds a little unnatural due to the ambiguity of ceva (see the remarks below the examples in (13)). But another anti-additive expression like fără (‘without’) is perfectly fine:
e. Ioana a ştiut răspunsul fără să fi citit ceva de Chomsky.
Joanna has known answer-the without Subj be read something by Chomsky
‘Joanna knew the answer without having read something by Chomsky.’
Between ceva and vreun, the latter is clearly perceived as the NPI. This is probably the effect of the possibility to associate a specific interpretation to ceva, something that never happens with vreun, or NPIs in general. As a result, in spoken language, and at least dialectally, one can encounter forms like vreoceva, where the speakers feel the need to disambiguate ceva to a pure NPI. Under such circumstances, it is hard to believe that vreun would resemble the Dutch ooit. In Romanian, the tendency seems to follow the other direction: vreun distinguishes as the negative part of the weak polarity pair vreun - ceva.
3. DENIAL
Considering the remarks above concerning the ungrammaticality of vreun with clausemate negation, one may wonder what makes (1a) possible in the vreun version. First of all, one should pay attention to the differences between the two variants: with vreun' and with the n-word. The most obvious such distinction involves intonation. With vreun, the NM has to bear a special emphasis9, which would result in ungrammaticality with the n-word:
(14) NU cunosc vreun / *niciun medicament care să-l ajute.
NM know v-any / no medicine that Subj-him help
If the negative marker nu bears neutral intonation, only the n-word is grammatical.
This particular intonational pattern points at another property of such constructions, which has to do with the previous discourse. If in a dialogue one speaker asserts (15a), the other speaker can reject it by uttering (15b), with or without giving his/ her reasons for the objection. In such a case, the NM nu is intonationally marked:
(15) a. Maria sigur a citit cartea asta.
Maria sure has read book this
‘Maria has surely read this book.’
b. Maria NU a citit cartea. (Mi-a zis ea mie astăzi.) Maria NM has read book (me-has told she me today)
‘Maria didn’t read the book. (She told me that today.)’
This is exactly the kind of context that makes the presence of vreun in (14) felicitous. If it is used, the sentence conveys the information that somebody in the discourse, or even the speaker himself, was expecting or presupposing that the one
9 I will mark such an emphasis by writing the word in capital letters, like in (14).
who utters (14) should know of a medicine that could help. A similar example is offered in (16) below:
(16) Context: A terrible accident took place and people around don't know how it happened.
a. Speaker A: Aştept să aud explicaţia pentru cele întâmplate.
wait Subj hear explanation-the for those happened
‘I’m expecting to hear the explanation for what has happened.’
b. Speaker B: NU există vreo explicaţie. (Nimeni nu ştie nimic.) NM exists v-any explanation (nobody NM knows nothing)
‘There is no explanation. (Nobody knows anything.)’
The utterance made by Speaker A presupposes that there is an explanation which (s)he is missing. With (16b), Speaker B cancels A’s assumption.
3.1. Metalinguistic negation
Horn (1985) and Horn (1989) discuss at large such instances of negation, which are called “metalinguistic negation”, as opposed to “descriptive negation”.
In such constructions, Horn argues that negation affects only the pragmatics of a sentence, and that we have to assume that there is one more negative operator, that comes in play only on such occasions. The term used by Horn and his ambiguist analysis brought about several reactions, that were mainly due to a narrow understanding of “metalinguistic”. In this sense, Van der Sandt (1991) offers a refined analysis of the same phenomenon, which he calls ‘denial’. Various linguists have employed one of the two terms, but I will use denial in oder to avoid misinterpretation. I am also of the opinion – expressed in Van der Sandt (1991) and Geurts (1998) among others – that there is no need to employ a different negative operator to explain denial.
There are several instances of denial, which can be classified like in (17) below10:
(17) a. Proposition denial:
A: It’s raining.
B: It’s not raining.
b. Presupposition denial:
A: Fred has stopped smoking.
B: Fred hasn’t stopped smoking: he never did smoke.
10 The examples are taken from Geurts (1998).
c. Implicature denial:
A: Julius had six beers.
B: He didn’t have six beers: he had at least seven.
d. Form denial:
A: Kurt swallowed a whole to[ma:]to.
B: He didn’t swallow a to[ma:]to but a to[mei]to.
One of the merits that Horn’s thorough investigation has is that it offers three reliable diagnostics to identify denial. They address the relationship between denial and the possibility to prefixally incorporate, the interaction with polarity, and a particular syntactic pattern with contrastive ‘but’, later on addressed in McCawley (1991). Here I will present these tests, in order to show that they predict the compatibility of vreun with denial.
3.2. Incorporation
One way to delineate denial is to test its possibility to incorporate prefixally.
In contexts with descriptive negation, either a prefix or a word, the negative operator yields similar grammatical results:
(18) a. The king of France is {not happy / unhappy} – he has some health problems.
b. The queen of England is {not happy / unhappy} – she’s worried about her son’s future.
If a prefix, Horn shows, the negative operator loses the chance to take scope over presuppositions and implicatures, as the unavailability of the incorporated versions of negation can prove in (19):
(19) a. The king of France is {not happy/ # unhappy} – there isn’t any king of France.
b. The queen of England is {not happy/ # unhappy} – she’s ecstatic.
3.3. Polarity
Another property that Horn attributes to metalinguistic negation (or denial, in our case) is the two-sided atypical interaction with polarity phenomena. On the one hand, it does not trigger negative polarity items, and on the other hand, it does not inhibit the presence of positive polarity items. Usually, the negative operator integrates in the structure of the sentence and it licenses NPIs, while it inhibits the presence of PPIs. Denial seems to display exactly opposite traits:
(20) a. Chris didn’t manage to solve the problems.
i. Chris didn’t solve the problems, it was too difficult for him.
ii. Chris did solve the problems, it was quite easy for him.
b. Chris didn’t manage to solve any / *some of the problems. (= i)
c. Chris didn’t manage to solve some / *any of the problems, he solved all of them.
(= ii)
A sentence like (20a) alone can have both readings in (i) and (ii), with the negative operator receiving either a descriptive or a denial interpretation. The examples in (20b) and (20c) show how these two readings interact with polarity: an NPI is possible only with the descriptive version, while a PPI only with denial.
3.4. Contrastive but
Either across speakers or within one speaker's contribution in a discourse, denial tends to occur in contrastive contexts. In accord with this observation and following the model given by Anscombre, Ducrot (1977) for French, Horn (1989) shows that the two uses of the English conjunction but offer a syntactic test for denial: contrastive but can only appear with denial, while the concessive version appears with descriptive negation. The difference between the two uses of but has a syntactic character, exemplified below:
(21) a. It isn’t hot, but it is warm. (concessive but) b. It isn’t hot, but scalding. (contrastive but) c. It isn’t hot, (# but) It’s scalding.
The examples above indicate that contrastive but is only a rectifier, not a true sentential connective like the concessive: its inability to show up with a full sentence constitutes clear evidence in this sense (see 21c). A concessive understanding of (21c) is pragmatically excluded, since it is inconsistent to affirm that something is scalding, yet not hot.
4. “VREUN” AND DENIAL
The assumption made here is that vreun appears with clausemate negation only if the latter expresses denial. If this is the case, vreun should verify the predictions of the three diagnostics above. Incorporation however is not applicable in our case, since prefixal negation has a lexical nature, and thus fails to display syntactico-semantic interactions with the rest of the sentence. Vreun being an instance of polarity, the negative operator that licenses it should be of a higher syntactic level than that of a prefix. But the interaction of denial with polarity and contrastive but offers us more information is this respect.
For English, Geurts (1998) shows that NPIs are not completely excluded with denial as Horn was trying to show. For (22), the NPI is more natural than the PPI:
(22) Walter didn’t give his ukulele to {?somebody / anybody}: he never owned a ukulele.
But this happens because anybody is a weak NPI. A strong NPI like one red cent is indeed totally excluded, confirming Horn’s test:
(23) *The king of France didn’t contribute one red cent, because there is no king of France.
Romanian employs vreun exactly with the same role that any plays in (22), and an NPI like o para chioară loses its idiomatic use in the context of denial, in the same way in which one red cent' in (23) above does:
(24) a. Ion NU şi-a dat chitara vreunui prieten; de fapt, n-a avut niciodată chitară.
John NM his-has given guitar v-any friend; of fact NM-has had never guitar
‘John didn’t give his guitar to any friend; in fact, he never had a guitar.’
b. *Regele Franţei NU a donat o para chioară. Nici măcar nu există un rege al Franţei.
king of France NM has donated a-red-cent. not even NM exists a king of France Concerning contrastive contexts, Romanian, like Spanish and German, and unlike English and French, uses a different lexical item from concessive but. The pair contrastive – concessive is represented by ci and însă / dar, respectively. In (26), we can follow the interaction between ci / însă and vreun:
(25) a. NU e căldură, ci caniculă.
NM is warmth, but heat
‘It’s not warm, but hot.’
b. Supa nu e fierbinte, însă e caldă.
soup-the NM is hot, but is warm
‘The soup is not hot, but it is warm.’
(26) a. Context: For the meeting, every participant was asked to bring relevant poems for the discussion.
Ion NU a adus vreo poezie, ci nişte nuvele.
John NM has brought v-any poem, but some short stories
‘John didn’t bring poems, but some short stories.’
b. Context: For the meeting, every participant was asked to bring both poems and short stories relevant for the discussion.
Ion nu a adus *vreo/ nicio poezie, însă a adus multe nuvele.
John NM has brought v-any/ no poem, but has brought many short stories
‘John didn’t bring any poems, but he brought many short stories.’
In (26a), we are dealing with a context where it is assumed that everybody who was present at the meeting had brought some poems. The speaker who utters the sentence objects to this assumption, by saying that somebody actually brought short stories. This is a contrastive context, where as expected Romanian employs the use of ci. Vreun is here perfectly grammatical. In (26b), there was a requirement for everybody to bring both poems and short stories. John is one participant who didn’t bring poems at all, but he at least brought plenty of short stories, which compensates for the lack of poems. This is a concessive context, where Romanian uses însă. But here vreun is not felicitous; only an n-word can appear. If we consider Horn’s claim that before concessive but denial is excluded, the ungrammaticality of vreun in (26b) is easily explained.
The syntactic pattern of contrastive negation seems to be the typical one for vreun, since here the presence of an n-word is not an option anymore. The example in (1c) makes this obvious. Moreover, the syntactic varieties of contrastive negation that McCawley (1991) presents are all compatible with vreun:
(27) Contrastive negation:
a. basic form:
Ion a împrumutat bani NU de la vreo rudă, ci de la asociaţii lui.
John has borrowed money not from v-any relative, but from partners his
‘John borrowed money not from a relative, but from his business partners.’
b. reverse form:
Ion a împrumutat bani de la asociaţii lui, NU de la vreo rudă.
John has borrowed money from partners his, not from v-any relative
c. anchored form:
Ion NU a împrumutat bani de la vreo rudă, ci de la asociaţii lui.
John NM has borrowed money from v-any relative, but from partners his
d. basic expanded form:
Ion NU a împrumutat bani de la vreo rudă, a împrumutat (bani) de la asociaţii lui.
John NM has borrowed money from v-any relative, has borrowed (money) from partners his
e. reverse expanded form:
Ion a împrumutat bani de la asociaţii lui, NU a împrumutat de la vreo rudă.
John has borrowed money from partners his, NM has borrowed from v-any relative Although in some of the contexts above (more precisely in those where negation is on the verb) an n-word is acceptable too, ‘vreun’ is always grammatical: with a basic or a reverse form (like in 27a and 27b), and with the anchored and the two expanded forms in (27c) and (27d) - (27e), respectively.
5. TOWARDS AN ANALYSIS AND FURTHER MATTERS
5.1. Some analytical consequences
The observation above with respect to (27) goes very well with Horn’s argument that contrastive contexts are indicators of denial, and with the claim in this paper, that vreun appears with denial. Moreover, the fact that on the one hand denial is incompatible with strong NPIs, and not excluded with PPIs and weak NPIs, while on the other hand the status of vreun is undecided between weak NPIs and weak PPIs within van der Wouden’s typology points again at the strong connection between denial and vreun.
However, note that trying to classify vreun according to the negative contexts established by van der Wouden is doomed to failure, since nu (‘not’) is antimorphic both in (2b), where it appears with an n-word and excludes the occurrence of
‘vreun’, and also in (2c) and (27a,b) where ‘vreun’ is grammatical, and the n-word is not an option anymore. So antimorphicity cannot tell us anything about the distribution of ‘vreun’, unlike the semantic use of negation as denial.
Another fact that can be easily explained by considering denial the right context for vreun is Farkas’s observation with respect to the difference between the two variants in (1b). She says that the vreun version is “less categorical” than the n- word version. If we consider the presentation of the two semantic roles of negation in Horn (1989: 132-144), denial would represent the weaker, contradictory paradigm which is entailed but cannot entail the other kind of negation. Assuming, along with Przepiórkowski, Kupść (1991), that n-words show up with the stronger negation, it is expected that the version with denial and vreun is weaker (i.e. also
“less categorical”) than the one with the n-word.
5.2. Denial and vreun
The question that arises now is how to account for denial and how to predict the distribution of vreun. For “metalinguistic negation”, Horn was arguing for a pragmatic ambiguity, and his solution was to call in another negative operator, that
would apply exclusively to non-truth-functional phenomena like presuppositions or implicatures. But he was not considering cases of proposition denial like in (16b) and (17a) above, and in this sense, Van der Sandt clearly shows that, no matter how pragmatic this ambiguity is pretended to be, it practically boils down to lexical ambiguity for the negative operator, which is not desirable in linguistic analysis.
More than that, the particular behaviour of `vreun' with clausemate negation is showing now that there is more syntactico-semantic interaction between denial and the rest of the sentence than Horn was ready to take into account.
Although a full analysis of the phenomenon is not aimed at here I defend the idea that negation should be regarded as semantically ambiguous between
“predicate negation”11 and denial. In HPSG, this ambiguity can be easily implemented by stating a subdivision of the semantic type that would be usually assigned to negative expressions. The type hierarchy would thus include two more types. The compromising lexical ambiguity that Horn was promoting is thus replaced by a type subspecification. A richer type hierarchy does not create any problem for grammar implementation systems like TRALE12, which are built specifically for such purposes.
With respect to the lexical entry of vreun one can successfully adopt Przepiórkowski and Kupść (1991)’s way of characterizing Polish n-words as idiosyncratic items sensitive to a particular type of negation. In its lexical entry, vreun would ask for a denial13 semantic object to retrieve the existential quantifier that it contributes. The quantifier retrieval and the whole licensing can be very well worked out as in Przepiórkowski, Kupść (1991).
5.3. Prefixal vs. lexical nu
A matter that would simplify the whole analysis is to import the distinction that Barbu (2004) makes with respect to nu (‘not’). It is argued that when preceding the verbal complex (like in (1a,b)), nu behaves like an affix, while in the other circumstances it acts as a full lexical word, and it functions as a constituent modifier (like in 1c). Apparently, the use of ‘nu’ as a modifier is limited to contrastive contexts like the ones we’ve seen above, so it always bears denial.
In conclusion, a lexical ambiguity would arise concerning nu' only that this will not affect the lexicon itself, because the prefix would be attached to the verbal complex via a lexical rule, and nu remains to have a lexical entry as a modifier, and
11 I use ‘predicate negation' here with the same meaning that ‘descriptive negation’ has been used so far in opposition to denial.
12 More information on TRALE (an implementation platform for HPSG grammars) can be found at http://www.sfs.uni-tuebingen.de/hpsg/archive/projects/trale/ and http://www.ling.ohio- state.edu/~vmetcalf/trale_prolog.html
13 Of course, in order to include DE expressions and other contexts where ‘vreun’ appears, one could specify a supertype that would include all of these environments.
its semantics will be specified as containing semantic negation under the form of denial. Modifier nu would thus be able to license vreun anywhere inside the modified constituent.
6. CONCLUSION
This paper has mainly dealt with the behaviour of the NPI vreun with clausemate negation. It was argued that the exceptional cases when it appears and the particular characteristics that the constructions display are due to a special semantic function of negation, that of being used as denial. These observations prove that denial is not only a pragmatic phenomenon, but it has syntactico- semantic properties that should be taken into account in the perspective of developing a consistent grammar of the Romanian negation system.
REFERENCES
Anscombre, J.-C., O. Ducrot, 1977, “Deux mais en français?, Lingua, 43, 23-40.
Baker, C. L., 1970, “Double negatives”, Linguistic Inquiry, 1, 169–186.
Barbu, A.-M., 2004, “The negation nu: lexical or affixal item?”, in: E.Ionescu (ed.), Understanding Romanian Negation, Bucureşti, Editura Universităţii din Bucureşti.
Farkas, D., 2002, “Extreme non-specificity in Romanian”, in: C.Beyssade et al. (eds), Romance Languages and Linguistic Theory 2000, John Benjamins.
Farkas, D., 2005, “Free choice in Romanian”, to appear in: G. Ward and E. Birner (eds), Festschrift for Larry Horn,.
Geurts, B., 1998, “The mechanisms of denial”, Language 74, 274–307.
Hoeksema, J., 2002, “A lexicon for linguists: the lexicon of Dutch negative polarity items”, at http://odur.let.rug.nl/hoeksema/lexiconbestanden/v3document.htm.
Horn, L. R., 1985, “Metalinguistic negation and pragmatic ambiguity”, Language 61, 121–174.
Horn, L. R., 1989, A Natural History of Negation, The University of Chicago Press, Chicago, London.
Ladusaw, W., 1980, Polarity Sensitivity as Inherent Scope Relations, New York, Garland Press.
Linebarger, M., 1980, The Grammar of Negative Polarity, PhD Thesis, MIT.
McCawley, J. D., 1991, “Contrastive negation and metalinguistic negation”, Chicago Linguistic Society. The Parasession on Negation 27.2, 189–206.
Przepiórkowski, A., A Kupść, 1999, “Eventuality negation and negative concord in Polish and Italian”, in R.Borsley and A.Przepiórkowski (eds.), Slavic in HPSG , CSLI, Stanford.
Sailer M., B. Trawinski, 2006, “The Collection of Distributionally Idiosyncratic Items: A Multilingual Resource for Linguistic Research”, in Proceedings of LREC 2006.
Van der Sandt, R., 1991, “Denial”, Chicago Linguistic Society. The Parasession on Negation 27.2, 331–344.
van der Wouden, T., 1994, “Polarity and ‘illogical’ negation”, in: M. Kanazawa, C. J. Pinon (eds), Dynamics, Polarity, and Quantification, CSLI, Stanford.
van der Wouden, T., 1997, Negative Contexts. Collocation, polarity and multiple negation, Routledge, London, New York.
