(Sobre a Natureza e as Condições de Licenciamento de Sintagmas-N em Português)

The present paper focuses on the syntactic and semantic nature of the expressions identified in the literature as n-words (preferably, n-phrases), and on their licensing conditions. Concerning their status, arguments will be given in favor of Ladusaw’s 1992 thesis that these are existential (and non-specific, it is claimed) indefinites. In a brief excursus, it will be shown that other constructs engage in the process known as “negative concord”. In the final part, an attempt will be made to offer a systematic picture of the intra– and cross-sentential licensing conditions of classical n-phrases. In this regard, the paramount importance of contextual decreasing monotonicity becomes apparent.

and on their licensing conditions.Concerning their status, arguments will be given in favor of Ladusaw's 1992 thesis that these are existential (and non-specific, it is claimed) indefinites.In a brief excursus, it will be shown that other constructs engage in the process known as "negative concord".In the final part, an attempt will be made to offer a systematic picture of the intra-and cross-sentential licensing conditions of classical n-phrases.In this regard, the paramount importance of contextual decreasing monotonicity becomes apparent.KEY WORDS: N-Words, Negative Concord, Indefinites, Decreasing Monotonicity RESUMO: O presente artigo centra-se na questão do estatuto sintáctico e semântico das expressões que têm sido identificadas na literatura como nwords (preferivelmente, n-phrases, sintagmas-n) e nas condições do seu licenciamento.Sobre o seu estatuto, argumenta-se em favor da tese de Ladusaw 1992 de que se trata de expressões com valor indefinido existencial (e não-específico, acrescenta-se).Num breve excursus, mostra-se que outras estruturas além de sintagmas deste tipo se envolvem no processo conhecido como "concordância negativa".Na parte final, tenta-se sistematizar as condições de licenciamento intrafrásico e transfrásico dos sintagmas-n clássicos, ressaltando a importância das propriedades de monotonia decrescente dos contextos relevantes.

PALAVRAS-CHAVE: Palavras-N, Concordância Negativa, Indefinidos, Monotonia Decrescente
The relevant facts in (1)-( 3) and its congruent expansions can be summarized in the following terms: (i) when in pre-verbal position, an n-word cannot, on one side, co-occur with an overt (say, propositional) negation operator, and, on the other side, does induce per se a negative value (to be specified later) in the sentence, as is clearly visible in formulas (1c) and (3c); (ii) when in post-verbal position, an n-word requires the presence of an overt negation operator acting as a licenser, which consequently suggests that post-verbal n-words are negative polarity items (somehow misleadingly, since, as is well-known, n-words can be licensed in other kinds of contexts); (iii) as shown in (3a), the negative licenser of a post-verbal n-word can be a pre-verbal n-word, whose presence precludes the emergence of an overt propositional negation.It should be noticed that the property in (ii) is nothing but the process of negative concord, as characterized in Labov 1972. Indeed, what is at issue there is the fact that an expression that in a pre-verbal position has the capacity to convey a negative value can also appear in a configuration where its negative force can be dispensed with, given the presence of the propositional negation operator.This peculiar arrangement raises the idea of "concord": with some similarity with gender and number concord, where just one marker is vital, the others being mere redundant replicas, in the sequences with a negation operator licensing an n-word only one negation value is needed, and, much more importantly, only one can be computed.
In general, the domain of n-words is extended in the literature to languages like French, Greek, Hungarian or the Slavic languages, where the pre-verbal occurrences of the expressions at stake require, just like the post-verbal ones, the presence of an overt negation operator (which amounts to, grosso modo, saying that, in the structures corresponding to (1a) above, the asterisk would lie outside the parentheses).Obviously, the pertinence of such an extension crucially depends on a clear definition of n-word.Furthermore, it can also be argued -e.g. in Acquaviva 1993(according to Giannakidou 1997) and in Peres 1995Peres /1997 -that a unified cross-linguistic account of the relevant expressions can be formulated, which encompasses the standard English sort of languages.
In the next section, I will try to sketch a hopefully convincing concept of n-phrase.Before engaging in that, let me settle a minor, but not negligible, terminological matter: henceforth, I will use the label nphrase instead of n-word, considering that the former seems much more in accordance than the latter with the internal composition of the expressions under scrutiny.

The indefinite theory of n-phrases
The exact semantic and syntactic nature of n-phrases has been the object of considerable and divergent discussion in the literature.If we restrict ourselves only to the languages of the predominant Romance type, the different theories on the categorial status of n-phrases can be divided in three groups.In a first cluster of theories (including, among others, those sketched in Rizzi 1982, Longobardi 1987and Dowty 1994), n-words are treated as NQ's or indefinites, depending on syntactic properties in the first two authors and on a computation of semantic values of monotonicity in the latter.In a second group (see, e.g.Laka 1990, Ladusaw 1992, Zanuttini 1994, Suñer 1995, and Peres 1995/1997), n-words are systematically treated as indefinites, or, from another point of view, as existentially quantified expressions.Finally, in a third group (see e.g.Zanuttini 1991 or Haegeman and Zanuttini 1991), n-words are always treated as NQ's whose negative value can be canceled.In the present section, focusing on Portuguese data, I will aim at showing that the indefinite theory of n-phrases, besides offering a general and economic account of the constructions at stake, explains a larger number of facts than other theories.
Ladusaw's basic idea of an indefinite theory of n-phrases, which I will subsequently argue for, can be summarized in the following terms: "in a NC language in which only one of these expressions [-negative terms -] can express negation in a particular clause, the way is open for proposing that the negative phrases in fact never express negation.In fact, we could propose that they are univocally interpreted as NPI indefinites and that it is not necessary that any visible formative of Sstructure actually express negation" (Ladusaw 1992: 251).The answer to the question "how does the negation get expressed and how are these polarity items licensed" is phrased as follows: "The answer (...) must be (...): a negation operator, preferably (anti-morphic) negation.(...) It need not be part of lexical meaning: it may be constructional, in the sense that it is associated with some structural feature not necessarily visible in the clause" (cf.ib.: 251-2).In Peres 1995/1997, I proposed an extension of this view of n-phrases to languages which, like standard English, only obey the first of the above mentioned two properties, that is, which lack negative concord.

Arguments in favor of the indefinite theory of n-phrases
In this subsection, I will bring together several arguments in favor of treatments of n-phrases that are basically equivalent to the indefinite theory.Some of these arguments appear in Klima 1964, some in Peres 1995/1977, Peres 1977and Peres 1978, and a few others are, to my knowledge, now presented for the first time.As for the arguments contra an indefinite theory -or, for that matter, in favor of the (total or partial) negative quantifier approach -, having, I believe, shown the immateriality of the most important of them in my 1995/1997 paper, I should now only pay attention to the absolutely argument.However, for absolute time and space constraints, I will not address this issue, just suggesting the idea that a treatment of that sort of expressions as discourse operators on propositions -not on noun phrases -can lead to showing that also this argument fails to prove that n-phrases are not indefinites. 1

The paradigms of indefinite n-phrases
The paradigmatic relations of n-phrases constitute a vantage point for understanding their nature, which, in my view, strongly favors the indefinite analysis, and which, to my knowledge, has not been paid noticeable attention.As we shall promptly see, the relevant paradigms are particularly rich and revealing in a language like Portuguese.To start with, let us observe a basic instance of an n-phrase involved in negative concord: For the sake of simplicity, I am ignoring the possible so-called (nonspecific) "generic" (preferably, "kind") interpretation of a sequence like (4a), where a natural kind is involved more directly than common individuals.Such reading is more easily available if certain factors are present, like a generic aspectual value or a contrastive context: 1 The same sort of approach may well be able to overcome difficulties raised by the test with virtually that Jack Hoeksema proposed during the Salford conference on negation, in November 1998.I thank him for having subsequently given me further details on this test, that I will not be able to discuss here.
( If we skip this interpretation and stick to the existential one -which is entailed by the former -, we can say that the sentence in (7a) below, also in its existential interpretation (specific -in the sense of "somehow known to the speaker" -or not, it is now irrelevant), is the contradictory of the existential reading of (4a): The interesting fact to be noticed at this point is that (4a) is only one of a bunch of (grosso modo) free variants that express -by intrasentential means, not by some logically legitimate expansion of formulas -the contradictory of (7a).In Portuguese, one of these variants is obtained by assigning the relevant constituent (in the case under scrutiny, an object NP) a prosodic focus, as symbolized by the capital letters in (8): (8) O Paulo não leu UM livro (que fosse).
'the Paulo not read a book (that be SUBJ )' Paulo didn't read any book (whatsoever).
As for the prosodically more neutral form, represented in (9) below, it only has a wide scope (that is, specific, in another sense) reading for the quantifier, and therefore isn't a variant of either (4a) or ( 8 Interestingly -not to say surprisingly -enough, the paradigm in (11) is still incomplete.In fact, the formula in (4b) can be expressed by all the sentences below (possibly with some discourse, not semantic proper, subtleties involved), where the subjunctive appendix is now ignored, for the sake of simplicity: If we take into account the crucial correspondence between, on one side, sentence (7a) and its meaning translation in formula (7b), ∃x [book (x) !read (p, x)], and, on the other side, all the sentences in (12) and their meaning translation in formula (4b), ¬∃x [book (x) !read (p, x)], the contradictory of (7b), the conclusion can be drawn that, in a language like Portuguese and at least in the syntactic context at issue (namely, under the scope of sentential negation), non-specific existential indefinites (NSEIs, for short) must be subject to a particular (prosodic or morpho-syntactic) formatting -which results in the distinguished forms (12a-f) -and are in free variation with a qualquer(any)-type NP, as shown in (12g).In this perspective, and if only the context presented above is considered, the n-phrases under discussion can be taken as just one of the forms that non-specific existential indefinites can assume in such context.
The behavior of existential indefinites I have been describing in connection with a post-verbal object position under the (syntactic and semantic) scope of negation grosso modo generalizes to every post-verbal position.In other words, in languages like Portuguese, paradigms like the one given in (12) above are in general valid for any post-verbal argument or adverbial position within a given syntactic domain (and, under certain conditions, not to be specified now, even across sentential boundaries).Nevertheless, in some cases there may arise some blocking effects that prevent all the expressions at stake to be free variants of one another, even in the negative configurations we have been considering.For the sake of simplicity, I will ignore here such possible effects.As for pre-verbal positions, the relevant facts are disclosed by the following pattern, where the asterisks mark either ungrammaticality or the unavailability of the intended meaning of negation of existence: Regardless of the idiosyncratic wealth of variants that ( 14) reveals, the general syntactic pattern, namely as exhibited in (14b), is a wellknown one, which consists of a typical form of n-phrase occurring preverbally.It is common to the traditionally called "negative concord languages", but also to languages like standard English, usually classified as "double negation languages".We will return to this distinction in section 2.3 below.
In my view, this is the right ambiance for an initial approach to the n-phrases we are now dealing with, insofar as the observed paradigms constitute strong evidence that such expressions belong in the class of non-specific existential indefinites.In the next subsection, further evidence in this direction will be adduced.

Klima's (1964) tests
The first battery of tests that suggests itself as a tool for evaluating the negative character of sentences with pre-verbal n-phrases is devised in Klima (1964).All such tests involve subtypes of elliptic constructions that seem to require the presence, in the first member of the dyadic structure, of some sort of negative operator.Putting things in these terms, I am implying not only the old and trivial claim that different sorts of negation are available in a language, but also that, as should be expected, the differences in attachment and scope of negative operators bear upon the viability of certain constructions.However, contrary to my previous contentions on this matter (cf.Peres 1995Peres /1997)), I now believe that, while Klima's tests indicate that some kind of negative value (possibly just decreasing monotonicity in one or more cases) is at work in the structure, it is not yet clear which.I will not further elaborate on this point, merely suggesting for now, with data presented in ( 15)-( 17), that different levels of negative values may license the constructions at stake: (15) a. Os estudantes leram o livro, *(não) leram?
The The students didn't read the book, not even Rui.c. Nenhum estudante leu o livro, nem mesmo o Rui.
Few students read the book, not even Rui ?(did).
'the Vera not read the book and the Rui also *(not)' Vera didn't read the book, neither did Rui.Vera didn't read the book, and Rui didn't either.c.Nenhum estudante leu o livro e o Rui também *(não).
'no student read the book and the Rui also * (not)' No student read the book, neither did Rui / and Rui didn't either.d.Poucos estudantes leram o livro e o professor também não ?(leu).
Few students read the book, and the teacher didn't ?(read it) either.

Arguments in favor of the indefinite Further theory of n-phrases
The arguments in favor of the indefinite theory of n-phrases that will now be put forward can be divided into two groups: those that strengthen the idea that a constructional (arguably, contradictory) negation is available in the relevant structures, which are grouped below under (i)-(iii), and those that confirm the (existential) indefinite character of n-phrases, which constitute the group (iv)-(vi).I will keep on skipping the indispensable syntactic characterization of the relevant negation operator that is at issue in the present discussion, which, in very vague terms, I conceive of as lying between constituent negation and a fully external propositional negation: "a mode of predication, a recipe for combining subject (...) and predicate (...) to form a proposition or sentence (...), rather than an operation on a fully formed proposition or sentence" (cf.Horn 1989: 469).

(i) Other cases of sensitivity to negation
The constructions given in ( 18)-( 20) point in a more distinct direction than Klima's tests.Since they will not be described in detail, suffice it to say that they all can be considered as composites of negative propositional components.
(18) a. Durante este depoimento, os fotógrafos NÃO poderão ficar na sala NEM as câmaras de televisão poderão filmar.During this deposition, the photographers will be allowed to remain in the room nor will the TV cameras be allowed to shoot.
[cf.Peres 1995Peres /1997: 292] : 292] During this deposition, no photographer will be allowed to remain in the room nor will the TV cameras be allowed to shoot.
( (20) a. Eu NÃO disse que estava a chover, (mas) APENAS que podia chover.'(I) not said that was at rain, (but) just that could rain' I didn't say that it was raining, just that it could rain.
b. NINGUÉM disse que estava a chover, (mas) APENAS que podia chover.'nobody said that was at rain, (but) just that could rain' Nobody said that it was raining, just that it could rain.
(ii) Anaphoric do the same constructions The third source of evidence in favor of the postulation of a negative operator concerns the anaphoric do the same construction (cf.Peres 1995Peres /1997: 292): 292), whose interpretation in the relevant cases must include a negation operator.'none of-the my colleagues accepted the invitation and I go do the same' None of my colleagues has accepted the invitation, and I will do the same.
(iii) Structures with pre-verbal n-phrases licensed by an overt negation operator The next piece of evidence (again, extracted from Peres 1995/1997) concerns the occurrence of n-phrases in pre-verbal position even when immediately preceded by an overt sentential negation operator (sem / without) -cf.(22a) below -, which also licenses post-verbal realizations of the same items -cf.(22b).Furthermore, it is shown that when such an operator is neutralized by genuine (adjacent) double negation, yielding a positive context, the licensing is blocked -cf.(23b We now move to the evidence in favor of the analysis of n-phrases as indefinites.

(iv) Anaphoras of n-phrases
The data in this item show that empty constituents or even pronouns whose interpretation is dependent on n-phrases can only receive an existential reading (cf.Peres 1997: 2).
"more nobody spoke" b.Plus personne n'a parlé.
"more nobody not has spoken" c.No one else spoke.

(vi) in particular-constructions
This construction has grosso modo the same restrictions as the previous one (cf . Peres 1997: 2).In fact, the adjunct in particular can only be computed in combination with an expression conveying an existential value.
(27) They chat about nothing in particular.
(28) Are you looking for someone in particular?No, I am looking for no one in particular.
From all these facts together, the reasonable conclusion to be drawn seems to be that, in languages like Portuguese, both pre-and post-verbal n-phrases are, at least in the kind of contexts examined so far, nothing but NSEIs under the scope of (not necessarily visible) negation.When pre-verbal, they are negation inducers which, in an abstract syntax, may be considered to have undergone a process of agreement with a negative head towards which they act as specifiers (cf., e.g., Zanuttini 1991).When post-verbal, they enter a negative concord process with a visible negation operator, which amounts to saying that they keep the same form as when they are negation inducers, but in reality do not perform such role, but only that of NSEIs.I believe that it was abundantly evidenced that only under the categorial assignment and constructional postulation just discussed can the right semantics be obtained in all the relevant cases, namely in the distinguished structures that were presented above as a diagnostic for the existentiality and indefiniteness of n-phrases.In its general features, this is a view that was proposed in Ladusaw 1992 for so-called negative concord languages and generalized to so-called double negation languages, like standard English, in Peres 1995/1997.

A concept of n-phrase and some typological matters
From the evidence presented so far, the notion of n-phrase, as applying to the Portuguese expressions we have been considering, can be associated with the following features: (i) the marked expression of non-specific existential indefiniteness in special contexts, namely in those that can be characterized as involving the "mode of predication" sort of negation; (ii) the capacity these marked NSEIs have, when located in at least one distinguished syntactic position (in the language under consideration, a pre-verbal position) and in the absence of any visible sentential negation operator, to make their clausal domain exhibit certain properties that are exclusive of the subtype of negative constructions mentioned in the previous item; (iii) the restraint to the mere role of NSEIs -without any conveyance of a negation value -, when in the appropriate dependency vis-à-vis another appropriate conveyor of negation (in short, the engagement in what Labov termed "negative concord").Now, it is just a terminological matter to decide whether or not for an expression to be classified as an n-phrase all the above features are required.My position is that the notion will gain in generality and expressive power if only the first two properties are summoned into its definition.In fact, given such a comprehensive outlining, not only languages like Italian, Portuguese and Spanish but also languages like English are endowed with n-phrases, a claim that allows a unified treatment of the pre-verbal behavior of the relevant expressions.Clearly, this was (avant la lettre in what concerns the labels n-word/n-phrase and related concepts) the view adopted in Klima 1964, where such expressions were derived in English by means of a rule of "incorporation" of (an abstract) negation (constituent) -"neg-incorporation". Sticking to Klima's terminology, all the languages in the wider group just mentioned are neg-incorporation languages.Without further speculation, it must be stressed that, under the definition of n-phrases that I have just adopted, languages like Greek, Hungarian and the Slavic languages -and even European French, although somehow more arguably, given that the omnipresent negation operator is a weak form (as opposed to pas) -appear not to be neg-incorporation languages.The obvious entailment is that the latter are not negative concord languages either.In fact, if the expression "negative concord" is to be taken at its face value, as it should, then an expression can be involved in negative concord if and only if it can also be by itself a conveyor of negation, which, in the cases under analysis, requires that it undergoes a process of the neg-incorporation type.From the above enumerated negincorporation languages, Italian, Portuguese and Spanish are genuine NC languages, while English is not 3 .As for French, Greek, Hungarian and the Slavic languages, all that can rigorously be said is that they possibly have marked NSEIs, namely for negative contexts, but not that they admit negative concord.
From this standpoint, the typological question can be further extended to the following issue: is the markedness of NSEIs in distinguished negative contexts a universal feature of languages or not?According to Bernini and Ramat 1996 the answer is negative, considering the behavior of Basque, two Celtic languages -Irish and Gaelic -, and three from the Finno-Ugric family -Lapp, Finnish and Estonian (cf.p. 182).Furthermore, if the evidence in (29) -from Evenki ("one of the ten Manchu-Tungusic languages", spoken in Siberia and Northern China see Nedyalkov 1994: 1) -truly corresponds to a general pattern, also this language belongs in the same group, the same being the case with Zazaki ("a language (or network of dialects) spoken in the region of the head waters of the Euphrates river in east-central Anatolia" see pp. 125 ff.), following the description found in Sandonato 1994.[cf.Nedyalkov 1994: 25] 2

. Additional instances of n-phrases and/or Negative Concord
In general, the work on negative incorporation and negative concord -or, briefly, on n-phrases -has focused on n-indefinites, notwithstanding the fact that Klima (1964) had already stressed the parallel between what have come to be known as n-words and other kinds of phrases, namely the neither... nor sequences.In the present section, I will stress that the field where these processes take place is much wider than the domain of indefinites.Subsequently, I will briefly list what I think to be the types of Portuguese structures that involve such processes, without attempting to establish a rigorous characterization of the material.Pending further research, suffice it to say for the moment that the data below may contain cases of n-phrases of the type we have been considering so far -that is, supporting both negative incorporation and negative concord -, but also cases where, the "mode of predication" negation not being involved, negative incorporation has a different morpho-syntactic and semantic architecture, although negative concord may survive as the semantically void concord process we know from the foregoing.Besides Portuguese (PT), occasionally examples are given for some other Romance languages.
[PT] 'not-even twenty students went to-the party' a'.NÃO

foram à festa (NEM) vinte estudantes. [ambiguous without nem]
'not went to-the party not-even twenty students' b.Not even twenty students attended the party.
When definites are involved, the same facts occur, with an additional implicature value: (31) a. NEM ao Domingo descanso.
[IT] 'not-even the Sunday me rest' b'.NON

Back to n-indefinites and their licensing conditions
In many languages that have n-phrases, these, namely in the indefinite variety, can occur in contexts where a negation value can only be obtained via indirect computations (for instance, interrogatives, comparatives or the antecedent of conditionals).As I tried to show in Peres (1998), Portuguese, contrary to other Romance languages, is in this respect a very restrictive language, which almost exclusively allows indefinite n-phrases in contexts where a basic negation operator is appropriately available.In the present section, I will only take into account this kind of licensing, both in single sentential domains and cross-sententially.

Intra-sentential anti-veridical licensing
All the occurrences of n-phrases discussed so far took place in a context where overt or covert sentential negation was present (in the second case, once the indefinite theory of n-phrases is assumed).Using the terms of Giannakidou (1997), this amounts to saying that up to now we have only considered the licensing of n-phrases in "averidical contexts" (or in the author's 1998 revised terminology, "anti-veridical"), according to the following definition: (38) Let OP be a monadic sentential operator.The following statements hold: (i) OP is veridical just in case OP(p) → p is logically valid.Otherwise, OP is nonveridical.
(ii) A nonveridical operator OP is anti-veridical just in case OP(p) →¬ p is logically valid.[cf.Giannakidou 1998: 106] Besides the common sentential negation operator, another operator that can be considered anti-veridical is sem (and its equivalents in other languages, like English without).Accordingly, such operator licenses n-phrases (that is, it gives rise to negative concord), as shown by the following data4 : As shown by Zwarts (1995) and Giannakidou (1997), in some languages the anti-veridical version of operators like temporal before license n-phrases and other items that are sensitive to anti-veridicality.French and Spanish examples are: (40) Partez avant que personne vous voie.[Grevisse-Goosse 1993: 1461] 'leave before that nobody you see' Leave, before anyone sees you.

Cross-sentential licensing
Developing the trend open by Ladusaw 1979, where decreasing monotone contexts are the crucial factor in the licensing of negative polarity items, Dowty 1994 propositions the idea that, in so-called negative concord languages, these items, including n-phrases, can only occur in positions that are subject to that sort of inferences, which they are aimed at marking.He re-elaborates the logic of monotonicity defined in Valencia (1991) as a tool for calculating monotonicity, and, as mentioned before, adopting an ambiguity (between NPI's and NQ's) treatment of n-phrases, he evaluated his theory in the domain of simple sentences.In Peres 1994Peres and 1995Peres /1997, I tried to reach the following two goals with respect to the licensing of n-words: (i) showing that, in a language like Portuguese, Dowty's monotonicity effects were preserved across clause boundaries, that is, in cases where NPI's, n-words included, are located within an embedded clause, the licenser being in the matrix; (ii) proving that the indefinite categorization of n-words and the invisible negation approach I was adopting, while being strictly opposed to Dowty's ambiguity treatment of n-words, was fully compatible with his basic intuition concerning the role of monotonicity not only of negation in the licensing of n-words.In face of the limited amount of data I analyzed at the time, the results seemed to confirm Dowty's hypothesis rather straightforwardly.Crucial cases were sentences like (45) a. NÃO me surpreende que NINGUÉM tenha telefonado à Maria.
'not me surprises that nobody has called the Maria' ≠ It doesn't surprise me that someone has called Maria.
'not me surprises that the Maria has called to none of-the colleagues' It doesn't surprise me that Maria has called any of her colleagues.
(46) a. O ministro NÃO exigiu que NINGUÉM fosse preso."the minister not required that nobody be arrested" The minister did not require that nobody / anybody be arrested.
b. O ministro NÃO exigiu que a Polícia prendesse NINGUÉM."the minister not required that the Police arrested nobody" The minister did not require that the Police arrested anyone.
Laterally, it must be pointed out that the classification of a verb like surpreender (surprise) as downward entailing, which appears recurrently in the literature, is not accurate, at least if the verb is taken to be veridical (or, more strongly, factive).In fact, as Zwarts (1995) has proven, an operator cannot be both veridical and downward entailing with respect to the same argument position.At any rate, the more accurate characterization of the verb as nonmonotone with respect to its sentential argument does not preclude the final value obtained under negation to be increasing monotonicity, for which reason no ambiguity arises with the pre-verbal instance.Concerning (46), where the inference is decreasing -as a result of the neutralization of the increasing value of the verb by the decreasing value of negation -, such ambiguity surfaces.However, this neutralization does not take place in every case, namely when factivity is present, leading to the conclusion that the negation of an increasing monotone predicate of force does not preserve the direction of the inference, but does not necessarily reverse it.This is the case with the next example, which contains a higher verb that (arguably, leaving apart the vexing controversy on the closure of the complements of positive epistemic predicates under logical consequence) is increasing monotone regarding its complement, and where nonmonotonicity is obtained after the application of negation.In such context, n-phrases are not licensed cross-sententially.
'I not knew that here nobody speaks Chinese' ≠ I didn't know that somebody here speaks Chinese.
'not (I) knew that you spoke no Asian language' The hypothesis I raised at the time in order to account for the kind of discrepancies I then noticed was that at least one modal feature of verbs -EPISTEMIC -plays a crucial role in the reading selection.The general picture is given in Table 1 (from Peres 1994).

Table 1. Cross-sentential licensing of n-phrases in Peres 1994
In defiance of the apparent regularity of this general picture, some problems remained unsolved, namely cases, which I then overlooked, with MON↓ predicates, trivially yielding upward entailment under negation [(-, -) / +], but (at least to some degree) licensing n-phrases: (48) a. Eu NUNCA EVITO que ninguém me critique.
'I never avoid that nobody me criticizes' ≡ I don't want to prevent anyone from criticizing me.
b.A Ana NÃO EVITA que os colegas leiam NENHUM dos seus poemas.'Ana not avoids that the colleagues read SUBJ none of her poems' ≡ Ana does not prevent her colleagues from reading any of her poems.
'the Pedro not avoids talk with nobody' Pedro doesn't avoid talking with anyone.
(50) a. O presidente da sessão NÃO IMPEDIRÁ que NINGUÉM se pronuncie.'the chairman of-the session not prevent FUT that nobody himself intervene' ?≡The chairman of the session will not prevent anyone from intervening.
'the chairman of-the session not prevent FUT that one talk about no subject' ?≡The chairman will not prevent any subject from being addressed.
Bearing in mind the fact that Giannakidou 1997 found a correspondence between nonveridicality and the licensing of (what she considers to be) n-phrases in Greek, the attempt to compute this value instead of decreasing monotonicity seems to be a reasonable trend of research.The prima facie evidence is that all the above cases are now solved in a simpler way (assuming a non-standard computation of monotonicity that I will not elaborate on): However, a minute scrutiny of further data suggests that the domain resists simple solutions, more factors being involved than those considered both in Table 1 and in Giannakidou's approach.In fact, not only it cannot be said that embedded nonveridical contexts constitute a sufficient condition for the licensing of n-phrases, but also the veridical ones are not regular in respect to the expected anti-licensing effect.The first relation, between nonveridicality and licensing, can be abundantly illustrated.The following data portray embedded nonveridical contexts which are also decreasing monotone due to combinations of negation and upward entailing predicates in the relevant argument.It should be noticed that henceforth no mention will be made of the availability of the reading in which (within a framework that incorporates the indefinite approach) the pre-verbal n-phrases are being licensed in the lowest clausal domain, since it is irrelevant for the purpose of the argument.
'not me remember of have written nothing about that subject' ≡ I don't remember having written anything about that subject.
A clear weakening of the likelihood of licensing originates when nonveridicality combines with the nonmonotonicity of declarative verbs and their kin (for instance, prove or demonstrate) regarding their complement: O pianista NÃO gostou que NINGUÉM ocupasse as filas da frente.'the pianist not liked that nobody occupied SUBJ the front rows' ≠ The pianist didn't like that (??)anybody occupied the front rows.
The pianist didn't like that nobody occupied the front rows.
'the Maria not goes say that nobody her helped' ??≡ Maria is not going to say that someone helped her.
b. ??A Maria NÃO VAI DIZER que ajudou NINGUÉM.'the Maria not goes say that (she) helped nobody' ??≡ Maria is not going to say that she helped someone.
The same, even more dramatically, is the case when nonveridicality is associated with increasing monotonicity, as was the case in ( 48) and ( 50), and, in addition, the higher verb can be considered epistemic (even though with some alethic flavor to it): (56) a. NÃO É IMPOSSÍVEL que NINGUÉM fale sobre a questão.
'not is impossible that nobody speaks about the issue' ≠ It is not impossible that anyone speaks about the issue.
'not is impossible nobody speak about the issue' ≠ It is not impossible that anyone speaks about the issue.
'not is impossible the Ana speak with nobody' By the way, notice that in metalinguistic realizations of negative sentences, the licensing of n-phrases is almost invariably admitted.For instance, sentence (55a) above is perfectly sound with cross-sentential licensing -that is, with the meaning corresponding to the English translation in (55) -if it is used metalinguistically, as in the following dialogue: (58) Sp 1 A Ana vai dizer que alguém a ajudou.
'the Ana goes say that someone her helped' Ana is going to say that someone helped her.
Sp 2 A Ana NÃO dirá que NINGUÉM a ajudou: é demasiado orgulhosa para isso.'Ana not say FUT that nobody her helped: (she) is too proud for that' Ana will not say that someone helped her: she is too proud to do that.
In ( 61)-( 62), we have an epistemic and a declarative predicate that can be considered decreasing monotone in its internal argument: (61) a. NÃO DUVIDO de que NINGUÉM votará a favor da proposta A.
'not (I) doubt of that nobody will-vote to favor of-the proposal A' ≠ I don't doubt that someone will vote for proposal A.
'the Maria not regretted that nobody her had passed' ≠ Maria didn't regret that someone had passed her.
b.A Maria NÃO LAMENTA que o Pedro tenha escrito NENHUM destes artigos.'the Maria not regrets that the Pedro has written none of-these articles' ?≡Maria doesn't regret that Pedro has written any of these articles.
(66) A Maria NÃO LAMENTA ter escrito NENHUM destes artigos.'the Maria not regrets have written none of-these articles' ≡ Maria doesn't regret having written any of these articles.
The whole array of features and values considered in the observed examples is given in Table 2, where the monotonicity values are the result of combining the downward monoto-nicity of the negation operator with the value the predicate exhibits and, additionally, considering factivity.The reading of Table 2 is quite linear, with three facts about Portuguese imposing themselves: (i) only decreasing monotonicity of the embedding position under negation is a sufficient (but not necessary) condition for the cross-sentential licensing of n-phrases; (ii) nonveridicality is neither a sufficient nor a necessary condition to that effect; (iii) n-phrases are not anti-licensed by veridicality; (iv) under increasing monotonicity, the licensing factors consist of rather restrained combinations of values in semantic and syntactic features (namely, in the latter domain, nonfiniteness of the embedded clause and the postverbal position of the affected indefinite).
Table 3 constitutes one of several possible compressed views of the information contained in Table 2.The shaded areas express the apparent insignificance of the values at stake.It should be noticed that the cluster of values "OK/?" can easily be split, as Table 2 clearly shows.

Table 3. A condensed view of the semantic and syntactic relevant conditions for the licensing of n-phrases according to the examined data
Let me try a brief and informal account of the findings we have been discussing in this section.In the first place, it seems to be the case that, in consonance with a basic idea found in Dowty 1994, which was later explored in my 1994 and 1995/1997 papers, in a language like Portuguese the computation of monotonicity values plays a prominent role in the cross-sentential licensing of n-phrases, for which decreasing monotonicity is a sufficient condition.In the second place, it seems that such kind of language is crucially sensitive to the modal character of the higher predicate, namely to whether it has an epistemic or declarative component.In a manner that demands explanation well beyond the limits of this study, these modal values tend to block the licensing (or, in other words, the concord) process.Interestingly, as Marques 1995 revealed, the same modal values appear to be anti-licensers of the subjunctive mood in languages like French, Italian, Portuguese and Spanish, contrary to what happens in languages like Greek, Hungarian and Rumanian, where the same mood value allegedly is anti-licensed by veridicality.
In merely intuitive terms, what appears to be the case in the licensing of n-phrases in languages like Portuguese is that, in the absence of a value of decreasing monotonicity in the embedding context, with which the indefinite would be prepared to engage in a process of (negative) concord (of downward monotonicities) -which ultimately is what the licensing at stake amounts to -, it is nonveridicality that, by inducing a virtual negative value in the lower sentence (say, to the extent that no assurance is given about its truth), as it were supplies a (semantic) concordant element for the negative concord process.However, this operation can only take place given certain modal properties of the matrix predicate.On the other side, the matrix negation can by itself, even in the absence of nonveridicality in the complement clause, be responsible for the licensing effect, provided that the syntactic conditions already discussed above apply.
A second point that must be stressed is that the above dependencies reinforce another of Dowty's basic intuitions, namely that n-phrases are markers of downward entailing positions.Indeed, in all the data under scrutiny that involve n-phrase licensing, the position of the indefinite is invariably a downward entailing position.The crucial licensing examples to be checked are those where the monotonicity computation à la Dowty yields a positive value, which is the case in ( 48)-( 50).I will exemplify this verification with one of the relevant predicates: (67) O Pedro não evita falar com nenhum colega.
'the Pedro not avoids talk with no colleague' Pedro doesn't avoid talking with any colleague.→ O Pedro não evita falar com nenhum colega mais novo.
'the Pedro not avoids talk with no colleague more young' Pedro doesn't avoid talking with any younger colleague.
What, in my view, this entailment reveals is that, regardless of the computation of mono-tonicity in terms of Valencia-Dowty's calculus, down to the position of clause embedding (resulting from combining the values of the negative operator and that of the higher predicate), the increasing monotonicity of the indefinite can be reversed by the matrix clause negation operator, if the (preferably post-verbal) position of the indefinite becomes transparent to that operator, by a sort of clause-union.Nevertheless, as we have seen, the possibility exists that a post-verbal indefinite in a finite clause be accessible to the negation in the matrix.Regarding this latter case, although one can intuitively perceive that it makes all sense that such access is allowed only under a non-epistemic verb in the matrix, as is the case in (65b), the reasons why this might be so will have to be the object of more extended and sophisticated investigation.

Conclusion
The main focus of this paper was the nature of n-phrases and their licensing conditions in Portuguese.In the first two sections, I tried to reach a clear definition of this class of expressions and of the related notions of "incorporation of negation" and "negative concord".The general purpose was to show that a theory like Ladusaw's indefinite theory of n-phrases fits the facts about these expressions, namely in Portuguese, better than concurrent theories.Working in this direction, I discussed several tests for indefiniteness.Moreover, close attention was paid to the paradigmatic relations of n-phrases, which were taken as a major source of evidence for a categorial settlement.In section 3, data were presented that clearly show that the characteristics of n-phrases that have to do with negation are manifest in several constructions of Portuguese and other languages which cannot be considered as simple noun or adverbial phrases with indefinite value.In the final part of the paper, I aimed at showing that the cross-sentential licensing of n-phrases in Portuguese -and presumably in several other languages -is subject to a rather well-defined network of semantic and syntactic properties, where, in partial confirmation of Dowty's 1994 hypothesis, the role of decreasing monotonicity is of paramount importance.
19) a. Espero que você NÃO se sinta aqui mal, mas SIM como na sua '(I) hope that nobody not himself feel here badly, but yes as in-the his own home' I hope nobody feels uncomfortable here, but instead as if being at home.
'no law forbids the-what [we] are at do and if [ec] did [ec] should be revoked' No law forbids what we are doing, and if (?it/one) did it should be revoked.(25) Não te peço nenhum livro emprestado, porque não sei quando to poderia devolver.'[I] not you ask no book borrowed, because [I] not know when you-it could return' I won't borrow any book from you, because I don't know when I would be able to return it.
mi riposo NEANCHE la domenica.".Eu NÃO conto tudo o que sei, NEM que me matem.I won't tell everything I know, even if they kill me.b.Ni aunque me maten contaré todo lo que se.[SP; same ambiguity] 'not me rest not-even the Sunday' c.I can't rest, even on Sundays.a Ele NÃO NEGOU que NINGUÉM o tinha autorizado a entrar.'henot denied that nobody him had authorized at get-in' ≠ He didn't deny that someone had authorized him to get in.b. *Ele NÃO NEGOU que tinha autorizado NINGUÉM a entrar.'henot denied that had authorized nobody to get-in' ≠ He didn't deny that he had authorized someone to get in.