Negation in Catalan : some Remarks with regard to no pas

NEGATION IN CATALAN. SOME REMARKS WITH REGARD TO NO PAS M. Teresa Espinal Universitat Autbnoma de Barcelona In this paper I shail approach certain questions that can be posed in the study of Catalan negative constructions: (a) how many negative markers can be distinguished; (b) how the difference between no and no pas should be described in the core grammar of this language, and (c) what sort of syntactic representation should be assigned to negative sententiai constructions in Catalan. The structure of the paper is organized as follows. Section 1 is devoted to describing the syntactic disttibution of the various negative markers existing in Catalan. It is proposed that, apart from no, there are two more negative markers: no pas I, which is a negative head, and no pas 2, which is a negative specifier. Section 2 describes the semantic properties that characterize these negative markers. It is argued that the association with focus is not an inherent property of the logical semantics of no pas, and that what singles out this operator, in contrast with no, is that it has a quantificational domain over a set of conceptual entities of which a subdomain is selected as the negative contrasted item. Section 3 deals with the sort of syntactic structure that should be assigned to Catalan negative sentences and it iniroduces the proposal that the relevant structure has a NegP as the root node and an AgrP as iis complement. TP is the complement of the latter functional category. 1. Syntactic Distribution I shall approach the first question by briefly sketching out the basic syntactic distribution of the simple lexical item no and the compound lexical item no pas in Central Catalan.1 A close examination of the order in which negative structures are internalized in the process of first language acquisition and an inquiry into the frequency of appearance of the various existing Catalan Working Papers in Linguistics (CWPL), 1991: 33-63 Universitat Authnoma de Barcelona negative structures suggest that it is convenient to distinguish within the grammar of this language between marked and unmarked negative constructions, with the syntactic distribution shown in (1). UNMARKED MARKED no no V no X UNMARKED no pas MARKED I I no pas I: no pas 2: Nego SPEC [ + NEG] no pas X El It is important to notice that V can be kFIN], and V* is either a V or a complex verb. On the other hand, X cannot be a finite verb form.2 Some examples which illustrate this syntactic distribution are given in (2). (2) a. La Maria no va aprovar the Maria not PAST pass 'Maria did not pass' b. *La Maria va no aprovar c. No tots els estudiants van aprovar not all the students PAST pass 'Not all the students passed' d. La Maria no va pas aprovar the Maria not PAST pas pass 'Maria did not pass' (cancelling or confirming the speaker's expectations)3 e. La Maria no va aprovar pas (2) f. *La Maria va pas aprovar g. *LaMariavaaprovarpas h. No pas LA MARIA no va aprovar not pas the Maria not PAST pass 'It was not Maria who passed' (cancelling or confinning the speaker's expectations on Maria) i. *No pensava que la Maria aprovaria pas not think-PAST that the Maria pass-COND pas 'I didn't think that Maria would pass' j. *No temia poder pas aprovar la Maria not fear-PAST be able pas pass the Maria 'I had no fear of being able to pass Maria' k. No crec que aprovi pas la Maria not think that pass-SUB pas the Maria 'I don't think that { Maria is going to pass, I'm going to pass Maria )' Exarnples (2a,b,c) show the distribution of the simple lexical item no, either preceding a verb forrn or a non-verbal item. However, in this paper I am more concerned with negative no pas constructions. Notice that in (2) no pas appears in two different syntactic environments (compare (2d,e) with (2h)). With regard to this fact two distinct negative markers will be postulated. No pas 1, as illustrated in (2d,e), is characterized by coappearing with a verb form (either simple or complex) and by being a discontinuous lexical item. Note also that in the dialect that I describe, pas cannot occur alone, without no (see examples (2f,g)), which suggests that -unlike French and Occitanin current Central Catalanpas is not an independent lexical unit, but a morpheme which is lexically and syntactically inseparable from no. Furthermore, the ungrammatical sequences given in (2ij) reveal that the series of verbs that interrupt the compound lexical item no pas 1 cannot contain a sentential boundary. In this respect, the grammaticality of an exarnple such as (2k) seems to be related to the tense-mood of the subordinate clause. According to Picallo (1984), the Tense node of a subjunctive clause lacks the features necessary to be fully specified as [+ TNS], for it is bound to the Tense specification of the main clause. On the other hand, pas at S-structure must always be c-commanded by no (cf. May (1985)), which means that there cannot be any maximal projection between no and pas . Therefore, in the syntactic context illustrated in (2k) it seems as if the CP node of the subordinate clause in the subjunctive tense is not visible with regard to binding effects. One possible interpretation of these data is to assume that no pas 1 is -1ike no - the head of a ~e~p.4 What differs between them is that no pas 1 requires an explicit complement at Sstructure, and this complement must have the form of an inflected verb, as the answers in (3B) reveal. (3) A. VindriilaMaria? come-FüT the Maria 'Will Maria come?' B. a. No b. *Nopas c. No vindrh pas From a syntactic point of view it is important to point out that no pas can also occur at a different syntactic environment, exemplified so far in (2h). No pas 2 , as I shall refer to this negative marker, is in a similar syntactic distribution to other lexical items, such as nome's, solament 'only', encara 'still' and fins i tot 'even'. In fact, all these adverbials can occur at several syntactic positions, modifying either the verb or some other constituent of the sentence. Among them, no pas introduces a negative polarity effect. More exarnples which illustrate the location of these adverbials in syntactic positions other than at the left of the verb are given in (4), (5) and (6) (4) a. { Només, solament, fins i tot, encara ) VOSALTRES sereu sancionats [ only, only, even, still ] you be-FUT punished '{ Only, even, still ) you will be punished' b. No pas VOSALTRES no sereu sancionats not pas you not be-FUT punished 'It's not you who will be punished' (5) a. Només una mica de PA ens han donat only a little of bread us have given 'Only a Iittle bread has been given to us' b. No pas gens de PA no ens han donat not pas nothlng of bread not us have given 'No bread at all has been given to us' (6) a. La Maria compra pa només al forn the Maria buys bread only at the bakery 'Maria buys bread only at the bakery' b. La Maria compra pa no pas al forn, sinó a la pastisseria the Maria buys bread not pas at the bakery, but at the pastry shop 'Maria buys bread not at the bakery but at the pastry shop' In (2h), as in (4b), (5b) and (6b), there is no superficial discontinuity between no andpas. This property seems to suggest that no pas in these examples is -as beforea single lexical item which occupies a single syntactic position. Notice, furthermore, that it would be incorrect to claim that in this sort of structure no pas is the head of a NegP, for it would be quite anomalous for items of the adverbial category to have indistinctly nominal, quantified or prepositional complements: a specific property of no pas 2 is that it can precede any major lexical category, with the exception of finite verb forns. If adverbs which behave like quantifiers are not heads, the only position they can fill is the SPEC position of a major syntactic category: NP, QP, PP, etc. A further distributional property relates to the notions of scope and focus, due to the fact that those lexical items which in other languages behave like no pas, només, solament, etc. are known in the linguistic literature -arnong other temsas "focus adverbs", "scopal adverbs", "scalar particles" and "focusing particles". Since the constituent associated with this type of adverbial has been sometimes named focus and other times scope, in the next section I shall describe the distribution of no pas with regard to focus and scope. 2. No pas As a Quantifier-like Operator No pas, just like only, is always semantically associated with some sentential constituent, but this semantic association does not have to be revealed in the syntax through precedence structural relations. Most often, the constituent associated with no pas is either identified with the focus of the sentence (see (4b) and (5b)) or with a constituent of a contrastive structure (no pas X, sinó Y 'not X, but Y'), as exemplified in (6b). In the first case, the constituent associated with no pas tends to be intonationally more prominent.5 However, it is not right to assume that no pas is automatically associated with the prosodic focus of the sentence, for it is possible to find examples where the focus of the sentence does not coincide with the constituent that no pas precedes.6 Let us consider the paradigm of sentences given in (7). (7) a. La Maria no ha pas vingut the Maria not has pas come 'Maria has not come' b. LA MARIA no ha pas vingut 'It is Maria who has not come' (7) c. La Maria no ha pas VINGUT 'Maria has not COME' d . No pas LA MARIA no ha vingut 'It is not Maria who has come' In (7a) there is no prosodic focus, whereas in the other examples the constituent in capital letters is the most stressed element and corresponds to the bearer of prosodic peak. (7a), therefore, illustrates that the association with prosodic focus is neither an essential property for the occurrence of no pas, nor a necessary condition for licensing this operator at LF. What should be noticed is that these examples show an important distinction between negative structures with no and those with no pas , for only no pas contains within its scope the contrast focus of the sentence. According to Brugman (1986), the contrast focus of a certain structure has to be identified with the contrasted item, which need not be identical with the stressed element, but can properly include the bearer of prosodic peak. In the examples I am describing, the contrast focus does not necessarily have to correspond to the constituent that no pas precedes at any level of syntactic representation. Actually, in (7a-c) the contrasted item, or quantificational domain (in Rooth's (1985) terms), for the meaning of no pas can be any sentential constituent, either the VP, or the NP, or the main verb. This is so because at the level where instructions for meaning representation are codified (i.e. at LF), the negative operator modifies and c-commands every constituent within its absolute scope, which -following May (1985)- is the CP, the most immediate maximal projection dominating all sentential operators. Therefore, the quantificational domain of an operator such as no pas can be any constituent within its absolute domain.7 In accordance with this claim, (7a-c) may be interpreted in the terms exemplified in the paraphrases in (8), (9) and (10). (8) a. La Maria no ha pas vingut, qui ha vingut ha estat el Joan the Maria not has pas come, who has come has been the Joan 'Maria has not come, the one who has come is Joan' b. La Maria no ha pas vingut, simplement ha telefonat the Maria not has pas come, simply has phoned 'Maria has not come, she has simply phoned' (9) a. LA MARIA no ha pas vingut, qui ha vingut ha estat EL JOAN 'Maria is the one who has not come, the one who has in fact come is Joan' b. LA MARIA no ha pas vingut, ELLA simplement ha telefonat 'Maria is the one who has not come, in fact she has simply phoned' (10) a. La Maria no ha pas VINGUT, en realitat s'ha limitat a TELEFONAR 'Maria has not actually come, in fact all she has done is to telephone' b. La Maria no ha pas VINGUT, per6 en Joan si que ha vingut 'Maria has not actually come, but Joan has in fact come' One of the things these data suggest is that, in order to represent adequately the semantic vagueness of the sequences in (7a-c), at LF the negative adverb should be in a sentential position from which it could c-command every constituent of the negative structure. This position could be an adjunct IP position. (1 1) a. La Maria no ha pas vingut LF: [cp [c' [?? no pas [jp la Maria ha vingut I]]] b. LA MARIA no ha pas vingut LF: [CP [SPEC~ la Maria ] [c' [?? no pas [~p ei ha vingut I]]] (1 1) c. La Maria no ha pas VINGUT LF: [cp [c' [ci vingut 1 [?? no pas [~p la Maria ha ei ]I]] Notice that any of these representations introduce vague interpretations, because the scope of no pas is the set of constituents it c-commands at LF, and this set includes but is not necessarily identified with the emphasized constituent. A further observation is that most often the unmarked reading which corresponds to (7b) or (7c) is that interpretation according to which the emphasized constituent is taken as providing the set of relevant entities which bear the quantificational domain of no pas. The reason why this happens should be attributed to the position that the constituent marked with the abstract operator [+ EMPH] occupies at LF. I assume that, since the constituent identified with the prosodic focus of the sentence occurs at the most prominent LF position, it corresponds to the kind of entity over which -other things being equalit is more reasonable to understand as the no pas quantificational domain. Otherwise, in any vague example it is going to be the contextual information which will circumscribe the target of negation. Contrasting with (7a-c), the LF corresponding to (7d) would have the logical representation in (12). (12) No pas LA MARIA no ha vingut LR [cp [SPEC~ no pa la Maria ] [CI [?? no [~p ei ha vingut I]]] In this structure the absolute scope of negation is still the whole CP, because it is the minimal maximal projection dominating every constituent within the negative sentential structure. What is crucial is that in (12) a vague interpretation is not licensed, and this fact can only be explained if the position that no pas has at LF is taken into consideration. Note that in this structure no par is placed within the SPEC of the CP, together with the emphasized constituent.8 In conclusion, the fact that no pas might be associated with a prosodic focus does not seem to be part of its logical semantics. What really defines this adverb semantically is that it has a quantificational domain or contrast focus. Therefore, the semantic distinction between no and Whereas clitics have to precede the finite forms of the verb and follow the non-finite forms, no must always precede both the verb -either finite or non-finiteand the clitics adjoined to it, and pas must always follow both the verb forn -either finite or non-finiteand the clitics adjoined to it. This syntactic distribution is schematized in (21). (21) a. no + (cl) + V b. V+(cl)+pm As already noticed by Zanuttini ( 1990), if the syntactic structure of languages like Italian and Catalan was the one in (18a) -postulated by Pollock (1989)-, then we should expect that the movement of the negative marker from the head position of NegP to the head of TP, together with the pronominal clitics, might lead to some sort of reordering process among these items. However, this reordering -at least regarding no -, never takes place. On the other hand, although I assume that pas must be reordered correctly at S-structure, it clearly differs from the clitic pronominal system as far as the second morpheme of no pas I is not restricted to follow non-fini te forms of the verb. 1 1 From these comments we can conclude the following: if the NegP is assumed to be a syntactic projection higher than TP and AgrP, then there is no reason to suppose that Catalan negative markers are clitics and that they must be raised up to a higher position. 12 The hypothesis that the NegP is the highest node would also predict the possibility that both no and pas can bear stress (cf. Zanuttini (1990)), as well as the fact that in Catalan the verb can never occur to the left of no. (22) a. *Vaig no veure'l (pas) PAST not see-him pas 'I didn't see him' (22) b. * Vindh no (pas) come-FUT not pas 'He will not come' c. *Ser no (pas) puntual és un mal costum be-INF not pas punctual is a bad habit 'Not to be punctual is a bad habit' In addition to accounting for the non-clitic nature of negation and the relative position of negation with regard to the verb, the analysis of sentential negation must also predict the morphological characteristics of the Catalan verb system. According to the Mirror Principle (cf. Baker (1985)), it is predicted that morphological derivations must directly reflect syntactic derivations (and vice versa). This means that, taking into consideration the morpheme structure of the verb forms and, more specifically, the relative order of tense and agreement morphemes with respect to the verb, the projection of Tense cannot be higher than the projection of Agreement (cf. Belletti (1988, 1990b), Sola (1989)). (23) a. pintaven /pínt+á+ba+n/ paint + lconj.+ imperfective tense + 3p.pl. They painted' b. dormiríem /dórm+i+ría+m/ sleep + 3conj. + conditional tense + 1p.pl. 'We would sleep' As illustrated in (23), if the syntactic and morphological order of Catalan verb forms is V+T+Agr, then at D-structure AgrP must be higher than TP; which is quite the reverse of what Pollock (1989) postulates. Paying attention to this remark, and assuming that in Catalan -as in French, Italian and Spanishaffixation is done through a head movement rule which raises all verbal heads to higher hierarchic positions (in order not to violate the Head Movement Constraint and the Empty Category Principle; cf. Chomsky (1981), Travis (1984)), then the proposals (18b) and (18d) are the only ones that can correctly predict the right order of verbal suffixes in this and similar languages. But, if at the same time we attempt to explain the non-clitic nature of Catalan negative markers, then we are left with the syntactic relation between functional categories postulated in (18d). In summary, the answer to the question of what acts as complement of negation seems to point to the functional category AgrP, which in turn would have TP as its complement.13 The advantage of (18d) is that it provides an account of the sort of data given in (24). (24) a. No obrir (pas) la porta quan truquen pot ser prudent not open-INF pas the door when call may be sensible 'Not to open the door when someone is calling may be sensible' b. No obrint (pas) la porta, és difícil que et passi res not open-GER pas the door is difficult that you happens something 'Not opening the door, it is unlikely that anything will happen to you' c. *No oberta (pas) la porta, hauris d'entrar pel garatge not open-PP pas the door have-FUT to go-in through the garage The door not (being) open, you will have to go in through the garage' d. *No obre (pas) la porta not open-IMP pas the door 'Do not open the door' e. No obri (pas) la porta not open-SUB pas the door 'Do not open the door' (2 p. sg.) (24) f. No obriu (pas) la porta not open-1ND pas the door 'Do not open the door' (2 p. pl.) 14 In order that a negative i tem might be licensed as the bearer of sentential negation, a condi tion on the assignment of sentential scope must be satisfied. (25) Constraint on the Assignment of Sentential Scope to Negation Negation can take sentential scope only if -at S-structure - it is in a position from which it c-commands both the TP and the AgrP (cf. Zanuttini (1989: 14)) In accordance with this condition, the negative adverbs which occur in (24c,d) cannot be licensed as propositional operators, for in these examples they coappear with verb forms which are not marked for Tense. With regard to past participial clauses, I follow the idea, already pointed out by Belletti (1989, 1990b), that they are AgrPs which lack a temporal specification. This AgrP corresponds to the projection which in the full clausal structure represents the past participial complement of Aux: AgrP0bject (cf. Chomsky (1989)). In this syntactic context negation is not allowed, because NegP fills a position outside the past participial AgrP. Lack of Tense and Agreement features with the subject can also be postulated in relation to true imperative forms. Notice that this mood is fully defective: in most Catalan verbs the imperative is reduced to a single form, which is used in association with the second person just because the imperative is supposed to appeal to the addressee's attention, but note that a verb fonn such as the one in (24d) (obre ) does not bear any second person agreement features.15 Consequently, (24d) is ungrammatical, because a negative adverb occurs with an imperative form which is characterized by its lack of Tense and subject Agreement features. The examples in (24e,f) illustrate what are actually known as imperative suppletive forns, either taken from the present indicative tense or the present subjunctive tense. The verb form in (24e) (obri ), taken from the subjunctive, may be defined by means of the features [-T, +Agr] (cf. Picallo (1984)). Being positively defined for at least one of these features, the adverbs no and no pas 1 can occur with this verb form without leading to any sort of ill-formedness. Likewise, the verb form in (24f) (obriu ), taken from the indicative, is positively defined for both Tense and Agreement and, therefore, it can freely appear with either negative adverb. With regard to the infinitive and gerund forms, exemplified in (24a,b) (obrir, obrint ), notice that in Catalan they do not show any sort of Agreement features with the subject of the sentence. Nevertheless, it seems reasonable to assume that these forms hold some aspectual features (imperfective in the infinitive forms, and progressive in the gerund forms) which are inseparable from their temporal specifications. In accordance with this assumption infinitives and gerunds may be said to have at least a Tense projection and, as a consequence, can freely coappear with a negative adverb. As a conclusion, the set of data which has been described in this section provides severa1 arguments for postulating that the NegP must be in a structural position higher than AgrP and TP. In the final section I shall address the question of what sort of formalism could allow the location of the negative morpheme pas in the appropriate syntactic position within a sentential structure. I arn going to assume the existence of a process of syntactic incorporation (cf. Baker (1988)) according to which the verb complex is adjoined to the head of the NegP. My proposal is that only after the application of this verb movement, which also affects infinitive forms, pas can find an appropriate position within the new syntactic constituent built through adjunction. 3.3. Incorporation to Negation It has already been argued (cf. Llinis (1990)) that in Catalan a complex verb Aux + V is an Xo syntactic unit, which implies that this verb sequence cannot be intermpted by any independent syntactic constituent, for example an AdvP or a NP (cf. Espinal (1983)). (26) a. *Ha sovint telefonat? has often phoned 'Has he often phoned?' b. *Va en Joan telefonar? PAST the Joan phone 'Did Joan phone?' The only constituents that can interrupt the components of this verb complex are a group of items strongly related to the verb: either the clitic pronouns or the negative morphemepas, as the exarnples in (27) illustrate. (27) a. Quan va haver-la vista, se'n va anar when PAST have-her seen, left 'Once he had seen her, he left' b. No ha pas vingut not has pas come 'He has not come' In order to explain this distribution, I am going to have recourse to a process of syntactic incorporation, which results in the configuration specified in (28).16 (28) Process of Syntactic Incorporation (cf. Lliniu (1990)) It has been observed that, once this incorporation process has been carried out, a clitic may change its precedence structural relations with regard to the complex verb (either before the finite verb, or after the non-finite verb), without affecting the grarnmaticality of the sequence. (29) a. Lava veure her PAST see 'He saw her' b. Va veure-la In a parallel way, my proposal is that after the verb incorporation process has applied, the verb complex is successively adjoined first to T, then to Agr and finally to Neg. 17 Along the same lines as it is currently assumed that T and Agr attract V only if they are provided with a certain morphological strength, I assume that, when negation is present, Neg attracts V. Further incorporation of this complex constituent to the head of CP is ruled out because no morphological or syntactic infonnation would be attached to V. The result of this head-to-head incorporation process is illustrated in (30). Nego I A no pas A Agr V 1 T Notice that this process of syntactic incorporation applies both to the right and to the left. More specifically, when a verb is moved in order to be adjoined to another lexical head (either an auxiliary verb or a negative adverb), it incorporates to the right. However, when a verb head (even a complex verb head) is adjoined to a non-lexical head (for example, Tense and Agreement), then it incorporates to the left. Since the verb forms in V2 (participles, gerunds and infinitives) have already undergone agreement attachment (cf. Belletti (1990a,b)), no more affixes can attach to them, and therefore the Tense and Agreement features specified in (30) can only be attached to the head of V1. Once this morphological process has applied, a syntactic structure such as (3 1) is obtained. NegP \ It is precisely at this point whenpas, following lexical requirements peculiar to no pas 1, is reordered within the new Nego constituent and, as a result, is placed in any postverbal position that can be found within this complex constituent. The data in (32) provide evidence in support of this claim. (32) a. No ha pas estat trucant not has pas been ringing 'It has not been ringing' b. No ha estat pas trucant c. No ha estat trucant pas Notice that whatever the syntactic position of pas may be, no is still c-commanding both Tense and Agreement and, therefore, the sentential scope of this operator is adequately licensed at the level of LF. OThis paper was delivered at the Primer Coloquio de Gramática Generativa (Madrid, 1991). Many thanks to C. Plcallo for her comments. This work has been partly supported by a grant from the DGICYT (number PB89-0323). Notes 1 This is the dialect broadly spoken in the provinces of Barcelona, Girona and Tarragona. 2 Actually, the structure no X seems to be lexically and syntactically restricted to certain quantifiers: no tot(s) 'not all', no gaire(s) 'not many', to certain quantifier-like adverbials: no nom3, no solament 'not only' and to other specific lexical items. (i) a. No tots els estudiants aniran a la universitat 'Not all the students will go to the university' b. No gaires estudiants aniran a la universitat 'Not many students will go to the university' c. No només els millors estudiants aniran a la universitat 'Not only the best students will go to the university' d. Se celebrar& una conferkncia internacional dels paisos no alineats 'An international conference of the non-aligned countries will be held' 3 From now on each example containing no pas can be interpreted either as a presuppositioncancelling type of negation (rejecting some of the speaker's assumptions with regard to what is being said), or as a presupposition-confirmation type of negation (reinforcing these assumptions) (cf. Espinal ( 1990), Wilson-Sperber ( 1988)). Analysing pas as a specifier (cf. Pollock (1989)), of the same sort as clitics, articles and weak possessive forms would incorrectly predict that pas at S-structure should be able to emerge at the left of no. However, this possibility only yields ungrammatical sequences.