Two Squibs on Modality and Negation

TWO SQUIBS ON MODALITY AND NEGATION M. Teresa Espinal Universitat Autbnoma de Barcelona This paper is concemed with two different linguistic problems raised by Catalan nopm negative clauses, and will examine their representational effects. It is argued that pas can occur within the boundaries of a Tense Dependent subordinate clause if, and only if, there is a modal constituent c-commanding the NegP, and the structural requirements for the condition of logical absorption are not satisfied. The first squib deals with general restrictions on the use of no-pas in structures which contain a subordinate clause, while the second squib deals more specifically with restrictions on the use of rw-pas in expletive contexts. These restrictions are exemplified in (1) and (2). Compare on the one hand (la) with (lb), which for some speakers is absolutely ungrammatical, whereas for others it shows some degree of ill-formedness. On the other hand, compare (2a) with (2b), which cannot be used under the expletive interpretation licensed by (24. (1) a. No diu pas si vindd. not says pas whether come-3sg-FUT '(S)he doesn't say whether (s)he is going to come.' ('not' against what you appear to think, 'not' against what somebody might suppose, 'not' confirming what someone might believe, etc.) b. *I?? No diu si vindrh Pas. not says whether come-3sg-FUT pas Catalan Working Papers in Linguistics (CWPL) 3.1 (1993): 113-138 Universitat Autbnoma de Barcelona (2) a. Tinc por que (no) nevi. have-lsg fear that not snow-3sg-SUBJ-PRES 'I'm afraid that it willlwill not snow.' b. Tinc por que no nevi Pas. havelsg fear that not snow-3sg-SUBJ-PRES pas. 'I'm afraid that it will not snow.' ('not', for example, against my expectations of going skiing) The common syntactic property of (Ib) and (2b) is thatpas occurs at S-structure next to the verb of a [+ Tns Dep] subordinate clause, and the hypothesis I shall put forward is that this configuration is possible just in case (i) there is a constituent with modal specifications ccommanding the NegP, and (ii) the stmctural conditions for logical absorption of negation are not met. 1. No-pas in Clausal Structures In (1) I have posed a problem: the existence of restrictions on the distribution of no-pas in subordinate clauses. And this problem may be solved as follows. First, I am going to claim that no-pas is licensed in subordinate clauses if, and only if, at the level of LF either an affirmative operator or a modal operator c-commands the negative operator, which shows that in order to license no-pas a specific stmchral environment which grammaticalizes modality is required. Second, since no other kind of operator is able to license no-pas at the subordinate clause, I shall argue that modality (represented by modal verbs, the subjunctive mood, and particles of affirmation) is concemed, in its tum, with the grammaticalization of subjectivity (i.e. speaker's subjective attitudes and opinions; cf. Coates (1990)) and, because of this, at the final level of meaning representation access to somebody's expectatlons, as needed by no-pas (cf. Espinal (1993)). is still possible. My first claim focuses on the important structural effects that modality has on negation. It is my intention to argue that some explicit form of modality is required in order to license Catalan nopas at the subordinate clause, while this is not so in order to license the unmarked negative lexical item no. Some relevant data showing that the grammaticalization of modality is compulsory in order to license no-pas in a [+ Tns Dep] subordinate clause is given in the pair (lb,c). Notice that, although speakers prefer (la), whenpas occurs within the boundaries of a subordinate clause a modal verb is responsible of its licensing. Unlike (lb), (lc) is well-formed both under the root sense and the epistemic reading of the modal verb.' (1) a. No diu pas si vindrii. not says pas whether wme-3sg-FüT '(S)he doesn't say whether (s)he is going to come.' b. */??No diu si vindd PS. not says whether come-3sg-FUT pas c. ?No diu si podrii pas venir. not says whether may-3sg-FüT pas wme '(S)he doesn't say whether {(s)he will be able / it will be possible for her / him) to come.' Additional data can be observed in (3) and (4).2 Notice on the one hand the contrast between (3c) and (3d), and on the other the contrast between (3d) and (3e). See Picallo (1992) for a syntactic study of modal verbs in Catalan. Restrictions on the use on no-pas in [+Tns Dep] subordinate clauses are widely observed in conditional clauses, finite complement clauses, nonfinite complement clauses, relative noun clauses, restrictive relative clauses, nonfinite adjunct clauses, etc. A couple of examples are given in (i). (i) a. CONDITiONAL CLAUSE * Si ell no diu pas res, li trucarem nosaltres. if he not sayspas anything 3sgDat phone-lpl-FUT we 'If he doesn't say anything, we'll phone him.' (3) a. No has telefonat? not have-2sg phoned 'Haven't you called?' b. No has pas telefonat? not have-2sg par phoned 'Do you happen to have called?' b. RELATIVE NOUN CLAUSES * Els qui no us esforceu pas no aprovareu. the who not yourself try-harder-2plpas not pass-2pl-FUT Those who don't try harder will not pass.' No-pas can only occur in the complement (indicative) clause if its Tense specifications are not bound to the upper Tense specifications; ibis happens when the C of the subordinate clause denotes the speech time. (ii) a. Diu que no es pensa pas afaitar. says that not himself thinks pas shave 'He says that he doesn't intend to shave himself.' b. Has de saber que no estic pas contenta de tu. have-2sg to know that not ampas happy of you 'You need to know that I'm not happy with you.' Exclamative sentences and non-reshictive relative clauses aiso ailow the presence of no-pas, which suggests that they should be analysed as Tense independent clauses. See Espinal (1991a) for an analysis of non-restnctive relatives as disjunct constituents. (ui) a. Entra! Que no hi ha pas ningú! come-in-2sg-IMP that not there has pas anybody 'Come in! There is nobody!' b. Si ell no diu pas res! QUI parla és la seva dona. if he not says pas anything who talks isthe his wife 'He doesn't say a word! The one who is talking is his wife.' c. Els terroristes, que no es van pas identificar, sortiren emmascarats per la televisib. the terrorists that not themselves PAST3plpas identify appeared-3pl put-on-mask by the TV 'The terrorists, who did not identify themselves, appeared under a mask at the TV.' c. Que no has telefonat? IM* not have-2sg phoned 'Haven't you called?' d. * Que no has pas telefonat? IM not have-2sgpas phoned e. {Eh, oi) que no has pas telefonat? ~ffpart* IM not have-2sg pas phoned 'You haven't called, have you?' In (4) the contrast between on the one hand (4a,b), which have indicative tenses in the subordinate clause, and on the other hand (4c-f), which have either a subjunctive mood or a modal verb in indicative I subjunctive, should be noticed. (4g) with a modal adverb modifying an indicative verbal form at the subordinate clause is out. (4) a. * No sabia que tomava pas. not knewlsg that retumed3sg pas b. * No sabia que tomaria Pas. not knew-lsg that return-3sg-COND pas c. No sabia que tornés pas. not knew-lsg that return-3sg-SUBJ-PAST pas 'I didn't know that (s)he might come back.' d. ? No sabia que havia pas de tomar. not knewlsg that had3sg pas to return 'I didn't know that {(s)he had 1 it was necessary for her) to come back.' e. ? No sabia que hauria pas de tornar. not knew-lsg that have-3sg-CONDpas to return * IM stands for Interrogative Marker * * AffPart stands for Affirmative Particles. f. No sabia que hagués pas de tomar. not knewlsg that have-3sg-SUBJ-PAST pas to return e. * No sabia que probablement {tomava, tomaria} Pas. not knew-lsg that probably retumed3sg retum-3sg-COND pas What is of interest to us is that all these examples are fully well-forn4 with the simple negative item no. And, therefore, this paradigm of sentences seems to show that the structural conditions that license no and no-pas in Catalan differ significantly. First, from a syntactic point of view the inadequacy of the data presented so far must be correlated with the occurrence of a [+ Tns Dep] CP. Second, the ill-formedness of those examples which have no in the main clause and pas in indicative contexts of a subordinate clause must be pointed out. These data should be contrast4 with further exarnples which have pas either in subjunctive contexts or next to a modal verb in indicative I subjunctive mood (compare (lb) with (lc), and (4a,b) with (4c-f)), commonly evaluated as well-fonned by native speakers. Third, the last example, (4g), shows that modal adverbs and modal verbs are members of different syntactic categories: modal adverbs do not seem to be the head of a Modal Phrase (MP), a syntactic category which (following Laka (1990)) has been postulated in linguistic theory to include modal verbs, and the subjunctive mood (in addition to other morphosyntactic features). Fourth, notice that no-pas, although interpreted as a single negation (as the English glosses illustrate), may appear as a discontinuous item, which suggests that in (lc) and (4c-f) the subordinate CP does not count as a structural barrier between no and pas (cf. Chomsky (1986b)). From these observations several syntactic questions can be raised: (i) whether discontinuous lexical items do in fact exist in natural languages, (ii) whether the locality conditions on government such as minimality and banierhood are relevant in these configurations, and (iii) whether syntactic incorporation (of V into Neg, and of the constituent with modal specifications into the head of CP) is the right process which must be postulated in order to explain the distribution ofpas in the configuration of (1) and (4). I am not going to address any of these issues in great detail, since they are the subject of my future research. Instead, I am going to devote this squib to the representation of meaning that corresponds to these examples at the linguistic level where meaning is supposed to be codified. I assume, following the generative paradigm, that the contribution of grammar to meaning is represented at the level of Logical Form (LF), which interfaces the theories of linguistic form (the language system) and interpretation (the Conceptual-Intentional system). All the data appear to show that at LF no and pas must be in the same intensional domain, since in spite of the fact that at S-structure we all recognize the existence of a subordinate clause, one single negative domain is to be licensed at the level of LF.3 With regard to this point an important syntactic difference between Catalan on the one hand, and French and Occitan on the other must be pointed out: in Cataianpas is not an independent lexical unit, but a morpheme which at the present stage is syntactically and semantically inseparable from no. This means that, following Ladusaw (1993). if we consider the head-particle cycle represented in (i) (which is glossed in (ii)), we should conclude that Cataian is at stage (i b), whereas French is at stage (i c). (i) a. ne b. *( ne) ...p as c. (ne) ... pas d (*ne) ...p us (ii) a. Head expression of negation b. Head supported by sirengthener c. Sirengthener able to express negation and be concordant d Former strengthener now independent expressar of negation Now, if (lc) and (4c-f) license one single sentential negation, whose scope domain corresponds to the main clause, several implications must be drawn as to the form of the representation of their meaning. The first is that no marks the position of the structural head of the negative operator. In other words, the position that no has at S-structure corresponds to the structural position of the negative head at LF. Following a hypothesis put forward in Espinal (1991b), my claim is that this negative head can be filled at any level of syntactic representation by a compound negative item, such as no-pas. Therefore, in (lc) and (44 only one single sentential negation is licensed, whose scope domain corresponds to the main clause. Notice, furthermore, that in languages such as Catalan, Italian, and Spanish the Negative head in itself suffices to express sentential negation, which means that quite often there is a non-overt negative operator to satisfy the Neg Criterion postulated by Haegeman and Zanuttini (1991).4 A second implication is that the blocking effects of the subordinate CP can be neutralized due to a constituent with modal specifications, either an affirmative particle, a modal verb, or a subjunctive mood (which I define as a [+ Mod] Tense head) that c-commands the negative head. A language will have reached the end of the cycle by the time it is no longer possible to doublepas with ne /no. With respect to this implication I assume the following definitions: (i) Neg operator: a Neg phrase in a scope position (cf. Haegeman (1992)) (ii) Scope position: a left-peripheral A'-position (either a Spec or an adjoined position) (cf. HaegemanZanuttini (1991)) (iii) The Neg critmion (cf. Haegeman (1992). Hxgeman-Zanuttini (1991)): a. Each Neg XO must be in a Spec-Head relation with a Negative operator b. Each Negative operator must be in a Spec-Head relation with a Neg 9 (iv) The Scopeprinciple: an operator A has swpe over an operator B in case A c-commands a member of the chain containing B (cf. Emst (1991:753)). Let us first consider the data in (3). It should be noticed that in (3e) the modality particles {eh, oi} c-command the negative head and are the unique constituents responsible for the contrast between (3d) and (3e). From a lexical point of view, both eh and oi are particles used to demand agreement, assent or confirmation of the content of the proposition they precede. (Particle oi can even be used to express confirmation of the content of the preceding utterance). From a structural perspective what is important to point out is that these modal particles are in a scope position, in the sense that they are in a left-peripheral A'-position. More specifically, these particles are in complementary distribution with quexpressions.5 (5) a. {Eh, oi} que no vas telefonar? AffPart IM not PAST-2sg phone 'You didn't phone, did you?' Notice, furthmore, the following contrast: (i) a. {Eh, OI) que no vas pas telefonar? AífPart that not PAST-2sg pas phone 'You didn't phone, did you?' b. *I?? Per quP no vas pas telefonar? why not PAST-2sg pas phone Why didn't you phone?' c. ? Per que no vas pas poder telefonar? why not PAST-2sg pas can phone Why couldn't you phone?' The ill-fonnedness of (i b) should be attributed to the lack of a modal constituent c-commanding the marked negative head. either a modal particle, a modal verb or the subjunctive mood in a scope (i.e. c-commanding) '*. position. Notice also that if modals were assumed simply to be adjoined to the most immediate sentential domain in the LF-component, this hypothesis would neither allow us to explain the contrast in grammaticality between (lb,c), nor would it account for the fact that pas can cross a CP boundary in (Ic) but not so obviously in (Ib). So far I have considered the set of structural specifications which seem to be required in order to license no-pas in syntactic contexts which contain a Tense Dependent subordinate clause. The next question I shall address is why it is the case that modality has such a significant effect on the process of licensing, that is, on the process of providing a full interpretation for marked negation in subordinate clauses. In other words, what is the relevance of the sort of representation I have postulated by the time the process of utterance interpretation is reached? It is widely accepted in the linguistic literature on modality that modal forms appear to be the chief lexical exponents of subjectivity. Palmer (1986) and Coates (l990), among others, argue that modality in natural languages is concerned with the subjective characteristics of an utterance, and therefore allows the grammaticalization of the speaker's subjective attitudes and opinions. The data we are concerned with reveal that access to either the speaker's or the hearer's expectations is legitimate, as determined by no-pas, just in case there is some other constituent which attends to the matter of bringing forward the required subjective information; othenvise, no-pas is not licensed in subordinate clauses.10 And the unique constituent which makes this 10Two distinct negative lexical items exist in Catalan, which contribute differently to the Propositional Form corresponding to the negative structure in which they occur. No-pas differs from no in that it not only contributes to the proposition expressed -exactly as no does-, but it aiso contributes to implicatures, and therefore to the process of enrichment of a linguisticaily underdetemined language expression, by implying a non-descriptive use of negation. No-pas can actually be said to constrain the comprehension process by access to people's expectations possible is a modal constituent: either a modal verb, a modal particle, or a subjunctive mood. Modal forms appear to imply a subjective commitment to the proposition expressed. Consequently, whereas from a grammatical point of view they c-command the negative constituent and may contribute to the propsition expressed, from a pragmatic view point they are concerned with the subjective information carried by an utterance. Modal particles (contrary to Wilson-Sperber's (1993) analysis) seem to encode linguistic information which is necessary to infer the proposition expressed. An example such as (3d) evidences that the absence of a modal particle results in an ill-formed linguistic expression. Accordingly, modal particles seem to contribute to the truth conditions of the utterance which carry them. In addition, they seem to constrain the sort of propositional attitude that the speakerlhearer must hold towards the propsition expressed. From this point of view, they are said to encode procedural information on attitudinal indicators, such as confirmation and assent. On the other hand, modal verbs and the subjunctive mood encode linguistic information which contributes clearly to explicit content, by encoding not only a grammatical relation with regard to the negative operator, but also higher-level explicatures (i.e. the speaker's opinions made manifest in the English translation of (lc) and (4c-f)). That is, they entail conceptual information which contributes to the speaker's op~nlons and attitudes towards the hearer for instmcting the hearer that (s)he has to take the proposition in which no-pas appears as a premise, and ennch this proposition in such a way that some cognitive effects might be drawn: either the reinforcement or cancellation of that proposition, or certain inferences derived from what the speakerlhearer considers as a desirable thought (in interrogative contexts), or a prohibition reinforcement or threat (in imperative contexts). And the claim that nopas conveys an interpretive use must be related to the restrictions on the use of no-pas in subordinate clauses. SeeEspinai (1993) for further details on the interpretation of no-par in Catalan what is said in the propositional form; in other words, the logical specifications of these constituents provide information on the speaker's commitment to the propsition expressed. To conclude, both modal particles on the one hand, and modal verbs and the subjunctive mood on the other, seem to provide the required subjective information for licensing no-pus in Catalan. Finally, 1 would like to pint out that the difference between modal adverbials and modal specifications on verbal heads observed in (4c-g) must be assumed to be determined from the lexicon. Whereas the modal operator corresponding to adverbials expresses objective modality, and contributes solely to the proposition expressed by the utterance, the modal operator corresponding to the subjunctive mood expresses epistemic possibility, and the one corresponding to modal verbs can either express subjective epistemic modality or deontic modality (cf. Lyons (1917)), and encodes logical information which not only contributes to the propositional form, but also to higher-order explicatures, to statements of opinion and attitude. Thus, it can be concluded that only expressors of subjective modality fulfil the semantic requirements of no-pas. In the second squib, 1 shall deal with a different group of data which shows that unlike no nopus cannot be interpreted expletively when it is the head of a negative subordinate clause. 2. No-pus in Expletive Contexts First, it should be noted that the term expletive negation (EN) refers to a Neg syntactic constituent which appears in certain syntactic environments, but makes no effective contnbution to the interpretation of the whole stnng containing this constituent. Some examples are given in (14), which have both an expletive and a negative reading. (14) a. Tinc por que (no) nevi. (= (2a)) have-lsg fear that not snow-3sg-SUBJ-PRES 'I'm afraid that it willlwill not snow.' b. Impediu que la policia (no) trobi l'assassi. stop-2pl-IMP that the police not find-3sg-SUBJ-PRES the-murderer 'Stop the police from findinglnot finding the murderer.' In a previous paper (cf. Espinal (1992)) I have approached the question of why negation is interpreted expletively in some syntactic contexts. I have argued that EN is a natural language phenomenon induced by specific lexical items appearing under specific structural conditions. This phenomenon is accounted for if, and only if, at the level to which the Principle of Full Interpretation applies (i.e. at LF) the effects of the negative operator are absorbed, or cancelled, by the logical specifications stipulated in the lexical entries corresponding to the set of lexical items which semantically select and govern a negative head at the complement clause (amongst others: 'before', 'until', 'to fear', and 'to prevent'). Thus, logical absorption is conceived as a licensing condition on LF representations, for this condition explains why certain negative constituents are not licensed as independent negative concepts. Negation is absorbed in accordance with the definition of EN given in (15) (cf. Espinal (1992)):" I have been pointed out by the audience of a lecture given at Stanford University on July 17 of 1993 that the tem 'entailing' has a sense too specific within formal semantics to be used appropriately in the definition of logical absorption given in (15b). Thus, lexical items such as 'false' and 'wish', although they seem to entail some negation, do not license expletive negation. (15) a. Expletive negation occurs at LF whenever there is a set of contiguous nodes in a tree T (a lexical category X, a complementizer que, and a Neg head), which form a subtree of T, and X, the governor, can absorb the effects of negation due to its logical meaning b. a absorbsJ, a = a lexical category entailing negation and R = Neg, in the following configuration: [ ... a [ CO [ ... R ... I]] i ff b ' . Minimality is respected b". There is no logical operator intervening between a and B at the level of LF The question that still remains to be answered is why this operation cannot cancel out the specific logical effects of a marked negative lexical item, such as no-pas. Notice that, in contrast with (14a,b),,the examples in (16) can only be licensed under a negative reading. (16) a. Tinc por que no nevi Pas. (= (2b)) have-lsg fear that not snow-3sg-SUBJ-PRES pas 'I'm afraid that it will not snow.' b. Impediu que la policia no trobi pas I'assassi. stop-2pl-IMP that the police not find-3sg-SUBJ-PRES pas the-murderer 'Stop the police from not finding the murderer.' The situation becomes even more interesting when we consider data that contain a modal verb within the subordinate clause of a lexical item that licenses EN. Some relevant examples are given in (17). (17) a. Temien que la seva filla volguCs estudiar a I'estranger. feared3pl that the their daughter want-3sg-SUBJ-PAST study to the-abroad They were afraid that their daughter wanted to study abroad.' b. Temien que la seva filla no volgués estudiar a l'estranger. feared3sg that the their daughter not want-3sg-SUBJ-PAST study to the-abroad 'They were aftaid that their daughter wantedldid not want to study abroad.' c. Temien que la seva filla no volgués pas estudiar a I'estranger. feared3pl that the her daughter not want-3sg-SUBJ-PAST pas study to the abroad 'They were afraid that their daughter did not want to study abroad.' The data in (14), (16) and (17) provide evidence for a NegP in the subordinate clause, and this negative clause is semantically selected by a specific predicate that appears in the main clause (i.e. the Catalan counterpart of 'to be afraid of', 'to stop' and 'to fear'). Besides, the Tense specifications of the subordinate clause are linked to the Tense specifications of the main clause. It is important to notice that in this syntactic environment the negative item no-pas cannot occur in subordinate clauses unless a modal constituent (either the subjunctive mood, or a modal verb such as 'may', 'want' or 'have to') c-commands the negative marker. But notice that in this linguistic context no-pas constrains the negative interpretation. That is, the common property shared by the data in (16) and (17) is that when a modal constituent c-commands no-pas, only the negative reading is fully licensed. It would appear, therefore, that under specific structural configurations no-pas cannot be absorbed by the logical content of the set of lexical items which in Catalan select and govern a negative head at the complement clause. In the previous squib the logical requirements of modal items were mentioned, mainly the requirement that they must be licensed as sentential operators by the time the level of LF is built. These lexical instructions determine that, if the modal head has not already moved in the syntax proper, it must move -as a last resort operation (cf. Chomsky (1992))- in the LFcomponent, and must adjoin to the upper sentential node (the NegP in expletive contexts), since it must be interpreted as modifying the whole sentence in which it appears. This means that in (14), (16), and (17) two different constituents (the modal constituent, either a modal verb in the subjunctive mood or any main verb in the subjunctive, and the Neg head) must be adjoined to the subordinate NegP, and depending on the order in which these adjunctions apply, either the expletive or the negative reading is licensed.12 In the previous squib it was also mentioned that no-pas cannot occur in [+ Tns Dep] clausal structures unless it is c-commanded by a modal operator which carries out higher-level explicatures. The relevant output LF structure has a modal operator between a particular predicate governing the negative adverb, and the negative operator itself, and in this syntactic configuration negation cannot be absorbed, for the required structural contiguity (stipulated in (15)) between the lexical head X and the Neg head is interrupted by an adjoined modal constituent that has lexical and logical features which stop the absorption condition from being applied. This configumtion is represented in (18): X ' n X CP [+ Tns Dep] I C ' n C NegP /-'. (M) + [+Modi T NegP I Neg' I l2 See Espinal (1992) for further details of these adjunction constructions. -- In ~hort, this representation would suggest that NegP is a blocking category for government of Neg by X, since at LF it contains a logical operator adjoined to NegP (see (9b)). With this intervening operator, CP becomes a barrier for government of Neg by X by inheritance (see ( IOa)), and absorption is ruled out. Besides, no-pas has very strict lexical requirements which underdetermine the utterance's interpretation. Accordingly, there are difficulties in accepting expletive no-pas, for the negative proposition is already entailed logically by the logical content of the lexical head, and no specific reinforcement of the negative proposition is expected to be drawn in purely expletive contexts. To conclude, in this paper I have argued for a representational approach to the study of the significance of language. Specific linguistic problems involving the use of the compound negative lexical item no-pas in Catalan can only be appropriately accounted for under a representational semantic theory which distinguishes between: (a) information which is linguistically and conceptually encoded in the lexicon as a contribution to or a constraint on implicatures (e.g. no-pas logical specifications), (b) information which is linguistically encoded at the level of LF as a contribution to the proposition expressed (e.g. the c-command structural relation between modal operators and the negative operator, and the configurational constraint which controls the expletive reading), and (c) information which is linguistically encoded at the level of LF as either a contribution to or a contraint on higher-level explicatures (that is, conceptual structures representing some conceptual comment on somebody's subjectivity and propositional attitudes). More particularly, this study has evidenced the structural effects that modality might have on negation. Modal verbs and the subjunctive mood behave similarly with regard to licensing nopas in clausal structures. In fact, the logical specifications of modal lexical items already determine what their contribution to LF formation is going to be: they are to be licensed as propositional operators, and this requirement is implemented within grammar itself by movement of the propositional operator from some constituent-interna1 position to a configuration in which operator scope is syntactically expressed. This upward movement of the syntactic head which contains modal specifications (M and [+ Mod] T) can take place either before reaching the LF-component or at the level which is assumed to represent the syntaxsemantics interface. From squib 1 it can be concluded that, if the modal head moves in the syntax, it must be incorporated into the head of CP. This hypothesis accounts for both the position of no-pas in Catalan negative main clauses when pas crosses a CP, and for the position of modal verbs in interrogative sentences in English. From squib 2 it can be concluded that, if the modal head moves in the LF-component, it must be adjoined to the most immediately dominating sentential node. This adjunction structure accounts for the fact that a modal operator, when intervening between a lexical head (that selects and governs a negative marker) and the negative operator itself, stops the condition of logical absorption from being applied. Furthermore, it should also be concluded that Catalan no-pas is licensed within a [+ Tns Dep] subordinate clause if, and only if, at the level of LF either a truth operator or a modal operator c-commands the negative operator, which shows that in order to license no-pas a specific structurai environrnent is required. On the other hand, since no other operator is able to license no-pas at the subordinate clause, it should also be concluded that modality (being represented either by modal verbs, the subjunctive mood, or particles of affirmation) is concerned with the grammaticalization of subjectivity (i.e. speaker's subjective attitudes and opinions, hearer's attitudinal effects), and so it satisfies the logical requirements of no-pas. Some aspects of the content of this paper were delivered on May 1993 at the third International Colloquium on Cognitive Science (San SebastiBn) under the title 'Marked Negation: A Case Study in the Grammar-Cognition Interface'. I have benefited from a D.G.I.C.Y.T. grant (number PB 89-0323), and from comments on the data by M. Batllori, T. Cabrk, C. HernBndez, N. Martí, G. Rigau and A. Suñer. References Bonet, E. (1989) 'Postverbal Subjects in Catalan', ms. MIT. Chomsky, N. (1986a) Knowledge of language, Praeger, New York. Chomsky, N. (1986b) Barriers, The MIT Press, Cambridge Mass. Chomsky, N. (1992)'A Minimalist Program for Linguistic Theory', MIT Occasional Papers in Linguistics, number 1, Dept. of Linguistics and Philosophy, MIT, Cambridge Mass. Coates, J. (1990) 'Modal Meaning: the Semantic-Pragmatic Interface', Journal of Semantics, 7, 53-63. Ernst, T. (1991) 'On the Scope Principle', Linguistic Inquiry, 22,750-756. Espinal, M. T. (1991a) 'The Representation of Disjunct Constituents', Language, 67.4,726762. Espinal, M. T. (1991b) 'Negation in Catalan. Some Remarks with Regard to no-pas', Catalan Working Papers in Linguistics 1991,33-63. Espinal, M. T. (1992) 'Expletive Negation and Logical Absorption', The Linguistic Review, 9.4, 333-358. Espinal, M. T. (1993) 'The Interpretation of no-pas in Catalan', Journal of Pragmatics, 19, 353-369. Haegeman, L. (1992) 'Sentential Negation in Italian and the Neg Criterion', Geneva Generative Papers, 0, 10-26. Haegeman, L. and R. Zanuttini (1991) 'Negative Heads and the Neg Criterion', The Linguistic Review, 8, 233-251. Ladusaw, W. (1993) 'Negation, Indefinites, and the Jespersen Cycle', ms. University of California, Santa Cruz.