Nihil est in LF quod prius non fuerit in SS

Abstract In this paper it is argued that dislocates are interpreted in their surface position. Evidence from binding and scope phenomena is provided to support such strong hypothesis. Moreover, it is shown that its interaction with the Split-Topic Hypothesis, which argues for a partition of the focus-topic structure of the sentence, can explain some otherwise surprising asymmetries. Finally, it is argued for an integration of discourse and modality aspects in the computation of binding and scope facts. Key words:dislocation, topic, binding, scope, reconstruction. Resum En aquest article s’argumenta que els dislocats s’interpreten en la seva posició superficial. S’aporten evidències del lligam i dels fenòmens d’abast que donen suport a una hipòtesi tan forta. A més, es mostra que la seva interacció amb la Hipòtesi del Tòpic Dividit, que proposa una partició de l’estructura de tòpic-focus de l’oració, pot explicar altres asimetries, d’altra banda sorprenents. Finalment, es defensa la integració d’aspectes del discurs i de la modalitat en la computació del lligam i dels fenòmens d’abast. Paraules clau:dislocació, tòpic, lligam, abast, reconstrucció. *This is a shortened English version of the paper I presented at the IX Colloquium on Generative Grammar. Thanks are due for comments to Albert Fontich, Gemma Rigau, and Maria Luisa Zubizarreta. I also thank the audience of the Seminari de Sintaxi i Semàntica, where a version of this work was presented. All remaining errors are my own. The preparation of this paper was supported by grant 1999/SGR/00113 from the Generalitat de Catalunya awarded to the Grup de Gramàtica Teòrica of the UAB. CatWPL 7, 1999239-252 Nihil est in LF quod prius non fuerit in SS* Xavier Villalba Universitat Autònoma de Barcelona. Departament de Filologia Catalana 08193 Bellaterra (Barcelona). Spain [email protected]b.es http://ggt.uab.es/xavier Received: December 13th 1998 Accepted: March 17th 1999 Table of Contents CatWPL 7 239-252 13/6/00 12:34 Página 239 1. The Split-Topic Hypothesis 2. Rigid binding/scope relations 3. Apparent reconstruction 4. Conclusion References It is a well-known fact that dislocates behave as if they were in situ for binding purposes:1 (1) a. La Maria ja no confia en ella mateixa the Maria already not relies in herself b. En ella mateixa,la Maria ja no hi confia in herself the Maria already not LOC relies c. La Maria ja no hi confia, en ella mateixa the Maria already not LOC relies in herself This behavior led several authors to the conclusion that dislocated elements are literally reconstructed at LF (Cinque 1983, 1990; see also Saito 1989 for scrambling). In this paper I will provide empirical evidence running against such a proposal. I will show that dislocates are interpreted in their S-Structure position, dispensing with syntactic reconstruction. Such strong hypothesis nicely correlates with an independent analysis of the partition of the informational content of sentence: the Split-Topic Hypothesis (see Villalba 1998, in press). The next section will be devoted to offering a snapshot of it. In section 2, some evidence will be offered supporting an S-Structure based interpretation of dislocation and scrambling. Finally, in section 3, I will address some counterexamples to this analysis, giving them a principled explanation in terms of independent semantic and discourse factors. 1. The Split-Topic Hypothesis Throughout this paper I will assume the Split-Topic Hypothesis (see Villalba 1998, in press), namely that the basic difference between Clitic Right Dislocation (CLRD) and Clitic Left Dislocation (CLLD) lies in the different position they occupy in the clause, not in the level at which they move (cf. Kayne 1994).2Whereas CLRD overtly moves to the specifier of the Internal Topic Phrase (IntTopP) just over the VP, CLLD overtly moves to a higher one, that of the External Topic Phrase (ExtTopP) in the CP area (see Villalba 1998, in press; cf. Rizzi 1997). Schematically:3 240 CatWPL 7, 1999 Xavier Villalba 1. Throughout this paper the bound reading is marked by means of italics. Unless otherwise noted, all judgments are referred to the marked reading. 2. I assume without discussion that dislocation is derived by movement. See Kayne (1994), Rizzi (1997), and Villalba (1998, in press) for arguments favoring a movement analysis. It is also controversial whether the clitic is just a visible trace of the dislocate or rather the head of a projection including the dislocate (before it moves). I remain neutral on this issue here, since it does not affect the core of the analysis. 3. I remain neutral on the respective position of ExtTopP and CP because it is quite controversial and it doesn’t affect the discussion. The reader is referred to Rizzi (1997) for a comprehensive analysis of the set of elements occupying the left periphery of the sentence. CatWPL 7 239-252 13/6/00 12:34 Página 240 I cannot argue at length for this structure, so the reader is referred to Villalba (1998, in press) for details and arguments. In any case, in section 2 empirical evidence will be provided that right dislocates occupy a lower position than left dislocates in the sentence tree. We can formalize this process in terms of feature attraction. Let us assume, following ideas by Choi (1996), that sentence elements are specified for informational features [new], [prominent].4If we combine these features, we obtain: all [-new] elements —clitic left and right dislocates— move to IntTopP for checking such a feature; later on, those [-new] elements also marked as [+prominent] —clitic left dislocates— will move to ExtTopP for checking purposes (see Postal 1991 for a previous proposal that CLLD is derived via CLRD; cf. Zubizarreta 1998). A similar derivation extends to focus (i.e. [+new]) elements, under the assumption that two Focus Phrases exist: all [+new] elements —contrastive and non-contrastive focus— move to FocP for checking such a feature; later on, those [+new] elements also marked as [+prominent] —contrastive focus— will move to CP for checking purposes (for the existence of a FocP see Belletti and Shlonsky 1995, and references cited therein). See Villalba (in press) for details. Nihil est in LF quod prius non fuerit in SS CatWPL 7, 1999 241 4. Things are more complicated, however. For the purposes of this paper, the feature [new] must be interpreted in its habitual sense, and [prominent] as an abbreviation of «informationally prominent in the discourse». (2) CP ExtTopP CLLD IP ... FocP IntTopP CLRD VP +new -new +prominent contrastive focus CLLD -prominent non-contrastive focus CLRD CatWPL 7 239-252 13/6/00 12:34 Página 241 2. Rigid binding/scope relations Languages like Catalan have been considered to show ambiguous scope relations, without taking into account the facts concerning dislocation. In general, it has been assumed that CLLD and CLRD involved reconstruction at LF (see Cinque 1983, 1990), but no exhaustive examination of binding and scope facts have been carried out. In this section I will provide some data arguing for a scope rigid analysis of Catalan, at least with respect to dislocation. In short, it will be shown that the binding/scope possibilities of left and right dislocates are limited to their surface position (see the tree in (2)). Those cases that do not show such rigidity will be dealt with in section 3, where they will be considered a by-product of independent discourse factors. 2.1. Pronouns bound by quantifiers In Catalan, a pronoun within a dislocate may be bound by a quantifier regardless of its original position: (3) a. No la penso discutir amb tots els professors,la seva manera not her think discuss with all the teachers the his way d’ensenyar. of-teach ‘I won’t discuss their teaching method with all the teachers.’ b. No ho penso discutir amb cada professor, com pro ensenya not it think-I discuss with every teacher how (s)he teaches ‘I won’t discuss how (s)he teaches with every teacher.’ c. Enlluerna a qualsevol pare, el seu fill dazzles to any father the his son ‘His son dazzles any father.’ It is a mystery for the LF-reconstruction approach why none of these sentences has a well-formed reconstructed counterpart: (4) a. *No penso discutir la seva manera d’ensenyar amb tots els not think discuss the his way of-teach with all the professors. teachers b. *No penso discutir com pro ensenya amb cada professor. not think discuss how (s)he teaches with every teacher c. *El seu fill enlluerna a qualsevol pare. the his son dazzles to any father These sentences are, descriptively speaking, instances of weak crossover (WCO). 242 CatWPL 7, 1999 Xavier Villalba CatWPL 7 239-252 13/6/00 12:34 Página 242 Let us consider now the converse situation, namely one having a right-dislocated QP and a pronoun in the core of the sentence: (5) a. *La seva mare va aconseguir acompanyar tothom the his mother PAST manage accompany everybody b. *La seva mare el va aconseguir d’acompanyar, a tothom the his mother him PAST manage accompany to everybody c. A tothom el va aconseguir acompanyar la seva mare to everybody him PAST manage accompany the his mother ‘His mother managed to accompany everybody.’ (6) a. *La seva mare va decidir acompanyar a cada nen the his mother PAST decide accompany to every child b. *La seva mare va decidir acompanyar-lo, a cada nen the his mother PAST decide accompany-him to every child c. A cada nen va decidir acompanyar-lo la seva mare to every child PAST decide accompany-him the his mother ‘His mother decided to accompany every child.’ The bound reading is not available if the quantifier surfaces to the right of the pronoun (a WCO violation). If the quantifier precedes the pronoun the bound reading is possible. Two points are worth commenting on here. On the one hand, it seems clear that the radical reconstruction approach cannot give a proper answer to these asymmetries. On the other hand, note that the different behavior of left and right dislocates suggests they occupy different position in the tree structure. Similar data are reported in Hindi —exs. (7)-(8)— and Basque —exs. (9)- (10)—, two SOV languages. In these languages, postverbal elements are topics, in many respects equivalent to Romance CLRD (all Hindi data in this paper come from Mahajan 1998, and all Basque data from Albizu n.d.): (7) a. *uske bhaai-ne har ek aadmii-ko maaraa. his brother-ERG every man-OBJ hit-PERF ‘His brother hit everyone.’ b. har ek aadmii-ko uske bhaai-ne maaraa. every man-OBJ his brother-ERG hit-PERF ‘???Everyone, his brother hit.’ c. *uske bhaai-ne maaraa har ek aadmii-ko. his brother-ERG hit-PERF every man-OBJ ‘His brother hit everyone.’ (8) a. raam-ne har ek aadmii-ko lɔtaaii uskii kitaab. Ram-ERG every man-DAT return-PERF his book ‘Ram returned every man his book.’ Nihil est in LF quod prius non fuerit in SS CatWPL 7, 1999 243 CatWPL 7 239-252 13/6/00 12:34 Página 243 b. *raam-ne uske maalik-ko dii har ek kitaab. Ram-ERG its owner-DAT give-PERF every book ‘Ram gave every book to its owner.’ (9) a. Munizipalak ume bakoitzari bueltatu dio bere bizikleta. policeman-ERG child each return AUX his bike b. *Munizipalak bere bizikleta bueltatu dio ume bakoitzari. ‘The policeman returned each child his bike.’ (10) a. Pellok txeke bakoitza bueltatu dio bere jabeari. Pello-ERG check each return AUX its owner b. *Pellok bere jabeari bueltatu dio txeke bakoitza. Pello-ERG its owner return AUX check each ‘Pello returned each check to its owner.’ Two generalizations follow from these data: Generalization 1: Dislocates are not interpreted as if they were in situ, but rather in their SS position Generalization 2: Left dislocates occupy a higher position in the structure than right dislocated ones. 2.2. Principle C The hypothesis that dislocates are reconstructed at LF for interpretation purposes runs into problems when the following paradigm is taken into account: (11) a. *pro va dir convençuda les mentides que la Maria va PAST say convinced the lies that the Maria PAST inventar. invent b. *pro no va confirmar després les aptituds que la Maria not PAST confirm later the aptitudes that the Maria apuntava de jove. suggested of young (12) a. [Les mentides que la Maria] va inventar, pro les va the lies that the Maria PAST invent them PAST dir convençuda. say convinced b. [Les aptituds que la Maria apuntava de jove], pro no les the aptitudes that the Maria suggested of young not them va confirmar després. PAST confirm later 244 CatWPL 7, 1999 Xavier Villalba CatWPL 7 239-252 13/6/00 12:34 Página 244 (13) a. *pro les va dir convençuda, [les mentides que la Maria them PAST say convinced the lies that the Maria va inventar] PAST invent b. *pro no les va confirmar després, [les aptituds que la not them PAST confirm later the aptitudes that the Maria apuntava de jove] Maria suggested of young The sentences in (11) are standard examples of principle C violations. Under the hypothesis that dislocates are interpreted as if they were reconstructed in their original position at LF, we expect both (12) and (13) to be equally out. However, even though this prediction is borne out by the right-dislocated version, it is not by the left-dislocated one. Note that the contrast is the expected one under the analysis proposed here: both non dislocated constituents and right-dislocated ones surface in a position under the c-command domain of the pronoun, which makes coreference impossible. In the case of CLLD, the dislocate occupies a position higher than the pronoun, allowing coreference. Again, this behavior extends to Hindi and Basque (see also Villalba in press, and Cecchetto 1999 for similar facts in Spanish and Italian, respectively): (14) a. *us-ne siitaa-ko [tumhaaraa raam-ko likhaa hua pətr] he-ERG Sita-DAT your Ram-DAT written be-PERF letter dikhaayaa show-PERF b. [tumhaaraa raam-ko likhaa hua pətr] us-ne siitaa-ko your Ram-DAT written be-PERF letter he-ERG Sita-DAT dikhaayaa show-PERF c. *us-ne siitaa-ko dikhaayaa [tumhaaraa raam-ko likhaa he-ERG Sita-DAT show-PERF your Ram-DAT written hua pətr] be-PERF letter ‘*He showed a letter written by you to Ram to Sita.’ (15) a. *Bera ez du gonbidatu [Jon estimatzen duen] neskak he-ABS not AUX invited Jon-ABS like AUX girl-ERG b. [Jon estimatzen duen] neskak ez du bera gonbidatu Jon-ABS like AUX girl-ERG not AUX he-ABS invited ‘The girl that liked Jon didn’t invite him.’ From these data, it seems we can maintain the generalizations raised in the previous section: Nihil est in LF quod prius non fuerit in SS CatWPL 7, 1999 245 CatWPL 7 239-252 13/6/00 12:34 Página 245 Generalization 1: Dislocates are not interpreted as if they were in situ, but rather in their SS position Generalization 2: Left dislocates occupy a higher position in the structure than right dislocated ones. 2.3. Scope interactions Consider the following interchanges: [Diuen que quan greixem fem servir un pot de greix per màquina] ‘They say that when we grease, we use one can of grease per machine’ (16) a. Mentida! Aquest matí amb un pot de greix, hi he greixat totes les màquines. ‘That’s a lie! This morning with a can of grease I’ve greased all the machines.’ b. #És veritat! Aquest matí amb un pot de greix, hi he greixat totes les màquines ‘That’s true! This morning with a can of grease I’ve greased all the machines.’ The sentence in square brackets provides a context in which more than one can of grease is available, hence favoring the distributivity reading, i.e. ∀ > ∃. In (16), the answerer breaks this presupposition, implying that only one can is available, i.e. ∃ > ∀; since the sentence is felicitous, we can conclude that this reading, which corresponds to the linear order of the quantifiers is available. This sentence tells us nothing about the inverse scope reading, but compare with (16). Here the distributivity reading favored by context —i.e. ∀ > ∃— is confirmed by the answerer, but as the infelicity of the sentence shows, the structural configuration of quantifiers does not provide such a reading. Consider now: [Diuen que quan greixem fem servir un pot de greix per màquina] ‘They say that when we grease, we use one can of grease per machine’ (17) a. #Mentida. Aquest matí hi he greixat totes les màquines, amb un pot de greix. ‘That’s a lie! This morning I’ve greased all the machines with a can of grease.’ b’.És veritat. Aquest matí hi he greixat totes les màquines, amb un pot de greix ‘That’s lie! This morning I’ve greased all the machines with a can of grease.’ 246 CatWPL 7, 1999 Xavier Villalba CatWPL 7 239-252 13/6/00 12:34 Página 246 Again, the sentence between brackets provides a context where more than one can of grease is available. In (17) the context introduced by the answerer favors ∃>∀, and the position of quantifiers implies the pragmatically odd ∀ > ∃. Since the sentence is not felicitous, it seems clear that only the latter reading is available. This is confirmed by (17), where ∀ > ∃is both contextually and structurally favored. Again, such scope rigidity facts are found in scrambling languages, like Hindi or Hungarian: (18) a. raam sab-ko dikhaayegaa tiin kitaabe. [everyone>three] Ram everyone-DAT show(FUT) three books ‘Ram will show three books to everyone.’ b. sab khariide ge tiin ciize. [three>everyone] everyone-DAT buy(FUT) three things ‘Everyone will buy three things.’ (19) a. Hatnál több ember hívott fel mindenkit. [more than six>everyone] six-than more man called up everyone-ACC ‘More than six men phoned everyone.’ b. Mindenkit hatnál több ember hívott fel. [everyone>more than six] everyone-ACC six-than more man called up ‘More than six men phoned everyone.’ None of these facts are expected if dislocation/scrambling is reconstructed at LF, but straightforwardly follow from a surface-based approach. 3. Apparent reconstruction 3.1. Principles A and B The most celebrated examples for a reconstruction analysis of dislocated have to do with binding of anaphors and pronominals: (20) a. En ella mateixa,la Maria ja no hi confia. in her same the Maria already not LOC rely ‘On herself, Maria doesn’t rely anymore.’ (21) a. *A ella,la Maria la va veure de seguida. to her the Maria her PAST see of followed ‘*To her, Maria saw immediately.’ We will not discuss them, since there is a huge amount of work offering an independently motivated account for such facts without resort to LF reconstrucNihil est in LF quod prius non fuerit in SS CatWPL 7, 1999 247 CatWPL 7 239-252 13/6/00 12:34 Página 247