RIVISTA DI GRAMMATICA GENERATIVA Volume 26, anno 2001 Direzione: Guglielmo Cinque (Università di Venezia) Luigi Rizzi (Université de Genève e Università di Siena) Comitato di lettura: Manuela Ambar (Univ. de Lisboa) - Paola Benincà (Università di Padova) - Adriana Belletti (Università di Siena) - Luciana Brandi (Univ. di Firenze) - Luigi Burzio (The John Hopkins Univ.) - Noam Chomsky (MIT) - Patrizia Cordin (Univ. di Trento) - Violeta Demonte (Univ. Autonoma de Madrid) - Alessandra Giorgi (Università di Bergamo) - Giorgio Graffi (Univ. di Verona) - Richard Kayne (New York University) - Michael Kenstowicz (MIT) - Giulio Lepschy (Univ. of Reading) - Giuseppe Longobardi (Univ. di Trieste) - Lidia Lonzi (Univ. di Milano) - Maria Rita Manzini (Università di Firenze) - Joan Mascaró (Univ. Autonoma de Barcelona) - Marina Nespor (Univ. Di Ferrara) - Jean-Yves Pollock (Univ. de Rennes II) - Annarita Puglielli (Univ. di Roma) - Andrew Radford (Univ. of Essex) - Lorenzo Renzi (Univ. di Padova) - Alain Rouveret (Univ. de Paris VIII) - Leonardo Savoia (Univ. di Firenze) - Sergio Scalise (Univ. di Bologna) - Laura Vanelli (Univ. di Padova) - Jean-Roger Vergnaud (Univ. of Southern California) L‟INTERPRETAZIONE MORFOSEMANTICA DEL MODO CONGIUNTIVO IN ITALIANO E IN TEDESCO 1 Paolo CHINELLATO 0. Introduzione In questo lavoro propongo che l‟interpretazione semantica e il comportamento sintattico del modo congiuntivo siano strettamente legati alla sua costruzione morfologica e al numero di tratti che i morfemi riescono ad esprimere. Più precisamente, quello che dimostrerò attraverso la comparazione di dati del tedesco, dell‟italiano standard e di due varietà dialettali dell‟italiano (veneziano e bellunese), è che quando siamo in presenza di una forma sincretica che appartiene a due modi verbali diversi, il parlante interpreta tale forma come quella appartenente al modo meno marcato. Il lavoro è organizzato nel seguente modo: nella sezione 1 si discute un caso di selezione del modo congiuntivo in tedesco, più precisamente il discorso indiretto. Dopo aver esposto i casi di selezione di entrambi i congiuntivi, seguendo Cinque (1999) ho attribuito al congiuntivo presente (o congiuntivo1) il tratto [+evidenziale], e il tratto [+valutativo] al congiuntivo imperfetto (o congiuntivo 2); in 1.1 il confronto tra la coniugazione del congiuntivo e quella dell‟indicativo mostreranno che le forme che vengono interpretate dai parlanti come appartenenti al modo congiuntivo sono solamente quelle che possiedono un morfema di modo che veicola i tratti presentati al punto1. Nella sezione 2 tratterò un caso di selezione del modo congiuntivo in italiano, più precisamente la possibilità dei verba dicendi di selezionare un congiun1 Questo lavoro nasce come approfondimento di alcuni temi già presenti nella mia Tesi di Laurea. Durante quel periodo, il mio lavoro è stato profondamente influenzato dal lavoro di Alessandra Giorgi alla quale sono estremamente debitore. Ringrazio tutti coloro che hanno letto o discusso con me parti di questo lavoro: Paola Benincà, Gehrard Brugger, Guglielmo Cinque, Federico Damonte, Giuliana Giusti, Rui Riberto Marques, Nicola Munaro, Fabio Pianesi e Cecilia Poletto. Un ringraziamento speciale va ad Anna Cardinaletti, che ha seguito questo lavoro dall‟inizio, fornendomi preziosi consigli ed indicazioni per la stesura di ogni versione del lavoro. 3 Paolo Chinellato tivo: in 2.1 presenterò i tratti morfosemantici del modo congiuntivo italiano, proponendo che questo modo (insieme al modo condizionale), se selezionato da questa classe verbale, veicola il tratto [+valutativo], come il congiuntivo imperfetto tedesco; seguirà in 2.2 l‟analisi delle forme verbali, dove proporrò di considerare la desinenza della prima persona plurale –iamo come nuovo morfema di accordo e in 2.3 tratterò i casi ambigui nell‟interpretazione dei due modi; nella sezione 3 introdurrò il Principio di Sostituzione (Pds), il principio che regola l‟interpretazione di una forma sincretica come quella appartenente al modo meno marcato e che attua la sostituzione di tale forma con una più marcata, se il contesto lo richiede. In 4 darò degli argomenti a sostegno di tale principio: osserverò per prima cosa due dialetti italiani settentrionali, il veneziano e il bellunese. Il veneziano (4.1.1) si comporta come l‟italiano standard e applica il PdS in alcuni casi in cui la forma del congiuntivo è identica all‟indicativo, mentre il Bellunese (4.1.2) non lo applica proprio perché avendo due forme distinte tra indicativo e congiuntivo non ne ha bisogno. In 4.2 infine considererò il caso della cancellazione del complementatore che in Italiano Standard come prova dell‟interpretazione di tali forme sintetiche come forme non marcate: infatti, la cancellazione è impossibile con forme di congiuntivo uguali all‟indicativo. La sezione 5 concluderà il lavoro. 1. Un caso di selezione del modo congiuntivo in tedesco In tedesco le classi dei verba dicendi e verba sentiendi selezionano una frase subordinata al congiuntivo, anche se l‟indicativo è ugualmente ammesso: Er sagt/glaubt, dass Paul aus Italien kommt/komme/käme. 2 (lett.: lui dice/crede che Paolo dall‟Italia viene/venga/IND/KON1/KON2) il congiuntivo tedesco è ammesso quando il soggetto della frase principale e quello della subordinata sono diversi dal parlante (2), quando solo il soggetto della principale è diverso dal parlante (3), ma non quando il soggetto della principale coincide col parlante (4) o quando entrambi i soggetti delle frasi coincidono col parlante (5): Sein Arzt sagt, er müsse ausspannen, diesmal sei ernstlich krank (Il suo medico dice, lui deve / KON 1 riposarsi, questa volta è / KON 1 lui seriamente malato) Er sagte, dass ich krank sei (Lui disse che io malato sono / KON1) 2 4 D‟ora in poi mi riferirò al congiuntivo presente tedesco con la sigla KON 1 e al congiuntivo imperfetto con la sigla KON 2. L’interpretazione morfosemantica del modo congiuntivo in italiano e in tedesco * Ich glaube, dass er komme (Io credo che lui venga / KON 1) * Ich glaube, dass ich krank sei (* Io credo che io venga / KON 1) La frase (4) mostra chiaramente che il KON 1 non può segnalare nessun atteggiamento del parlante sul contenuto della frase, perché se così fosse sarebbe grammaticale come lo è (6): Er glaubt, dass Paul komme (Lui crede che Paolo venga / KON 1) Il KON 1 sembra quindi essere un segnale del discorso riportato; sembra esprimere, attraverso il morfema di modo, un tratto che chiamerò, seguendo 3 l‟impostazione di Cinque (1999), [+evidenziale] . Notiamo che l‟uso del congiuntivo tedesco differisce dall„italiano nel caso in (4), dove la corrispondente italiana è grammaticale, Io credo che lui venga ma è identico nel caso di (5), dove anche la corrispondente italiana (8) è completamente agrammaticale e deve essere sostituita con (9): * Io credo che io venga Io credo di venire La differenza tra (4) e (7) chiarisce la diversa natura del modo congiuntivo nelle due lingue. In italiano standard la classe dei verba sentiendi richiede il modo congiuntivo (consecutio modorum) come si vede dalla differenza di accettabilità tra 4 (10a) e (10b) (si veda la sezione 2.1): a. Credo che lui venga b. ? Credo che (lui) viene Per quanto riguarda la concordanza dei tempi, il congiuntivo presente esprime un rapporto di contemporaneità (11a) e di posteriorità (11b) con la frase principale, ma non contemporaneità con il verbo della principale al passato (12) (si veda la sezione 2.1): 3 4 Il nome [+evidenziale] corrisponde al termine [+riportato] che avevo usato in Chinellato (2000). Per una dettagliata analisi morfosintattica del modo congiuntivo italiano, si veda Giorgi & Pianesi (1997), cap. 5. 5 Paolo Chinellato a. Credo che in questo momento sia al ristorante. b. Credo che lui venga/verrà domani. * Credevo che lui venga Inoltre, come abbiamo visto, in italiano non è possibile avere un congiuntivo con una frase subordinata il cui soggetto coincide col soggetto della principale e con il parlante, come si vede dall„agrammaticalità di (13) e (16): * Credo che io venga Credo di venire * Credevo che io venissi Credevo di venire In tedesco il congiuntivo non è legato né a concordanza di modo né di tempo come si vede da (17), Hans glaubt/glaubte, dass Paul krank sei/ist (Gianni crede/credeva che Paolo malato sia/fosse/IND) ma è sempre agrammaticale con la prima persona nella frase principale (si veda (7) e (8)). Come si vede in (18)/(20), le forme del KON 1 sono uguali al presente indicativo alla prima persona singolare e plurale e alla terza persona plurale: Er sagt/glaubt, dass ich aus Italien komme (Lui dice/crede che io dall‟Italia vengo / venga / IND = KON 1) Er sagt/glaubt, dass wir aus Italien kommen (Lui dice/crede che noi dall‟Italia veniamo / IND = KON 1) Er sagt/glaubt, dass sie aus Italien komme (Lui dice/crede che loro dall‟Italia vengono / vengano / IND = KON 1) Nei casi (18)–(20), in cui le forme del KON 1 coincidono con le forme 5 dell‟indicativo, è possibile un‟alternanza con il KON 2 : Er sagt/glaubt, dass ich aus Italien käme (lui dice/crede che io dall„Italia vengo / venga / KON 2) 5 6 La grammatica tradizionale propone una sistematica sostituzione con il KON 2 ogni volta che il KON 1 è uguale all‟indicativo: come si vedrà in seguito, la regola della sostituzione proposta dalla grammatica tradizionale non sembra essere in grado di soddisfare i casi in (26) e (27). Per un‟analisi alternativa del congiuntivo tedesco si veda Chinellato (2000). L’interpretazione morfosemantica del modo congiuntivo in italiano e in tedesco (22) Er sagt/glaubt, dass wir aus Italien kämen (lui dice/crede che noi dall‟Italia veniamo / KON 2) (23) Er sagt/glaubt, dass sie aus Italien kämen (lui dice/crede che loro dall‟Italia vengono / vengano / KON 2) Osservando (24) e (25) (24) Hans sagt, dass seine Eltern in der Schweiz sind/seien/wären, aber das ist nicht wahr. (Gianni dice che i suoi genitori in Svizzera sono / IND / KON 1 / KON 2 /, ma questo non è vero) (25) Hans sagt, dass ich ihn geschlagen habe/hätte, aber das ist nicht wahr. (Gianni dice che io lui colpito ho / IND = KON 1 / KON 2, ma questo non è vero) notiamo che il KON 2 sottolinea la posizione del parlante riguardo a ciò che ha detto il soggetto della frase principale: questo tratto che chiamerò, sempre seguendo Cin6 que(1999) [+valutativo] , è espresso dalla morfologia passata del congiuntivo unita con il morfema di modo. Questa interpretazione del KON 2 è confermata dal fatto che l‟alternanza del KON 1 con il KON 2 alla prima persona singolare e plurale e alla terza persona plurale non è sempre possibile. Ci sono dei casi in cui il KON 2 risulta piuttosto strano, come in (26) e (27): (26) Hans sagte, dass seine Eltern in der Schweiz sind/seien/??wären, und er hat Recht, weil ich sie gesehen habe (Gianni ha detto che i suoi genitori in Svizzera sono / IND / KON 1 / KON 2 e lui ha ragione perché io li ho visti) (27) Hans sagte, dass ich geschlagen habe/??hätte, und er hat recht (Gianni ha detto che io lui colpito ho / IND = KON 1 / KON 2) Notiamo che le frasi con il KON 2 sono quelle in cui esso esprime il dubbio che viene sottolineato dalla frase seguente, come si vede da (24) e (25), mentre quando c‟è un contesto di certezza come in (26) e (27) il KON 2 risulta non accettabile. L‟indicativo, poiché non contiene nessun morfema di modo, si presenta come l‟opzione meno marcata. Osservando le frasi (26) e (27), possiamo stabilire l‟inaccettabilità del KON 2 (si veda la sezione 1.1.2) considerando la sua presenza insieme alle altre due forme in 6 Il nome [+valutativo] corrisponde al termine [+distanza] che avevo usato in Chinellato (2000). 7 Paolo Chinellato termini di opzione [±marcata]. In (26) il KON 2 è inadeguato perché ha un valore più marcato rispetto al contesto dell‟enunciazione e lo stesso vale per (27): questo valore viene veicolato dal tratto [+valutativo] che risulta inadeguato nei contesti di certezza espressi dalle due frasi. 1.1. I tratti morfosemantici del congiuntivo tedesco 1.1.1. Il konjunktiv 1 Confrontiamo la coniugazione del konjunktiv 1con il presente indicativo: (28) Indicativo Presente TEMPO MODO kommkommkomm- RADICE TEMPO MODO Ich komm- Du Er/sie Wir kommkommkomm- ee- Ihr Sie kommkomm- Ich Du Er/sie Wir Ihr Sie RADICE kommkommkomm- ACCORDO e st t en t en (29) Konjunktiv 1 e ACCORDO e st en t en In (29) il morfema –e– distingue il modo congiuntivo dall‟indicativo (i morfemi di modo saranno sempre segnati in grassetto): il morfema è presente alla seconda persona singolare e plurale e alla terza persona singolare. Riassumendo, la distribuzione del tratto [+evidenziale] è la seguente: 8 L’interpretazione morfosemantica del modo congiuntivo in italiano e in tedesco (30) PERSONA Ich Du Er Wir Ihr Sie MORFEMA DI MODO + + + - MORFEMA DI ACCORDO + + + + + 1.1.2. Il konjunktiv 2 Consideriamo ora il konjunktiv 2 confrontandolo con la struttura del preterito: Preterito del verbo gehen (andare) (31) Ich Du Er/sie Wir Ihr Sie TEMPO gingginggingginggingging- MODO Ø Ø Ø Ø Ø Ø ACCORDO Ø st Ø en t en TEMPO MODO ACCORDO gingginggingginggingging- eeeØ eØ Ø st Ø en t en (32) KON 2 del verbo gehen (andare) Ich Du Er/sie Wir Ihr Sie Come si vede in (32), la distribuzione del tratto [+ valutativo] è la seguente: 9 Paolo Chinellato (33) PERSONA Ich Du Er /sie Wir Ihr Sie MODO + + + + + ACCORDO + + + + 1.2. Casi ambigui Le frasi (34) e (35), inoltre, (34) Hans sagte, dass wir nach Italien gingen (35) Hans sagte, dass sie nach Italien gingen non vengono interpretate come in (36), ma come in (37); la forma gingen viene riconosciuta come preterito e non come KON 2: (36) Gianni ha detto che noi/loro andiamo/vanno in Italia (37) Gianni ha detto che noi/loro andavamo/andavano in Italia Le voci verbali della prima e terza persona plurale si comportano come le voci verbali dei verbi cosiddetti “deboli”, che hanno la stessa forma per il preterito e per il KON 2 , come si vede nella frase (38): la forma studierte non viene riconosciuta come KON 2, ma solamente come preterito e accettando la frase solo con l‟interpretazione data in (39). In entrambi i casi, l‟unica forma ammessa di KON 2 è 7 la WF come appare in (40) : (38) Hans sagte, dass ich Mathematik studiere/*studierte (Gianni ha detto che io matematica studio IND = KON 1/ PRET = KON 2) (39) Gianni ha detto che studiavo matematica (40) Hans sagte, dass ich Mathematik studieren würde 7 10 Se il verbo subordinato è un verbo regolare (o “debole”), la sostituzione proposta dalla grammatica tradizionale non sarà sufficiente, perché il KON 2 è uguale al preterito; la grammatica tradizionale propone di sostituire le forme ambigue con la perifrasi chiamata “würde-Form” (d‟ora in poi WF). L’interpretazione morfosemantica del modo congiuntivo in italiano e in tedesco Pertanto, al KON 2 le voci verbali della prima e terza persona singolare dei verbi irregolari e tutte le forme dei verbi regolari sono meno marcate perché in esse non è presente il morfema -e- che esprime insieme alla morfologia passata il tratto [+valutativo]. Un altro caso in cui, quando la forma del KON 2 è uguale al preterito, essa viene interpretata come indicativo è il periodo ipotetico. Le frasi (41) e (42), (41) Wenn wir einen Brief schrieben, schrieben wir mit einen blauen Kugelschreieber. (42) Wenn wir ein neues Auto kauften, kauften wir eine BMW vengono interpretate come (43) e (44) e non come (45) e (46): (43) Ogni volta che scrivevamo una lettera, scrivevamo con una penna blu (44) Ogni volta che compravamo una nuova auto, compravamo una BMW (45) Se scrivessimo una lettera, scriveremmo con una penna blu (46) Se comprassimo un‟auto nuova, compreremmo una BMW Anche qui, l‟unica possibilità di avere un periodo ipotetico è quello di sostituire il KON 2 con la WF: (47) Wenn wir einen Brief schreiben würden, würden wir mit einen blauen Kugelschreieber schreiben. (48) Wenn wir ein neues Auto kaufen würde, würden wir eine BMW kaufen. 2. Un caso di selezione del modo congiuntivo in italiano 2.1. I tratti del congiuntivo In questa sezione esaminerò i tratti del modo congiuntivo. Chiamerò tratto di concordanza quel tratto presente nel verbo della frase principale e della subordinata che lega tra loro i due verbi in un rapporto di tempo e/o di modo. Distinguerò, quindi, il tratto di concordanza temporale [+CT] dal tratto di selezione di modo: per concordanza temporale mi riferisco ad una relazione temporale definita del verbo principale che obbliga il verbo subordinato a concordarsi nello stesso tempo e al modo congiuntivo; per selezione di modo intendo ciò che ho illustrato nella precedente sezione per il tedesco e cioè la selezione del tratto [+evidenziale] (KON 1) o [+valutativo] (KON2). 11 Paolo Chinellato In Italiano, il verbo credere al passato instaura un legame temporale e di modo con il verbo subordinato (che deve essere un congiuntivo imperfetto). Il tratto [+CT] è presente nell‟italiano contemporaneo: (49) Gianni crede che Maria sia incinta. (50) Gianni credeva che Maria fosse incinta. (51) *Gianni credeva che Maria sia incinta. L‟agrammaticalità di (51) dipende dal fatto che il verbo subordinato non possiede il tratto [+passato], ma [+presente], che non concorda però con il tempo della frase principale. In (52) la frase è accettabile perché, nonostante il verbo non abbia instaurato la selezione modale con il verbo principale, possiede comunque il tratto [+passato] che si concorda con il tempo della frase principale: (52) ? Gianni credeva che Maria era incinta Considererò ora il verbo dire che seleziona il congiuntivo, che non è obbligatorio per ragioni di consecutio temporum, ma mostra un comportamento analogo al discorso indiretto tedesco. Alcuni linguisti sostengono che il congiuntivo in italiano non è ammesso se viene selezionato da un verbo di affermazione e se il soggetto è 8 espresso. Per questi autori (53) è agrammaticale : (53) Marco e Maria hanno detto che il signor Bianchi abbia lavorato in Africa Serianni (1988) e Portner (1997) sostengono il contrario portando come prova un esempio tratto da un testo letterario. Riporto l‟esempio parafrasato in italiano con9 temporaneo presente nel lavoro di Portner (1997) : (54) I ragazzi dicono che suo padre, il farmacista, stia sperimentando su di lui gli effetti di un nuovo sciroppo purgativo. Riscriviamo lo stesso periodo (con alcune modifiche) al passato con l‟indicativo, col congiuntivo e il condizionale nella frase subordinata: (55) Marco e Maria hanno detto che il padre di Gianni ha sperimentato un nuovo sciroppo. 8 9 12 Wandruszka (1991): “Il congiuntivo si può usare anche in dipendenza da locuzioni con soggetto indeterminato dicono, si dice, c’è chi dice, con le quali il contenuto della frase subordinata può essere riportato come opinione o punto di vista esistenti, senza impegno in merito da parte del parlante stesso.” La parafrasi in italiano contemporaneo presente nel Portner (1997) è di Raffaella Zanuttini. L’interpretazione morfosemantica del modo congiuntivo in italiano e in tedesco (56) Marco e Maria hanno detto che il padre di Gianni abbia sperimentato un nuovo sciroppo. (57) Marco e Maria hanno detto che il padre di Gianni avrebbe sperimentato un nuovo sciroppo. Tutti i tre modi sono accettabili, ma se proviamo a far seguire la frase da un contesto più preciso, vediamo che sia congiuntivo e condizionale risultano molto strani: (58) Marco e Maria hanno detto che il padre di Gianni ha/??abbia/??avrebbe sperimentato un nuovo sciroppo e hanno ragione perché io l‟ho visto farlo. (59) Marco e Maria hanno detto che il padre di Gianni ha/abbia/avrebbe sperimentato un nuovo sciroppo, ma questo non è vero. I tre modi italiani, indicativo da una parte, congiuntivo e condizionale dall‟altra sembrano comportarsi rispettivamente come l‟indicativo tedesco l‟uno e il KON 2 gli altri. La stranezza del congiuntivo e condizionale in (58) dipende dalla presenza del tratto [+valutativo] che è in contrasto con il contesto della frase. 2.2. Analisi delle forme verbali L‟analisi delle forme verbali dell‟Italiano standard delle prossime sezioni segue l„impostazione di Thornton (1999): Thornton propone il seguente schema di costituzione delle forme flesse italiane, valido per l„indicativo e per il congiuntivo e per altri modi verbali: (60) a. base radicale + accordo nelle persone prima, seconda e terza singolare b. base tematica + accordo nelle persone prima e seconda plurale c. terza persona singolare + morfo di plurale = terza persona plurale In questo lavoro considererò questo schema, modificando lo status della prima persona plurale. Essa infatti risulta una forma irregolare in quanto la desinenza -iamo risulta un‟espansione della forma del congiuntivo (cfr. Tagliavini (1969), Vincent (1980), Bruni (1984)). Come dimostreremo più avanti questa forma e la forma della seconda persona singolare della prima coniugazione vengono interpretate dai parlanti dell‟Italiano Standard come indicativo: propongo, quindi, di considerare -iamo come morfema di accordo della prima persona plurale: (61) noi am-iamo nostro padre (62) Gianni pensa [che noi am-iamo nostro padre] 13 Paolo Chinellato In questo caso la –i di –iamo non viene interpretata come morfema di modo congiuntivo proprio perché non esiste in Italiano Standard una prima persona plurale 10 come in (63) tale che la forma in (62) possa sostituirla come congiuntivo : (63) *noi ama-mo nostro padre Riassumendo propongo, per la prima persona plurale, indicativo e congiuntivo, il seguente schema: (64) 1^ persona plurale: Radice + Nuovo Accordo Per l‟analisi delle forme del congiuntivo imperfetto seguirò la proposta di Thornton (1999), che ipotizza per ogni tempo e modo una base radicale e una base tematica come nello schema riportato in (60). Secondo questo schema la base radicale dell‟imperfetto è costituita dalla base tematica del presente con il morfema di tempo –v-. La base tematica è formata da questa radice e dalla vocale tematica, come si può vedere in (65) (la base radicale e tematica dell‟imperfetto sono segnate in grassetto), mentre in (66a) la radice non mostra il morfema di tempo che è caduto, ma che era presente nella forma latina di provenienza (66b): (65) [[[[am]R –a ]T –v ]R –a T] (66) a. [[[[am]R–a]T ]R –ssi T] b. [[[[am]R –a]T –v]R –sset T] 2.3. Casi ambigui Un verbo della classe dei verba sentiendi in Italiano Standard seleziona il congiuntivo nella frase subordinata: consideriamo la frase (67), che contiene un verbo alla prima coniugazione alla seconda persona singolare, che non distingue tra indica11 tivo e congiuntivo : (67) Gianni pensa che proprio in questo momento tu parli (IND = CONG) alla conferenza. Per intensificare il rapporto di contemporaneità nel presente tra la frase principale e la subordinata, (68) sembra la sostituzione migliore: (68) Gianni pensa che proprio in questo momento tu stia parlando alla conferenza. 10 11 14 In varietà non standard si hanno prime persone plurali formate secondo lo schema proposto da Thornton (1999): cfr. i tipi parlamo, vedemo, sentimo. In questa sezione mi occuperò solo dell‟Italiano Standard. I parlanti che di solito selezionano un indicativo dopo i verba sentiendi, sentono quest‟uso, cioè (67), legato alla loro varietà regionale che non prevede il congiuntivo. Ringrazio Anna Cardinaletti e Giuliana Giusti per i giudizi su questo punto. L’interpretazione morfosemantica del modo congiuntivo in italiano e in tedesco Notiamo che in presenza di un verbo irregolare con le forme di indicativo e congiuntivo distinte, la sostituzione non sembra essere necessaria come non lo è per le seconde persone delle altre coniugazioni: (69) Gianni pensa che proprio in questo momento tu vada / stia andando al cinema (70) Gianni pensa che in questo periodo tu beva / stia bevendo troppo gin. (71) Gianni pensa che in questo periodo tu dorma / stia dormendo troppo. La stessa frase (67), risulta inaccettabile quando la subordinata esprime un rapporto di posteriorità come in (72) e diventa accettabile con la sostituzione in (73): (72) *Gianni pensa che domani tu parli bene alla conferenza (73) Gianni pensa che domani parlerai bene alla conferenza Anche qui la sostituzione non sembra obbligatoria per un verbo irregolare della prima coniugazione e per le altre coniugazioni: (74) a. Gianni pensa che domani tu vada/andrai al cinema b Gianni pensa che domani tu scriva/scriverai una lettera c Gianni pensa che domani tu dorma/dormirai fino a mezzogiorno 2.4. Riepilogo In § 1 abbiamo individuato i tratti espressi dal KON 1 e dal KON 2: (75) KON 1: [+evidenziale] KON 2: [+valutativo] In 2.1 abbiamo visto che se selezionato da un verbo di affermazione (dire) il modo congiuntivo e condizionale italiani esprimono il tratto [+valutativo]. 3. Il Principio di Sostituzione Dall‟analisi dei dati considerati in tedesco (§ 1) e in italiano (§ 2) sembra pertanto che vi sia un principio che regoli l‟interpretazione di due forme morfologicamente uguali: (76) Principio di Sostituzione (PdS): Se hai una forma morfologicamente identica per due modi verbali diversi, interpreta tale forma come quella appartenente al modo meno marcato e sostituiscila con una forma più marcata se il contesto te lo richiede. In tedesco le forme in (77a, b, c) 15 Paolo Chinellato (77) a. ich hab-e [-evidenziale] b. wir hab-en [-evidenziale] c. sie hab-en [-evidenziale] vengono interpretate solo ed esclusivamente come indicativi perché non possiedono un morfema di modo come quello presente nelle altre forme verbali in (77d, e, f) e non riescono ad esprimere il tratto [+evidenziale] d. du hab-e-st [+evidenziale] e. er hab-e [+evidenziale] f. ihr hab-e-t [+evidenziale] Allo stesso modo le forme in (78) (78) a. wir ging-en [-valutativo] b. sie ging-en [-valutativo] non possedendo il morfema –e non esprimono i valori del tratto [+valutativo], come invece le forme in (79), e vengono interpretate come voci dell‟indicativo preterito: (79) a. ich ging-e b. du ging-e c. er ging-e d. ihr ging-e [+valutativo] [+valutativo] [+valutativo] [+valutativo] In italiano forme come in (80) non esprimono alcun tratto morfosemantico e non possono instaurare un rapporto di posteriorità con la frase principale; il congiuntivo 12 non riesce ad esprimere il suo tratto [+irrealis] e di conseguenza viene interpretato come la forma meno marcata, l‟indicativo presente: (80) a. *Gianni pensa che domani tu parli bene alla conferenza b. Gianni pensa che domani tu parlerai bene alla conferenza [-irrealis] [+irrealis] In (81) entrambe le forme esprimono il tratto [+irrealis] e la scelta tra le due diventa in italiano standard opzionale: (81) a. Gianni pensa che domani tu scriva/scriverai una lettera [+irrealis] b. Gianni pensa che domani tu dorma / dormirai fino a mezzogiorno [+irrealis] 12 16 Cfr. Scorretti (1991) e Poletto (1995), (2000) per il tratto [+irrealis], del congiuntivo, del futuro e del condizionale. L’interpretazione morfosemantica del modo congiuntivo in italiano e in tedesco 4. Argomenti a sostegno del principio di sostituzione 4.1. Un’osservazione comparativa 4.1.1. Il Veneziano Consideriamo una varietà dialettale come il veneziano, che ha alla prima coniugazione il modo congiuntivo uguale all‟indicativo tranne che nella prima persona singolare: (82) Mario e Maria i sa che fumo massa (Mario e Maria sanno che fumo troppo) (83) Mario e Maria i crede che mi fuma massa (Mario e Maria credono che io fumi troppo) e alla seconda e terza coniugazione la prima persona singolare e la terza singolare e plurale distinte tra indicativo e congiuntivo: (84) a. Mario e Maria i sa che mi bevo/dormo massa b. Mario e Maria i sa che iù beve/dorme massa c. Mario e Maria i sa che iori i beve / i dorme massa (85) a. Mario e Maria i crede che mi beva/dorma massa b. Mario e Maria i crede che ‟l beva/dorma massa c. Mario e Maria i crede che iori i beva / i dorma Se il verbo subordinato deve indicare posteriorità, anche in veneziano la forma corrispondente ad (86), (87), risulta inaccettabile: (86) *Gianni crede che domani tu parli bene alla conferenza (87) *Nane el crede che domàn te parli ben ala conferensa Entrambe le forme sono perfettamente grammaticali se vengono selezionate da un verbo che ammette solo l‟indicativo come sapere: (88) Gianni sa che domani tu parli alla conferenza (89) Nane sa che domàn ti te parli ben ala conferensa Le frasi (86) e (87) andranno sostituite con una forma più marcata come il futuro, che contiene il tratto [+irrealis]: (90) Gianni crede che domani parlerai bene alla conferenza (91) Nane el crede che domàn te parlerà ben ala conferensa 17 Paolo Chinellato Anche in veneziano, come in italiano standard, quando la forma uguale ai due modi è usata per indicare posteriorità (caratteristica del modo marcato, il congiuntivo), essa risulta inaccettabile perché interpretata come indicativo, cioè la meno marcata. Anche in questo caso sembra trovare applicazione il Principio di Sostituzione. 4.1.2. Il Bellunese Il Bellunese, è una varietà più conservativa del veneziano. Quando la forma del congiuntivo è distinta da quella dell‟indicativo, essa risulta grammaticale per indicare posteriorità, senza dover ricorrere alla sostituzione col futuro indicativo. In (92) l‟indicativo è inaccettabile perché il verbo credere seleziona il modo congiuntivo, (frase (93)), e il futuro indicativo risulta strano (frase (94)) poiché il verbo della frase principale ha a disposizione una forma di congiuntivo (che non è disponibile in Italiano standard e in veneziano) e quindi non ha bisogno del Principio di Sostituzione: (92) *Nani ‟l crede che domàn ti te compra an vestito novo (Gianni crede che domani tu compri / IND un vestito nuovo) (93) Nani ‟l crede che domàn ti te compre an vestito novo (Gianni crede che domani tu compri / CONG in vestito nuovo) (94) ?? Nani ‟l crede che domàn ti te comprarà an vestito novo. (Gainni crede che domani tu comprerai un vestito nuovo) 4.2. Cancellazione del complementatore e congiuntivo L‟altro argomento a sostegno dell‟interpretazione di forme uguali interpretate come appartenenti al modo meno marcato, riguarda la cancellazione del complementatore che (d‟ora in poi CD, dall‟inglese Complementizer Deletion), possibile con il modo congiuntivo, ma non con l‟indicativo: (95) Gianni crede sia malato (96) *Gianni sa è malato La CD è possibile con il congiuntivo nei seguenti contesti: (97) Gianni crede tu sia malato (verbo essere) (98) Gianni crede tu abbia ragione (verbo avere) (99) Gianni crede tu debba farlo (verbo modale) (100) Gianni crede tu possa farlo (verbo modale) 18 L’interpretazione morfosemantica del modo congiuntivo in italiano e in tedesco (101) Gianni crede tu voglia andartene (verbo modale) (102) Gianni crede tu beva troppo (2^ coniugazione) (103) Gianni crede tu dorma troppo (3^ coniugazione) Con la seconda persona singolare della prima coniugazione e la prima persona plurale di tutte le coniugazioni, abbiamo agrammaticalità come si vede in (104)/(107): (104) *Gianni crede tu fumi troppo (105) *Gianni crede noi fumiamo troppo (106) *Gianni crede noi beviamo troppo (107) *Gianni crede noi dormiamo troppo Appare evidente che queste forme vengono interpretate come indicativi, poiché non ammettono CD. Questo dato è coerente con i risultati raggiunti nella sezione 2. 5. Conclusioni In questo lavoro ho proposto che i tratti presenti nei morfemi verbali di modo intervengono nell‟interpretazione semantica del verbo all‟interno di una frase subordinata. In primo luogo ho analizzato la selezione del modo in tedesco (sezione 1) e in italiano (sezione 2). Ho rilevato che le forme coincidenti tra congiuntivo e indicativo, cioè che non possiedono alcun tratto veicolato da alcun morfema di modo, vengono interpretate come forme meno marcate, cioè forme del modo indicativo. In base a queste osservazioni ho formulato un principio che sembra regolare l‟interpretazione di tali forme, che ho chiamato Principio di Sostituzione (PdS, sezione 3): Se hai una forma morfologicamente identica per due modi verbali diversi, interpreta tale forma come quella appartenente al modo meno marcato e sostituiscila con una forma più marcata se il contesto te lo richiede. Nella sezione 4 ho presentato due argomenti a sostegno del Pds: in 4.1 ho analizzato il caso di due varietà dialettali, Veneziano e Bellunese. Il Veneziano (4.1.1) sembra comportarsi come l‟Italiano standard e applica il PdS, mentre il Bellunese (4.1.2) non lo applica proprio perché avendo due forme distinte tra indicativo e congiuntivo non ne ha bisogno. In 4.2 infine ho considerato il caso della CD in Italiano Standard e la sua impossibilità con forme di congiuntivo uguali all‟indicativo. 19 Paolo Chinellato Bibliografia Bruni, F., (1984), L’italiano. Elementi di storia delle lingue e della cultura, Torino, Utet. Chinellato, P., (1998), Su alcuni aspetti della concordanza di tempo e modo in italiano e tedesco: un’analisi nel quadro della teoria generativa, Tesi di Laurea, Università di Venezia. Chinellato, P., (2000), Il modo congiuntivo tedesco: una proposta d‟analisi, in corso di stampa. Cinque, G., (1999), Adverbs and Funcional Heads. A Cross - Linguistic Perspective, Oxford University Press. Giorgi, A., & Pianesi, F., (1997), Tense and Aspect: from Semantics to Morphosyntax, Oxford University Press. Poletto, C., (1995), Complementizer Deletion and Verb Movement, in Venice Working Papers in Linguistics, vol. 5 n. 2., pp. 49-79. Poletto C., (2000), The Higher Functional Field. Evidence from Northern Italian Dialects, Oxford University Press. Portner, P., (1997), The Semantics of Mood, Complementation and Conversational Force, in Natural Language Semantics, vol. 5, n. 2., pp. 167-212. Scorretti, M., (1991), Complementizers in Italian and Romance, Tesi di Dottorato, Università di Amsterdam. Serianni, L., (1988), Grammatica Italiana. Italiano comune e lingua letteraria. Torino, Utet. Thornton, A.M., (1999), Diagrammaticità, uniformità di codifica e morfomicità nella flessione verbale italiana, in SLI 41, Atti del XXXI Congresso di Fonologia e Morfologia dell’Italiano e dei dialetti d’Italia, Bulzoni, Roma, 483-502. Vincent, N., (1980), Words versus morphemes in morphological change: the case of Italian -iamo, in J. Fisiak (a cura di), Historical Morphology, The Hague, Mouton, 383-398. Wandruszka, U., (1991), Frasi subordinate al congiuntivo, in L.Renzi e G. Salvi (a cura di), Grande grammatica italiana di consultazione, II, pp. 117–156, Bologna, Il Mulino. 20 THE ARTICULATION OF INFLECTION IN JAMAICAN CREOLE1 Stephanie DURRLEMAN 0. Introduction This paper explores the syntax of declarative clauses in Jamaican Creole (JC), focussing on the articulation of inflection in this language. The investigation situates itself within the Principles and Parameters framework, which upholds that the study of any natural human language reveals a complex linguistic system regulated by the Principles and Parameters of Universal Grammar (UG). Under this perspective, the examination of the architecture of the clause in JC is expected to show compatibility with the very elaborate functional clause-structure provided by UG (Pollock (1989), Belletti (1990), Cinque (1999)). This research verifies the validity of such a prediction. The work is organized as follows: section 1 presents the universal clause-structure in question, section 2 concentrates on the compatibility of the articulation of inflection in JC with this sophisticated clausal architecture, section 3 turns to theoretical issues raised by this comparison, section 4 highlights the evidence from JC for the framework provided by Cinque (1999), and section 5 concludes the discussion. 1. The Architecture of the Clause 1.1. The Split-Infl Hypothesis The structure in (1) illustrates a first application of X‟-schema to both lexical and functional categories of sentential structure: 1 This paper is a reduced version of Durrleman (1999). I thank Enoch Aboh, Guglielmo Cinque, Eric Haeberli, Liliane Haegeman, and Luigi Rizzi for their comments. Thanks are also due to Deborah DePass, Leah DePass and Evelyn Miller for their intuitions on the data considered. 21 Stephanie Durrleman (1) CP C‟ Spec C° IP I‟ DP I° VP DP V‟ V° XP Although this analysis is essentially on the right track, it has had to be modified into a more articulate structure on the basis of both conceptual and empirical shortcomings: (a) Conceptually, it is undesirable to associate a single X° with independent sets of features such as tense and agreement. (b) Empirically, the splitting of this projection is forced so as to accommodate word 2 order variations such as those found between French and English : (2) AGRP AGR‟ NP AGR TP T‟ T VP… (Pollock (1989), Belletti (1990)) A recent development of clause structure has led to the postulation of additional functional projections to those in (2). Much of the evidence for this stems from a meticulous observation of the behaviour of adverbs as discussed in Cinque (1999), as well as that of bound and free functional heads expressing Tense, Modality and Aspect. The present paper situates itself within the framework provided by Cinque (1999). More specifically, it aims at testing its validity on new evidence of the third sort: i.e. free functional morphemes as found in the grammar of JC. 2 22 C.f. Pollock (1989). The articulation of inflection in Jamaican Creole 1.2. Cinque (1999) 1.2.1. A hierarchy of classes of adverbs Adverbial hierarchy is established on the observation that certain adverbs (each representative of a class) necessarily precede others. For example, consider the data in (3) and (4) from French which examines the position of déjà in relation to pas and plus: (3) a. Si tu n‟as pas déjà mangé, tu peux le prendre „If you have not already eaten, you can take it‟ b. *Si tu n‟as déjà pas mangé, tu peux le prendre „If you have already not eaten, you can take it‟ (4) a. A l‟époque, il ne possédait déjà plus rien „At the time, he did not possess already any longer anything‟ b. *A l‟époque, il ne possédait plus déjà rien „At the time, he did not possess any longer already anything‟ These examples yield the relative orders: pas > déjà and déjà > plus. Through transitivity, it is expected that plus be more structurally embedded than pas. At first sight, this may appear difficult to prove since the two cannot co-occur: (5) a. *Ils n‟ont pas plus téléphoné They haven‟t not any longer telephoned b. *Ils n‟ont plus pas téléphoné They haven‟t any longer not telephoned Despite the impossible co-occurrence of pas and plus, evidence with respect to lexical infinitives (Pollock (1989)) upholds that, as expected from the data above, pas is higher than plus in the structure: (6) a. Ne dormir plus b. Ne plus dormir (7) a. *Ne dormir pas b. Ne pas dormir The data above can be accounted for in terms of the syntactic analysis in (8): (8) [ _ [ pas _ [ plus … dormir ]]] X 23 Stephanie Durrleman This argumentation upholds both a hierarchical organisation of adverbs, the lat3 ter‟s status as XPs , as well as the presence of intervening X° positions where, for example, a lexical verb [-fin] can occur. The hierarchy of AdvPs proposed in Cinque (1999) on the basis of evidence from Romance (Italian and French), is shown to hold cross-linguistically: English, Norwegian, Bosnian/Serbo-croatian, Hebrew, Chinese, and Albanian, to name a few languages, reveal a striking consistency in the ordering of adverb classes. Indeed, despite the occasional surface variations of adverbial orders from one language to the next, it is illustrated in Cinque (1999) that, upon closer examination, these variations remain consistent with a single overall order. An example of apparent crosslinguistic variation in the ordering of adverbs may be found between English and Romance always/sempre and any longer / più: (9) John doesn‟t always win his games any longer (10) a. Gianni non vince più sempre le sue partite G. [neg] win any longer always his games b. *Gianni non vince sempre più le sue partite However, “when both always and (not..) any longer appear before the verb, their order is just like that found in Italian (Romance)” (Cinque (1999, 33): (11) a. John doesn‟t any longer always win his games b. *John doesn‟t always any longer win his games The apparent subversion of the order of adverbs in (9) can therefore be analysed as XP movement of [always win his games] across any longer as shown in (12). Movement across any longer gives rise to a slight focus on this adverb: (12) John doesn‟t [always win his games] i any longer ti Through this type of meticulous consideration of the syntactic behaviour of adverbs, the single, universal order of AdvPs is identified: (13) Frankly > fortunately > allegedly > probably > once > then > perhaps > necessarily > possibly > willingly > inevitably > cleverly > usually> again > often > quickly > already > no longer > still > always > just > soon > briefly > characteristically > almost > completely > tutto > well > fast/early > co mpletely > again > often 3 24 Note that if adverbs were X°s, they should block X° movement (e.g. of the infinitive) under Relativized Minimality (Rizzi 1990). The articulation of inflection in Jamaican Creole Independently of AdvPs, Cinque (1999) determines the order of clausal functional heads based on evidence from bound and free morphological inflection. Once again, he examines a wide variety of languages, and the overt relative orders evinced in these support his claim that, as found to obtain for AdvPs, functional heads are universally organised in a single overall order. 1.2.2. Bound functional morphemes If a universal hierarchy of functional heads indeed exists, then head-initial languages such as English should reflect the mirror image to that observed in „headfinal‟ languages such as German: (14) English: These books have been being read all year → Tense > Aspectperfect > Aspectprogressive (15) German: …da er von der Bank angestellt worden sein muss … that he by the bank employed been have must „… that he must have been employed by the bank‟ → Aspectperfect > Tense > Modal The data above confirms that, as a consequence of the headedness parameter, German virtually duplicates the evidence from English for the order of functional heads. 1.2.3. Free functional morphemes Free functional morphemes behave differently to bound functional morphemes in that they disallow the head immediately below them to adjoin onto them. Consequently, in instances of head-initial languages evincing particles, the latter “allow us to directly observe the order of functional heads. One such case is provided by creole languages” (Cinque 1999, 58). Literature on Creoles has generally claimed that the ordering of functional particles is Tense - Mood/modal – Aspect, wherefore the reference to these as TMA markers. Cinque (1999, 59) underlines that “(t)hough in essence correct, this ordering is somewhat gross, and must be qualified. For one thing, various aspectual particles can co-occur, so that their relative order needs to be determined”. Cinque‟s careful analysis of data from head-initial Guyanese Creole, Haitian Creole, and Sranan leads to a refinement of this claim. For example, Guyanese Creole provides evidence for co-occurring aspectual particles: (16) Shi a aalweez/neva de a sing (Guyanese Creole; Gibson 1986, 852f) She HAB always/never DUR PROG sing „She usually always/never keeps singing‟ 25 Stephanie Durrleman Therefore different positions for Asp head positions must be provided in the structure. Secondly, Cinque (1999, 59) notes that “there are occasional claims in the literature for the order Modal > Tense rather than Tense > Modal”. Guyanese Creole once again gives insight for the postulation of various positions for modals: (17) Jaan shuda bin kyaan get fu gu (Guyanese Creole; Gibson 1986, 585) J. MODepistemic PAST MODr MODr COMP go „J. should not have been able to be allowed to go‟ The data in (17) shows different positions for epistemic versus root modals (ability > permission) with respect to tense: indeed whereas the former precedes T°, the latter follows it. Therefore Cinque (1999) refines the traditional analysis of TMA markers so as to derive a more articulate structure with different positions for accommodating different modal types. The more intricate structure of functional particles established on the basis of head-initial languages is further supported by evidence from „head-final‟ languages which also make use of functional particles (e.g. Kachin of the Tibeto-Burman area, or Sanio-Hiowe of New Guinea): The latter in fact display their particles sentence finally, in a mirror image order to that which is established on the basis of direct evidence from „head-initial‟ languages with particles. Putting the attested relative orders together, Cinque (1999) arrives at the single overall order in (18): (18) Moodspeech act > Moodevaluative > Moodevidential > Modepistemic > T(Past) > T(Future) > Moodirrealis Modnecessity > Modpossibility > Modvolition > Modobligation > Modability/permission > Asphabitual > T(Anterior) > Aspperfect/imperfect > Aspretrospective > Aspdurative > Asp generic/progressive > Aspprospective > AspCompletive > Voice > Aspcelerative > Aspcompletive > Asp(semel)repetitive > Asp iterative At this stage one can observe the two independently established hierarchies, namely that of AdvPs and that of functional heads, and see that they generally match semantically from left to right: (19) [Frankly Moodspeech act [fortunately Moodevaluative [allegedly Moodevidentia l [probably Modepistemic [once T(Past) [then T(Future) [perhaps Moodirrealis [necessarily Modnecessity [possibly Modpossibility [willingly Modvolition [inevitably Modobligation [cleverly Modability/permission [usually Asphabitual [again Asprepetitive(I) [often Aspfrequentative(I) [quickly Aspcelerative(I) [already T(Anterior) [no longer Aspterminative [still Aspcontinuative [always Aspperfect(?) [just Aspretrospective [soon Aspproximative [briefly Aspdurative [characteristically (?) [? Asp generic/progressive [almost Aspprospec26 The articulation of inflection in Jamaican Creole [completely Aspcompletive(I) [tutto AspPlCompletive [well Voice [fast/early Aspcelerative(II) [completely AspSgCompletive(II) [again Asprepetitive(II) [often Aspfrequentative(II) … tive Cinque (1999, 77) writes that “(i)n many cases a transparent specifier/head relation between a certain adverb class and the right-adjacent functional head is immediately recognizable”. He concludes that the essential compatibility of individual languages attested with this comprehensive order cannot be accidental. Rather, it follows as a consequence of the fact that the structure in (19) above is a fixed order determined by UG. Therefore Cinque (1999) predicts that no human language should prove incompatible with the hierarchy expressed in (19) or refinements of it. JC is a new territory on which to test Cinque‟s (1999) clause structure. 2. The articulation of Inflection in JC Inflectional markers are the overt manifestation of clausal functional heads, and their corresponding adverbs are the overt realisation of the specifiers of their projections. The investigation undertaken in this paper verifies if a rigid order exists amongst the functional material of the clause in JC, as predicted by the framework, and if such a hierarchy is compatible with that established in Cinque (1999). 2.1. The preverbal markers JC exhibits SVO surface order: 4 (20) Dem en/did nuo dat S V O Them [+past] know that „They knew that‟ Verbs are not conjugated via inflectional morphology in this language. In (20), an independent inflectional element, en (rural) or did (urban) depending on the variety of Creole, precedes the verb stem and gives rise to a past interpretation. These markers do not agree in tense or number with the subject: (21) Mi/yu/Im/Wi/Uno/Dem en/did nuo betta I/you/S/he/We/You[+plural]/They [+past] know better „I/you/S/he/We/You[+plural]/They knew better‟ 4 En and did express past tense in rural and urban Creole respectively. Some of the data drawn from the literature on JC use en, so I familiarise the reader with this marker here. Although this marker is not unfamiliar to me, I will not always give it in my examples as I am not personally a user of it. 27 Stephanie Durrleman JC makes use of markers to express not only tense, but also mood and aspect (TMA). All of these markers, when used, must intervene between the subject and the invariant verb form, as seen for past tense markers in (20) and (21). An example of a modal is given in (22), and an aspectual marker in (23): (22) Im shuda nyam di bammy lang taim S/he [+modal] eat the bammy long time „S/he should‟ve eaten the bammy a long time ago‟ (23) Im a nyam di bammy S/he [+progressive] eat the bammy „S/he is eating the bammy‟ Long sequences of these middle-field inflectional markers seldom occur in JC, but nevertheless they can potentially be used combinatorily, and occasionally are, as long as they fall in a fixed order: (24) a. Jan shuda en a ron S/he [+modal] [+past] [+prog] run „S/he should have been running‟ b. *Jan shuda a en ron c. *Jan en shuda a ron d. *Jan en a shuda ron e. *Jan a shuda en ron f. *Jan a en shuda ron Bailey (1966) 2.2. Modals 2.2.1. Mod(al) 1 The examples in (24a-f) suggest that in JC the structural hierarchy of inflectional markers gives rise to certain modals, here shuda, dominating tense and aspect markers. An inventory of these initial modals is given in (25): (25) shuda, wuda, maita, mosa, kuda A particularity of modals belonging to this class is that they tend to all end in a (26) Im neva shuda tief di mango-dem S/he never [+modal] thief the mango-[plur] „S/he never should‟ve stolen the mangoes‟ Modal-associated a is not a marker encoding past time (such as perfective have in the Standard), contrary to what the gloss for (26) may imply. Notice that stative verbs may follow these sequences and give rise to a present interpretation (27): 28 The articulation of inflection in Jamaican Creole (27) Betta Jan no tell im dat, caw im wuda tink seh im a di bess! Better John [neg] tell him that, „cause him [+modal] think that him [equative] the best „(It‟s) better that John doesn‟t tell him that, because he would think that he‟s the best!‟ The past interpretation is therefore not forced by the presence of this modal-a element at all. Modals from different sets can be combined in JC. The first set of modals in the hierarchy of modal markers is Mod(al)1. Modals of the same set cannot be combined amongst themselves: (28) a. b. c. d. e. f. g. h. *Im shuda wuda… *Im wuda shuda… *Im shuda maita… *Im maita shuda… *Im shuda mosa… *Im mosa shuda… *Im shuda kuda… *Im kuda shuda… etc. 2.2.2. Mod(al)s 2 & 3 In the event that the combination of modals takes place (29a, 30a), this combination must respect a certain ordering constraint, otherwise the result is ungrammatical (29b-d, 30b-d): (29) a. Im shooda muss kyan get tru „He must surely be able to succeed‟ Adams (1995) (30) a. dat-de biebi wuda mos hafi priti „That baby would have to be pretty‟ Bailey (1966) (29) b. c. d. e. *Im muss shooda kyan get tru *Im kyan muss shooda get tru *Im muss kyan shooda get tru *Im shuda kyan muss get tru (30) b. c. d. e. *Dat-de biebi mos wuda hafi priti *Dat-de biebi hafi mos wuda priti *Dat-de biebi mos hafi wuda priti *Dat-de biebi wuda hafi mos priti 29 Stephanie Durrleman As a result the first set of modals, Mod1, whose members always end in a, can be opposed to mos, hafi and kyan which occur deeper in the structure. Mos can in turn be opposed to hafi and kyan in that the former precedes the latter. On the basis of these observations, the positions of modals in the structure of JC 5 seem to fall in the fixed order given in (31) : (31) [Mod1 kuda/wuda/shuda/mosa/maita] > [Mod2 mos] > [Mod3 hafi, kyan] … As already attested for Mod1, modals belonging to the same set compete for the same position and consequently cannot be combined amongst each other: (32) a. *Im kyan hafi… b. *Im hafi kyan… 2.3. Tense 2.3.1. Past Tense In the sequence of inflectional markers in JC, just after the modals ending in a comes the past tense marker en in basilectal varieties, did in mesolectal ones: (33) Im wooda en say (Adams 1995) S/he would have [+past] say „S/he would have said‟ Im wooda did say S/he wouldhave [+past] say „S/he would have said‟ This tense marker is optional in sentences such as those given in (33): even in the event that en and did were done away with, the interpretation yielded could still correspond to the conditional past: (34) Im wooda say „S/he would have said‟ This is because null tense specification is the default mechanism for expressing past with non-stative verbs in JC: (35) Im say dat saim ting deh S/he say that same thing there „S/he said that very thing‟ 5 30 It is not immediately obvious why sequences such as …*kuda kyan…, and …*mosa mos…are banned, as the mutual exclusion of these elements can neither be accounted for in terms of competition for the same position, nor in terms of semantics alone. Notice that in French, a similar constraint on modals exists: *pouvait pouvoir, *devrait devoir. The articulation of inflection in Jamaican Creole Since [-stative] verbs have a default interpretation corresponding to a past-time reading, then in the case of these verbs the insertion of a [+past] tense marker did generally yields an anterior past interpretation. (36) Jan did nyam di bammi John [+past] eat the bammy „John had eaten the bammy‟ The potential temporal specifications for a non-stative verb such as ron are given in (37): (37) a. Mi ron b. Mi en ron I run (habitually); I ran I have run; I had run 6 Sebba (1993) The [±stative] nature of the verb plays a crucial role in determining tense in the absence of markers: we have seen that a non-stative verb without markers has a default interpretation corresponding to past-time. With respect to stative verbs, however, the absence of particles usually results in a non-past interpretation: (38) Jan tink seh im a di bess John think that he [equative] the best „John thinks/*thought that he is/*was the best‟ Therefore the [±stative] nature of a verb plays an important role in the temporal specifications assigned to this verb. On the one hand, the default time specification for non-stative verbs is past, so that when en/did combine with [-stative] verbs the result is an anterior past interpretation. On the other hand, the default time specification for stative verbs is present, so the insertion of the pre-verbal markers en/did are necessary in order to obtain a past interpretation for [+stative] verbs. 2.3.2. Future Tense The markers en/did considered above are tense indicators of past, in contrast to wi which is an indicator of futurity: (39) Jan wi nyam dat „John will eat that‟ Futurity may also be expressed by the aspectual a+go, which often becomes ao in rapid speech: (40) Im a (g)o nyam dat S/he [prog] [prosp] eat that S/he is going to eat that 6 Sebba (1993) draws on Bailey (1966). 31 Stephanie Durrleman Wi and a+(g)o behave differently with respect to en/did. The former, though not the latter, is in complementary distribution with the past tense markers. Indeed, the sense of imminence given by the Asp markers a+(g)o may very well serve in the description of some past event: (41) a. *Im en/did wi nyam dat b. *Im wi en/did nyam dat (42) Im en/did a (g)o nyam dat aaff, bot mi stap im “S/he was going to eat it all up, but I stopped her/him” Given the complementary distribution between wi and en/did, wi is analysed here as a T marker expressing the opposite value (future) to en/did (past). The mutual exclusion of these elements could then be accounted for in terms of their targeting the same T head. This analysis does not extend to a(g)o in light of the compatibility of a(g)o with en/did. A(g)o is therefore considered an aspectual marker of futurity. 2.4. Modals and T 2.4.1. Mod1 and T We have already observed in example (33) repeated here as (43) that mood (Mod1) must precede tense. (43) Im wooda en say (Adams 1995) S/he would+have [+past] say „S/he would have said‟ Im wooda did say S/he would+have [+past] say „S/he would have said‟ We can see now that the various tense markers must in turn precede aspectual ones: (44) a. b. c. d. e. Im did a (g)o nyam… *Im a did/en (g)o nyam… *Im (g)o did/en a nyam… *Im a (g)o did/en nyam… *Im (g)o a did/en nyam… Examples (44a-e) confirm that tense must precede aspect. (45) gives the order established so far: (45) Mod1 > T > Asp 2.4.2. Mod 2 & 3 and T Mods 2 and 3, unlike Mod1, cannot precede a tense particle. Instead, Mods 2 and 3 must follow did/wi: 32 The articulation of inflection in Jamaican Creole (46) *Im did shooda laugh (47) Im did hafi laugh „She had to laugh‟ (48) (Mi feel seh wen di taim come ) im wi kyan dwiit „(I feel that when the time comes) s/he will be able to do it‟ 7 (49) Im wi mos hafi tek dat „S/he will be obliged to take that‟ This means that Mod 2 and 3 are to be situated lower down in the structure than T. In the event that they behave like other lexical verbs, one could situate them under V. However this is unlikely: note that unlike lexical verbs, they cannot be situated below aspect because of the ungrammaticality of the following sequences: (50) *(Wen di taim come,) im a (o) hafi laugh vs. Im wi hafi laugh (51) *(Wen di taim come,)maybe im a (o) kyan dwiit vs. Maybe im wi kyan dwiit The ordering of TMA markers thus seems to be as follows: (52) [Mod1 wuda/shuda/mosa/maita] > [T en/did/wi] > [Mod2 mos] > [Mod3 hafi, kyan] > [Asp1 a [Asp2 (g)o] [V…] Therefore tense intervenes between epistemic and root modals. Recall that Guyanese Creole exhibits a similar phenomenon: (53) Jaan shuda bin kyaan get fu gu J. MODepistemic PAST MODr MODr go „J. should not have been able to be allowed to go‟. (Gibson 1986, 585) Cinque (1999) highlights that the distributional variation between different modal types in relation to T is linked to their correspondingly different interpretational values: Epistemic (pre-T° modals) are analysed as being “concerned with the speaker‟s deductions or opinions” versus root modals (post-T° markers) which, “in contrast to epistemic (…) are strictly subject oriented”. 8 In JC, an epistemic and a root version of the same modal exist : mosa and mos. The semantic contrasts between these two offers new ground where one can test this epistemic/root distinction: 7 8 See also Bailey (1966, 44) for more examples. Kuda has alethic particularities which render it difficult to strike a clear contrast with kyan. C.f. Cinque (1999, 78,79 & 198n3). 33 Stephanie Durrleman (54) Jan mosa did hafi tell dem „John most probably/more than likely had to tell them‟ / „*John was obliged to tell them‟ (55) Jan did mos hafi tell dem 9 „John was obliged to tell them‟ / „ % John probably had to tell them‟ Whereas the preferred interpretation for mos yields the notion of necessary obligation with respect to the subject Jan, this is not accessible to mosa, which does not emphasise Jan‟s obligation, but rather invokes the speaker‟s opinion with respect to Jan‟s plausible obligation. Another clear difference between the two has to do with their use in questions: mosa though not mos, gives rise to a distinct awkwardness in a question: (56) ??Jan mosa did hafi tell dem? (57) Jan did mos hafi tell dem? Jackendoff (1972, 103) writes that “(i)f epistemic modals are treated like speaker-oriented adverbs by the semantic component, this restriction will follow automatically”. The above observations give evidence for the structure below: (58) Mod epistemic (Mod 1) > T > Mod root obligation (Mod 2) > Mod root ability/permission (Mod 3) 2.5. Asp markers This section turns to aspectual markers in JC. Markers of aspect in this language, and seemingly in all languages, form the group of inflectional particles located closest to the VP (see Bybee (1983)). 2.5.1. Progressive Aspect This marker a precedes the [-stative] verb so as to give the action or event evoked by this verb an „on-going‟ interpretation: (59) Jan a nyam i’ John [+prog] eat it „John is eating it‟ 9 34 It would be inaccurate to conclude that mos cannot give rise to an interpretation where the speaker‟s opinion is involved. This ambiguous nature is also attested with dwe for some speakers of Haitian Creole (see Leblanc (1989, 51). The articulation of inflection in Jamaican Creole We observed that [-stative] VPs such as [nyam i’] are generally interpreted in the past in the absence of markers. In (60), the sole use of the particle a with this same VP implies that the event described is going on at the time of the utterance. In the presence of a, even with projections of [-stative] verbs, a past interpretation has to be specified: (60) Jan did a nyam uno bammi John [+past] [+prog] eat your[plur] bammy „John was eating your bammy (… good ting mi tell im fi stap) (… good thing I tell him to stop) (… good thing I told him to stop)‟ Therefore the use of the progressive suggests that the verb evokes something in progress, by default at the time of utterance. Recall that [+stative] verbs are, by definition, inherently on-going at the time of utterance: (61) Jan nuo dat John know that „John knows that‟ It is not surprising then that the redundant combination of stative verbs with the progressive is banned: (62) *Jan a nuo dat A combines with (g)o to yield a reading of futurity referred to as prospective aspect. 2.5.2. Prospective Aspect Cinque (1999, 99) explains that: “the term „prospective aspect‟ has come to be used for those grammatical forms (…) which mark “a point just prior to the beginning of an event” (Frawley 1992, 322). This is for example the case with the English construction “to be going to” / “to be about to”. (Comrie (1976, 64ff))”. Comrie (1976, 64) points out an important difference between “prospective meaning”: Bill is going to / is about to throw himself off the cliff and expressions of future time reference:Bill will throw himself off the cliff, since Bill‟s eventually not throwing himself off the cliff makes the speaker wrong in the second case, but not in the first. As already underlined, a reinforcement of the idea that wi warrants an analysis as the tense marker for future while a(g)o is aspectual is founded on the observation that wi is in complementary distribution with did, although this is not the case with a(g)o: (63) a. *Im en/did wi nyam i’ aaff b. *Im wi en/did nyam i’ aaff (64) Im en/did a (g)o nyam i’ aaff, bot mi (en/did) stap im „S/he was going to eat it all up, but I stopped her/him‟ 35 Stephanie Durrleman The fact that prospective go must be used in combination with the progressive (even though in rapid speech the latter may become somewhat shortened) can be seen by the fact that a minimal pair can be formed between preverbal go used without the progressive and one used with the progressive: (65) (Afta wa im seh…) yu go pick i’ up? Vs. y(u) a go pick i’ up? (After what s/he say) you go pick it up you [+prog] [+prosp] pick it up „(After what s/he said)you went and picked it up? /you‟re going to pick it up? 2.5.3. Retrospective Aspect In this work, the marker jus designating „immediate past‟ is referred to along the lines of Cinque (1999) as retrospective aspect (66) Im did jus a go dw i’ S/he [past] [retrospective] [progressive] [prospective] do it „S/he was just about to do it‟ 2.5.4. Completive Aspect and Anterior The inflectional particle done may precede the VP as do other aspect markers of JC, but it also has the particularity of occasionally occurring in a post-VP configuration. This was already noticed by Cassidy (1961) who gives the following description for done: “The participle done enters into a peculiar adverbial idiom. Placed after verbs it shows completion of the action. (…) „Me feed him dun dis long time‟ (…) done loses verbal force and becomes a modifier of the other verb.” In the event that done occurs in a pre-VP configuration with [-stative] verbs, it can yield two different interpretations. A sentence such as that in (67) is therefore ambiguous, giving rise to the interpretations in (67a) and (67b): (67) Im done nyam i’ S/he done eat it a. „S/he already ate it‟ b. „S/he finished eating it‟ The additional interpretation in (67a) is erased in a post-VP configuration, so the ambiguity observed for (67) no longer obtains in (68) where done follows the VP it modifies: (68) Im nyam i’ done S/he eat it done a. „*S/he already ate it‟ b. „She finished eating it (up)‟ 36 The articulation of inflection in Jamaican Creole I take this to suggest the presence of two different done markers in JC, one corresponding to the meaning [+completion] as given by the verb „to finish‟ in English, and the other corresponding to the meaning [+anterior], as given by the adverb „already‟ in English. These two done markers behave differently syntactically: Assuming that VP10 movement takes place to the Spec of the completive marker done when the latter appears in a post-VP configuration, according to the data in (68) movement of the VP projected by a [-stative] verb to the Spec of Anterior done is not accessible to VP-movement. In other words, if one were to translate the sentence in (69) into JC, one could not say (69a), only (69b): (69) S/he already ate it a. *im nyam i’ done b. im done nyam i’ A possible hypothesis to account for the fact that a VP cannot move to the Spec of [+anterior] done, is that the VP in JC cannot move as high as the [+anterior] projection. Movement of the VP in JC is limited as examples (70-2) illustrate: a VP cannot be found in the specifier of projections for retrospective aspect jus (70a, b), progressive aspect a (71a, b), or prospective aspect go (72a, b): (70) a. Im jus nyam i’ S/he just eat it „S/he just ate it‟ b. *Im nyam i’ jus (71) a. Im a nyam i’ S/he [+prog] eat it „S/he is eating it‟ b. *Im nyam i’ a (72) a. Im a (g)o nyam i’ S/he [+prog] [+prosp] eat it „S/he is going to eat it‟ b. *Im a nyam i’(g)o If movement of [-stative] VPs in JC cannot go as high as the specifiers of aspectual particles such as jus, a and (g)o, yet the specifier of the particle done 10 I assume leftward movement of a VP to be possible, in line with Cinque (1999, 190n26) who also makes use of this device to derive sentence final don for Guyanese Creole. I return to the discussion of VP movement in section 3. 37 Stephanie Durrleman [+completive] may host the [-stative] VP in JC, then it can be reasonably hypothesised that the projection of done [+completive] be situated quite low in the structure: lower than the aspectual markers considered here. If this were not the case, then movement to the specifer of completive aspect would be excluded since it would violate Relativized Minimality through having to skip intermediate specifier positions. This predicts a structure along the lines of (73): (73) [Intermediate Asp] > [done [+completive]] > VP X Indeed, this is confirmed by the examples in (74, 75) which overtly illustrate the distribution of this particle [+completive] as used in JC in relation to other inflectional markers: (74) Wentaim mi reach, im did jus done nyam i’ When I reach, s/he [+past] [+retrospective] [+completive] eat it „When I arrived, s/he‟d just finished eating it ‟ 11 (75) Mine! Im a go done nyam di whole a i’! Mind! S/he [+prog] [+prosp] [+completive] eat the whole of it „Careful! S/he is going to finish eating all of it!‟ Where done follows the lowest of the other overtly expressed aspectual heads in JC: (76) Asp [+retrospective] > Asp [+prog] > Asp [+prosp] > Asp [+completive] As expected, this occurrence of done, i.e. deeply embedded structurally, cannot correspond to an interpretation meaning [+anterior], but is reserved only for an interpretation implying that the action expressed is completed. It follows that [+completive] aspect is as hypothesised, situated low down in the clausal hierarchy: lower than the other aspectual heads already considered, and lower than T [+anterior]. Given the observations above, JC gives evidence for two different types of inflectional particles done: one [+anterior], the other [+completive]. Completive may occur either in a pre- or post-VP configuration, and is only compatible with [stative] verbs, Anterior can only occur in a pre-VP configuration, and may occur with both [+stative] and [-stative] verbs. Done when used in combination with [+stative] verbs in a pre-VP configuration, unlike that observed with respect to [-stative] verbs, does not give rise to ambiguity 11 38 Note that while done [completive] may potentially occur with prospective a (g)o, it does not occur easily with progressive a alone : ? ? ?im a done nyam i’. The articulation of inflection in Jamaican Creole in JC. Consider example (77) with the verb nuo, and the fact that the interpretation in (77a) can be derived, whereas the one in (77b) cannot: (77) Im done nuo seh mi like im a. „S/he already knows that I like her/him‟ b. „*S/he finished knowing that I like her/him‟ This amounts to saying that the only marker done which can occur with [+stative] verbs is the one which gives rise to the meaning [+anterior], as expressed 12 by the adverb „already‟ in English. Lamiroy (1987, 284) accounts for this by suggesting that: “since all the phases in a state are identical (…) (states) lack an internal dynamic structure. Aspect, however, crucially deals with the internal structure of situations. Therefore states and the expression of aspect are naturally incompatible”. It follows then that stative verbs may be specified for T [anterior] done, but not for Asp [completive] done. The hypothesis that there exist two different done markers in the clause structure makes a prediction as to the potential syntactic distribution of the projections of [+stative] verbs with respect to the particle done: If done [+anterior] cannot host VPs in its Spec for it is too far away, and only this done can occur with [+stative] verbs, then [+stative] VPs should never be able to occur in a pre-done configuration, as the latter configuration is derived by movement to the specifier of done [+completive] only. This prediction is borne out as the contrast between (78a) and (78b) illustrates: (78) a. Im done nuo dat S/he done know that „S/he already knows that‟ b. *Im [[nuo dat]i done] ti S/he know that done Done [+anterior] does not easily combine with other markers in JC, so although we have reason to believe it is relatively high in the hierarchy of markers, it would be difficult, maybe impossible, to locate its exact position in the structure if we did not have recourse to its corresponding adverb aredi, an overt manifestation of its specifier position along the lines of Cinque(1999, 94): Aredi, like done[anterior], has as “its core meaning (…) one of temporal priority (…), in fact, one of precedence with respect to a reference time. (…) This makes it plausible to locate it in the specifier position of the lowest TP (TP anterior)”. 12 In Da Cruz (1995, 368). 39 Stephanie Durrleman 13 Although aredi is, more often than not, placed at the end of a sentence (79a-c) , it can occasionally be inserted amongst certain of the other markers, as examples (79e-g) illustrate: (79) a. Im mosa gi ’im di gassip aredi „S/he must have given her/him the gossip already‟ b. Im did nuo dat aredi „S/he knew that already‟ c. Im a gwaan bad aredi „S/he is behaving badly already‟ d. Im (*aredi) mosa (aredi) gi ‘im di gassip S/he (*already) must [epistemic] (already) give her/him the gossip „S/he must have already given her/him the gossip‟ e. Im (*aredi) did (aredi) nuo dat S/he (*already) did (already) know that „S/he already knew that‟ f. Im (aredi) a (*aredi) gwaan bad S/he (already) [prog] (*already) go+on bad „S/he‟s already behaving badly‟ The data in (79d-f) gives evidence for the structure in (80): (80) Mod (epistemic) T (past/future) > aredi (anterior) > Asp prog Given the respective distributions of done [+completive] (76 repeated as 81) and aredi [+anterior] with respect to Asp [+prog] (80), the projection corresponding to [+anterior] is by transitivity higher in the structure than the one corresponding to [+completive] (82): (81) Asp [+retrospective] > Asp [+prog] > Asp [+prosp] > Asp [+completive] (82) T [anterior] > Asp [+prog] > Asp [+completive] 2.5.5. Continuative and Frequentative Aspects Cinque (1999) situates the adverb „still‟ in the specifier position of the continuative aspect projection. In JC, the position of this adverb indeed coincides with the Asp [continuative] projection in his structure: i.e. JC still can be shown to follow the root modal [ability/permission] kyan, and therefore all markers dominating this mo13 40 Sentence final aredi may possibly be derived through XP movement past this adverb, along the lines of (Cinque (1999). I return to this hypothesis in Section 4.3. The articulation of inflection in Jamaican Creole dal, and it can also be shown to precede those markers dominated by continuative 14 aspect in Cinque‟s structure (83) Jan wuda/did/mos/kyan > still > a/go chat John [Mod1] / T / [Mod2] [Mod3] > still > Asp [prog] Asp [prosp] talk Cinque (1999, 207n51) observes that “(c)ontinuative aspect is found expressed by particles, (…) or (apparently, more often) by reduplication of the verb stem” In JC, reduplication of the verb stem is also a productive process: (84) Yu nuh nuo im? A di same wan weh chat-chat wi business? You [neg] know her/him? [equative] the same one which chat [V reduplicated] we business „Don‟t you know her/him? S/he‟s the very one who‟s incessantly spreading our personal affairs‟ Whether or not reduplication of the verb in JC directly corresponds to continuative aspect, however, is not so clear-cut: Bailey (1966, 16) takes repetition of a verb to “refer to repetitive or habitual action”, while giving the examples taak-taak, and biit-biit which she translates as “talk continuously” and “whip constantly” respec15 tively . It seems therefore that what is stressed by the reduplication of a verb is not specifically the continuity of the action expressed, but rather more generally the frequency at which this action takes place. If reduplication were an expression of frequency in JC, then the fact that stative verbs do not reduplicate would follow: these verbs cannot be qualified frequentatively either. Recall that the adverb „still‟ and its analogues in different languages is taken by the framework here adopted, to be the overt realisation of the specifier position of the projection of continuative aspect. Notice in the example below that still can be separated from the reduplicated verb by the progressive particle: (85) im still a chat-chat di people-dem business bout di place S/he still [prog] Verb-Verb the people [+plur] business around the place „She‟s still incessantly spreading those people‟s private affairs all over town‟ 14 15 Still is not compatible neither with retrospective aspect jus, nor with anterior or completive aspects done. Indeed if jus occurs with still, jus cannot mean „a short while ago‟, but rather is automatically interpreted to mean „nevertheless‟: im still jus do wa im waan fi do: „s/he kept on nevertheless doing what s/he wanted to do‟ vs. „*s/he kept on a short while ago doing what s/he wanted to do‟. Anterior done, as well as its specifier aredi, are in complementary distribution with still: *im done still nyam; *im aredi still nyam. Similarly, completive done shows incompatibility with still: *im still done nyam. Bold letters are mine. 41 Stephanie Durrleman If still sits in the Specifier of AspContinuativeP, and a is the overt realisation of Prog°, then the preverbal copy cannot sit in AspContinuative°. A possible analysis to account for the verbal reduplication in JC is one which situates the preverbal copy in AspFrequentative°, i.e. the lowest functional head of the clausal system. This would explain why it can even follow done, the marker of completive aspect: (86) Im nuh done chat chat di people-dem business all now? S/he [neg] [completive] [frequentative] chat-chat the people [plur] business even now? „S/he hasn‟t yet finished incessantly spreading those persons‟ private affairs?‟ The interpretation and distribution of aspectual heads in JC considered here and, where possible, their corresponding specifiers, is compatible with the hierarchy in (87): 16 (87) Anterior < Asp Continuative < Asp retrospective < Asp progressive < Asp prospective < Asp completive < Asp frequentative 2.6. Overall order for TMA markers in JC (88) Mod epistemic (Mod 1) > T > Mod root obligation (Mod 2) > Mod root ability/permission (Mod 3) > Anterior < Asp Continuative < Asp retrospective < Asp progressive < Asp prospective < Asp completive < Asp frequentative 3. Theoretical issues This section returns to the case of the marker done in JC (section 2.5.4). The aim is to examine the validity of an IP-internal-movement analysis for sentence final material, as compared to a Serial Verb Construction (SVC) approach for Completive and distinct sentence-medial and sentence-final for AnteriorP. 3.1.1. Completive Aspect in JC and Fongbè: an SVC approach It has been pointed out that the marker encoding Asp [completive] behaves differently syntactically with respect to the other markers in JC: done [completive] has the 17 distributional particularity of optionally following the VP over which it takes scope : 16 42 Note, however, that the incompatibility of Anterior with AspContinuative and AspRetrospective makes it difficult to determine the precise hierarchy between these proje ctions in JC. The articulation of inflection in Jamaican Creole 18 (89) Uno jus nyam di bammi done ? Subj [retrospective] VP [completive] „You all have just finished eating the bammy?‟ Analogous patterns related to the completive marker are found crosslinguistically. Da Cruz (1995) observes that: Fongbè places completive markers fó and vò in a VP-final configuration: (90) Kòkú wà àzo ó fó Kòkú do work DET finish „Kòkú finished doing the work‟ (91) Kòkú kló katake le vò Kòkú wash high stool PL finish „Kòkú finished washing the high stools‟ Da Cruz (1995, 364) defines fó/vò as aspectual verbs, and distinguishes them from aspectual morphemes of Fòn as the word order of sentences containing the former is different to that of sentences containing the latter: “The word order (…) distinguishes the verbs fó and vò from the aspectual morphemes of Fòn” (da Cruz 1995, 364). Whereas irrealis ná, and habitual nó precede the VP over which they take scope, fó and vò follow it. Consequently, da Cruz (1995) quotes Avolonto‟s (1992) structure for aspect markers ná and nó as given in (92), and considers this an impossible underlying structure for fó and vò (93): (92) AspP Asp‟ Asp° ná/nó VP V‟ V° 17 18 NP It is worth noting that this option is less exploited by the younger generations than preverbal done. Whether or not there are restrictions on the type of VP which can precede done is a topic for future research. 43 Stephanie Durrleman (93) * …Asp‟ Asp° VP fó To account for the particularities of fó/vò sentences, da Cruz argues that they are serial verb constructions involving obligatory control. Therefore under his approach, a sentence such as (91), repeated as (94a), is assigned the structure (94b) below: (94) a. Kòkú kló katake le vò Kòkú wash high stool PL finish „Kòkú finished washing the high stools‟ (94) b. IP I‟ NP Kòkúi I VP1 V‟ V‟j V1 VP2 NP OPj kló katakè lé VP2 V‟ PROi V2 V‟ vò ej Da Cruz (1995, 374) explains his approach as follows: “In order to express the fact that fó and vò semantically select an event, I propose that (…) the complement of fó,vò is an empty verbal projection (minimally a V‟). I argue that the empty category which occupies this position is a variable bound by a null operator (Op) which is adjoined to the VP headed by fó,vò. The first VP and Op have the same reference. Op and the variable form a chain ; thus, by co-indexation, there could be transfer to the chain of the semantic properties of VP1.” He argues (p.377) that “The analysis of sentences with aspectual verbs fó and vò as obligatory control constructions has many advantages. The structure in (98b) accounts for the S-structure word order in Fongbè, and offers a mechanism for understanding the selectional restrictions on the verbs which combine with fó and vò. As a matter of fact, indicating that the subject 44 The articulation of inflection in Jamaican Creole of the construction finished realizing the event expressed by the preceding VP, fó and vò can only combine with verbs which are compatible with this property”. Da Cruz (1995) bases his analysis of sentences containing fó and vò as serial verb constructions on the observation that they have four characteristics of serial verb constructions as listed under (95) through (98) below. In both fó/vò sentences and serial verb constructions: (95) (96) (97) (98) “there is only one lexical subject and more than one verb” “there is only one expression of time and aspect” one can see that they “only contain one expression of negation” and, “just as there are semantic restrictions on serial verb constructions (…) there are semantic restrictions in the context of the aspectual use of fó and vò” On the other hand, da Cruz (1995, 370) recognises that “there is not object sharing with aspectual verbs fó and vò” although it has been argued in Baker (1989) that “the Projection Principle predicts that object sharing is not only possible in serial verb constructions, but obligatory”. Da Cruz then takes fó/vò constructions to be evidence for 19 the conclusion that argument sharing is not a necessity in serial verb constructions . I do not adopt the analysis of da Cruz (1995) in this paper, and instead tentatively propose an analysis along the lines of Cinque (1999) which involves move20 ment . The VP-movement analysis is seen to account for the ungrammaticality of fó/vò constructions as discussed in da Cruz (1995) independently of a consideration of such sentences as serial verb constructions involving obligatory control. The fact that fó/vò sentences do not involve argument sharing follows from this approach: The lack of argument sharing would then not in itself constitute an argument for abandoning Baker‟s (1989) hypothesis that serial verb constructions obligatorily involve argument sharing, but rather could be a consequence of the fact that fó/vò sentences are not serial verb constructions. 19 20 The fact that I do not treat fó/vò (Fongbè) and done (JC) constructions as serial verb constructions is inspired by the approach in Cinque (1999). The IP-internal VP-movement hypothesis applied here has implications for a serial verb approach. These implications deserve careful examination which is beyond the scope of this work. In this paper, I follow Kayne (1994), Cinque (1999) in assuming a head-initial X‟-schema. Notice that an alternative approach which would allow for head-final structure would imply that the constructions here considered could be derived independently of movement. 45 Stephanie Durrleman 3.1.2. Completive Aspect in JC and Fongbè: VP-movement The observation in (95) that fó/vò, like serial verb constructions, involve “only one lexical subject and more than one verb” is used by da Cruz (1995) to account for the ungrammaticality of the (99): (99) *Kòkú wà àzo ó Kòkú fó Kòkú do work DET Kòkú finish „Kòkú finished doing the work‟ However, the fact that there is “only one lexical subject and more than one verb” is a characteristic of any sentence containing a lexical verb augmented with an aspectual marker, once this marker is simply labelled as an aspectual „verb‟, which I take to be the case of fó in (99). Consider, for example, the contrast between the data in (100a) and (100b, c): (100) a. Jan a go nyam di bammi J. [aspectul verb 1 (progressive)] [aspectual verb 2 (prospective)] VP b. *Jan a go Jiemz nyam di bammi c. *Jan a Jiemz go nyam di bammi The contrast in (100) stems from the fact that aspect markers are functional material and consequently do not project an argument structure. To insert an additional external argument into a sentence with only one lexical verb would mean that this argument would be without a theta role and therefore entail a violation of the Theta Criterion. The second parallelism drawn by da Cruz (1995) between serial verb constructions and sentences containing fó/vò, is that “there is only one expression of time and aspect” in both. He illustrates the relevance of this hypothesis by means of the contrast between (101a) and (101b): (101) a. Asíbá ná sá sèn dó hòn ó fó Asíbá IRR pass on paint put door DET finish Asíbá will finish painting the door‟ b. *Asíbá ná sá sèn dó hòn ó ná fó Asíbá IRR pass on paint put door DET IRR finish However, if (101b) were not treated as a serial verb construction but rather as a run-of-the-mill sentence with multiple asp markers, it would be ruled out anyway since this sentence makes use the same marker twice: In (101b), the marker encoding irrealis is repeated within one single sentence. Notice that a double use of an aspectual marker in one sentence such as that encoding progressive from JC below also yields ungrammaticality (102b, c): 46 The articulation of inflection in Jamaican Creole (102) a. Im a go nyam I‟ S/he [prog] [prosp] eat it „S/he is going to eat it‟ b. *Im a go a nyam I’ c. *Im a a go nyam I’ The third parallelism between serial verb constructions and fó/vò sentences has to do with the two “constructions only contain(ing) one expression of negation”. The ungrammaticality of (103) is assumed to stem from the fact that it is a serial verb construction: (103) *Asíbá má sá sèn dó hòn ó má fó Asiba NEG pass on paint put door DET NEG finish However, once again, the ungrammaticality of (103) could also be linked to another factor: Multiple negation can give rise to ungrammaticality in almost any sentence involving an asp marker if this marker as well as the VP it modifies are both negated (104b): (104) a. Jan nuh jus pain di door deh? John neg [retrosp] paint the door there „Isn‟t it just a while ago that John painted that door?‟ b. *Jan nuh jus nuh pain di door deh? John NEG just NEG paint that door there Finally, da Cruz (1995, 366) remarks that “just as there are semantic restrictions on serial verb constructions in Fongbè in general (…), there are semantic restrictions in the context of the aspectual use of fó and vò”. He takes the ungrammaticality of (105) to be a consequence of its being a serial verb construction which violates a semantic selectional restriction: (105) *Kòkú mò Báyì fó Kòkú see Báyì finish „Kòkú finished seeing Báyì‟ Notice, however, that not only lexical verbs, but also markers of aspect commonly show semantic restrictions on the VP they select: recall, for example, that the progressive aspect marker cannot combine with a [+stative] VP: (106) *Jan a nuo dat John [prog] know that Therefore the semantic restriction on selection typical of fó/vò sentences does not necessarily lead to the conclusion that fo/vo are verbs in serial verb constructions. 47 Stephanie Durrleman Indeed fó/vò are likely to be aspect markers with selectional restrictions like those already observed with other aspectual markers. In conclusion, I do not believe that the characteristics sketched in da Cruz (1995) imply necessarily that fó/vò sentences involve serial verb constructions with obligatory control. Moreover, I believe this analysis has undesirable theoretical shortcomings: Serial verb constructions are typically constructions involving a sequence of verbs which share logical arguments. To illustrate this with an example from JC, consider the data in (107): 21 (107) (Yu nuo wa dat ginal do?) Im tek mi ackee (go) sell a maakit! (You know what that ginal do?) Him take me ackee (go) sell at market „(Do you know what that trickster did?) S/he actually (went and) sold my ackee at the market!‟ Both tek and sell are transitive verbs. The object mi ackee is „shared‟ by the verbs tek and sell: i.e. what is both „taken‟ and „sold‟ is „my ackee‟. Indeed according to the Projection Principle, this sharing of an object is considered by Baker (1989) to be obligatory in serial verb constructions. Da Cruz (1995) illustrates that fó/vò constructions do not involve object sharing: (108) Ajòtó lé xò kãnlìn lé fó Thief PL hit animal PL finish „The thieves finished hitting the animals‟ Da Cruz (1995) explains that the data from Fongbè in (108) above “mean(s) that the action of hitting (…) is finished. (…) The interpretation of (this) sentence() does not imply that (…) the animals are “finished”.” This reading extends to done in JC as the example (109) taken from Bailey (1966, 42) nicely illustrates: (109) Jiemz no riid di buk don yet Jamed [neg] read the book [completive] yet „James has not finished reading the book yet‟ It is indeed the act of reading the book, and not the book itself, which is qualified by the marker done. The data above illustrates therefore that object sharing is a characteristic of serial verb constructions which is not typical of fó/vò and done sentences. This suggests, according to the hypothesis in Baker (1989), that fó/vò and done sentences are not serial verb constructions. 21 48 Ackee is a fruit used to prepare Jamaica‟s national dish: ackee and saltfish. The articulation of inflection in Jamaican Creole Notice that the analysis proposed in da Cruz (1995) for fó/vò sentences is unable to capture similarities between fó and vò and markers of completive aspect in other languages such as English which do not at first sight exploit (serial verb) constructions with obligatory control. One such similarity is noted by da Cruz (1995, 372) himself: “(f)ó and vò present the same selectional restrictions as the aspectual verb „finish‟. They only combine with accomplishment verbs or with activity verbs which are interpreted as quasi-accomplishments”. The choice of a different syntactic apparatus for fó/vò on the one hand, and „finish‟ or „finir‟ on the other is based on the observation that “contrary to what is observed in English and French (…) the verbs fó and vò appear after the VP complement (da Cruz 1995, 369).” Postulating an entirely new analysis for constructions which share many points in common apart from surface word order is intuitively unattractive. What would be preferable is an approach which would reflect the underlying cross-linguistic parallelism between linguistic elements such as those considered here, while allowing for a derivation which could be responsible for their SurfaceStructure difference. On a language specific level, this analysis also fails to reflect underlying similarities between heads encoding aspect in Fongbè. It is shown in da Cruz (1995) that other aspectual morphemes of Fòn occur in a pre-VP configuration. Example (110) taken from da Cruz (1995) illustrates that markers encoding irrealis or habitual aspect select a VP complement which must surface to their right: (110) Kòkú ná/ nó wà àzo Kòkú IRR HAB do work „Kòkú will work / usually works‟ To postulate that markers encoding irrealis and habitual aspect select a phonetically realised VP complement, whereas what appears to be the marker encoding completive aspect selects a control construction strikes an undesirable inconsistency in the underlying grammar of Fongbè: the obligatory control construction postulated by da Cruz (1995) would be a structural particularity of the complement of the element encoding completive aspect as opposed to those encoding irrealis and habitual. It would be favourable to account for the respective Surface-Structure differences between the markers of this language with an analysis which retains an underlying similarity between them. It is also noteworthy that when fó and vò function as lexical verbs they must select a nominal complement which surfaces to their right (111a, b) just as JC done does (112): 49 Stephanie Durrleman (111) a. Fongbé: Kòkú fó àzo ó Kòkú finish work DET „Kòkú finished the work‟ b. Kòkú vò mólìnkún ò Kòkú finish rice DET „Kòkú finished the (plate of) rice‟ (112) JC: Im done di bammi S/he finish the bammy „S/he finished the bammy‟ If other aspectual verbs in Fongbè generate their verbal complement to their right (110) without recourse to control, and lexical verbs in this language also generate their nominal complements to their right (111), then the underlying structural tendency is for a X°, lexical or functional, to generate its complement to its right. It is therefore plausible that the VP which precedes fó and vò is its complement which has simply been generated to the right and has undergone leftward movement. Da Cruz (1995) himself touches upon this possible analysis: „In Fongbè, the NP complement of the verbs fó/vò is always on the right (…). If the VP (…) is generated in the same position, we would then have to explain the word order at S-structure by a movement of this VP (…). Thus, one could suppose that there is movement to the left of the VP complement.” Indeed this is what I argue to be the case. The VP-movement hypothesis is rejected by da Cruz (1995) for Fongbè for two main reasons: One reason is that “(i)t is impossible to have S-structure sentences like those in (113a, b) in which a VP with a phonological content is on the right of fó and vò. (113) a. *Kòkú fó wà àzo ó Kòkú finish do work DET „Kòkú finished doing the work.‟ b. *Kòkú vò kló katake le Kòkú finish wash high stool DET „Kòkú finished washing the high stools.‟” Another reason he gives for rejecting a VP-movement hypothesis is that “this hypothesis is difficult to defend, given that there is no independent motivation for such a movement of VP in Fongbè” (da Cruz 1995, 373). I believe that these reasons can be countered on the basis of the following observations: Firstly, other languages influenced by such African languages as Gbe, namely creoles, do allow S-structure sentences where the completive marker may surface to 50 The articulation of inflection in Jamaican Creole the left of a phonetically realised VP complement. This can be seen in data from JC (114) and Guyanese Creole (GC) (115, taken from Edwards 1991): (114) JC Jan done nyam i’? Subj [completive] V O „John finished eating it?‟ (115) GC Somtaim wen you don wok yu go an bai a dringk Subj [completive] V „Sometimes when you are finished working you go and buy a drink.‟ The fact that Gbe languages constitute part of the substratum of JC, and that JC allows the completive marker to optionally precede its VP complement reinforces the idea that when the VP surfaces to the left of completive markers fó/vò in Fongbè or done in JC, it has in fact originated in a post-VP configuration and undergone leftward movement: (116) … AspCompletive Asp‟ Spec Asp° fó/vò done VP Secondly, observe that this type of movement is plausibly exploited by Fongbè since another Kwa language of the Gbe group, namely Gungbe, exploits such movement of the VP and extended projections of VP. This is illustrated for purposeclauses, and imperfective/prospective constructions in Aboh (1998). It is therefore not implausible that this analysis be extended to Fongbè. One instance of the application of leftward movement of an extended projection of V in Gungbe is applied by Aboh (1998) to „purpose‟-clauses in Gungbe known as gbé-constructions (117): (117) Hwé-énenu Asíbá nò yì hwéví jrá gbé At that time Asiba Hab go fish sell Purpose „At that time Asiba habitually went out to sell fish‟ The analysis given to account for the structure of the purpose-clause in (117) is (118): 51 Stephanie Durrleman (118) AspP2 AspP2‟ Spec Asp°2 yì NomP Nom‟ spec Nom° gbé AspP3 Asp3‟ spec hwévíj Asp°3 jrái AgroP Agr‟ spec tj Agr° t‟i VP V ti DP tj Under the approach in Aboh (1998), aspect verbs like yì „go‟ select a syntactic unit NomP whose head may be realised by the purpose-marker gbé. The internal argument of this marker is a reduced clause: the aspectual projection AspP3. This entire aspectual projection must move to the specifier position of the projection to the immediate left of the small clause: [Spec,NomP] for nominalization purposes. This analysis is shown in Aboh (1998) to capture not only the syntactic particularities of purpose-clauses, but also those of imperfective/prospective clauses. Imperfective/prospective sentences in Gungbe always end in a low tone: `. This can be explained under the analysis that the imperfective marker tò, is situated under AspP2 in a structure like that given in (118). Tò is then logically in complementary distribution with aspectual verbs such as yì. Imperfective tò selects a NomP which is headed by a Nom° realised as `. When AspP3 moves to [Spec,NomP] for nominali- 52 The articulation of inflection in Jamaican Creole zation purposes in imperfective constructions, a logical consequence of this move22 ment is that a low tone ` always floats at the end of the sentence : (119) … tò [ Spec NomP [AspP3 nà O V]i [Nom ° ` ] t i imperf prosp O V Therefore, if movement of extended projections of VP arguably takes place in languages of the Gbe group, it is not implausible that VP movement occur in Fongbè, and ultimately in JC. Moreover, the scope properties of completive constructions follow from a VP-movement analysis: the marker encoding completive aspect c-commands the trace of its VP complement. No additional structure is needed to account for the scope properties of completive constructions under the analysis adopted here: That the completive marker takes scope over the VP which precedes it is a natural consequence of the movement hypothesis since according to the structure in (116), the marker encoding completive aspect c-commands the trace of its VP complement. Recall that the ungrammaticality of sentences (99), (101b), (103) and (105) was argued by da Cruz (1995) to stem from these sentences being serial verb constructions. Recall also that postulating this analogy is not the only option available. Notice now that the ungrammaticality of (99), (101b), (103) and (105) repeated here as (120), (121), (122) and (123) follow from the structural analysis in (114): Structure (114) leaves no space for an element to intervene between the VP in [Spec,CompletiveAspP] and fó/vò in CompletiveAsp°, which gives a syntactic account for (120) through (123): *Kòkú [Spec AspCompletive wà àzo ó] [ ? Kòkú] [AspCompletive° fó] Kòkú do work DET Kòkú finish „Kòkú finished doing the work‟ *Asíbá ná [Spec AspCompletive sá sèn dó hòn ó] [?ná] [AspCompletive° fó] Asíbá IRR pass on paint put door DET IRR finish *Asíbá má [Spec AspCompletive sá sèn dó hòn ó][? má] [AspCompletive° fó] Asíbá NEG pass on paint put door det NEG finish The ungrammaticality of sentence (105) repeated as (123) follows from feature incompatibility: the VP mò Báyì: “see Báyì” does not bear the feature [+completive] and therefore cannot occupy the specifier position of the Completive Projection: 22 I refer the reader to Aboh (1998) for details. 53 Stephanie Durrleman *Kòkú [SpecCompletiveAsp [mò Báyì [–completive]] [CompletiveAsp° fó] Kòkú see Báyì finish „Kòkú finished seeing Báyì‟ 23 The very motivation for movement of a VP to [Spec, CompletiveAspP] is explainable in terms of the presence of the completive aspect marker in CompletiveAsp° endowing this projection with a [+completive] feature. This feature is strong in Fongbè since the specifier of the CompletiveAsp projection in this language must be morphologically realised at S-structure. The [+completive] feature is less strong in JC, so that the filling of [Spec,CompletiveAsp] may optionally occur at the level of Logical Form. In short, leftward movement of the VP to [Spec,CompletiveAspP] has the theoretical advantage of accounting for the particularities of completive aspect constructions, while retaining an underlying structural consistency between markers of aspect both cross-linguistically, as well as within the internal grammars of languages such as Fongbè and JC. 3.1.3. Problem for the VP-movement analysis The analysis proposed here still faces the problem of accounting for the optionality of VP-movement to [Spec,CompletiveAspP] in JC. However, this would be a problem for the alternative analysis in da Cruz (1995) also: indeed if the surface order were to be the determining factor in the development of a syntactic apparatus for linguistic elements, then one would have to postulate completely different underlying structures for pre- and post-verbal completive done, although the two are semantically equivalent. The fact that younger speakers of JC use VP-final done more rarely than preverbal done, and more rarely than the older generations, may be an important point for the IP-internal VP-movement hypothesis: Possibly, the originally strong [+completive] feature emerging from substratum influence has entered into competition with a weak [+completive] feature resulting from superstratum influence. The optionality attested between pre- and post-verbal done in JC would be the result of the availability of both of these options of the completive feature. What seems to be happening at present is that the use of the strong option of the feature has become less common than the weak one. More specifically, the weak [+completive] feature is now more predominant than the strong one in the grammar of JC, implying that VP movement to [Spec,CompletiveAspP] is no longer forced at Surface-Structure. This analysis makes the prediction, therefore, that in future generations, VP23 54 Recall that preverbal done cannot occur with a [stative] VP either: Im done see mi: S/he has already seen me; *S/he has finished seeing me. The articulation of inflection in Jamaican Creole movement to [Spec,CompletiveAspP] may eventually cease to exist, as the weak option of the completive feature takes over entirely. Indeed this is already the case in JC varieties which oscillate between the mesolect and the acrolect. 3.2.1. Multiple base generation of Anterior tense The fact that the adverb aredi can occur in a sentence medial as well as a sentence final position can be accounted for in two ways: either it is directly generated in these two positions, or movement has taken place past it. This section briefly sketches these two analyses and argues that the movement hypothesis is the more favourable of the two. If aredi is the specifier of a functional projection as argued in Cinque (1999), then to generate it in two different positions implies that its corresponding functional projection can be generated in two different positions. However, as Cinque (1999, 22) notes: “it would make little sense to generate functional projections twice, once to the left, and once to the right of the verb (and its complements) (…) (T)he same rigid order of the AdvPs in post-complement position would have to be enforced through a specific principle duplicating the ordering principle for the functional heads in the pre-VP “space””. The uneconomical factor of an analysis generating aredi in two different positions renders this approach conceptually unsatisfactory. On an empirical level, this approach would fall short in accounting for the fact that the two independently generated adverbs cannot occur simultaneously: *Im aredi nyam di whole a i’ aredi S/he already eat the whole of it already „S/he already ate the whole of it already‟ Indeed where adverbs are generated in two different positions this is indicated by the fact that their simultaneous presence does not render the sentence unacceptable: John twice knocked on the door twice (Cinque 1999, 27) 3.2.2. Movement of (extensions of) VP across Anterior The fact that adverbs like „already‟ and its analogues in other languages can occur either sentence-medially or sentence-finally without yielding any perceptible change in interpretation seems best captured by the analysis whereby movement of the VP or of its extensions can occur across this adverb. This type of movement is demonstrated in Cinque (1999, 22) for Italian: (125) a. A Natale, credo che avesse completamente perso la testa di GIA „At Christmas, I think he had completely lost his mind already‟ b. A Natale, credo che avesse di già [completamente perso la testa] 55 Stephanie Durrleman Cinque (1999, 22) explains that “(u)nder this alternative, we can account for the „scope under reconstruction‟ property typical of movement (whereby completamente is under the scope of di già to its right), and at the same time derive the apparent subversion of the relative order of the AdvPs, otherwise unexpected in a non wh-type movement because of the ensuing Relativized Minimality violation. Given that the AdvP di già is crossed over not by the AdvP completamente directly, but by a larger phrase containing completamente, no Relativized Minimality violation takes place.” Notice that the relative order between aredi [anterior] and done [completive] was established as illustrated in (126): T [anterior] > Asp [+completive] This accounts for the grammaticality of (127a) and the ungrammaticality of (127b) a. Im aredi done nyam di whole a i’ S/he [anterior] [completive] eat the whole of it “S/he already finished eating it all” b. *Im done aredi nyam di whole a i’ Notice that although this fixed order cannot be subverted when both aredi and done precede the verb and its complements, it can be once aredi occurs in a post-VP configuration, as expected under the movement hypothesis illustrated in (128): a. Im aredi [done nyam di whole a i’] b. Im [done nyam di whole a i’]i aredi ti S/he [completive] eat the whole of it already „S/he finished eating it all already‟ Now consider the data in (129) which at first sight presents counter-evidence for Cinque (1999) where it is argued that functional projections respect the same fixed order in both pre- and post-VP configurations respectively: in (129) both done and aredi follow the VP yet the order established between the two in a pre-VP configuration (127) is subverted: Im nyam di whole a i done aredi S/he eat the whole of it [completive] [anterior] “S/he finished eating it all already” Under the VP-movement hypothesis established for completive done, however, this can be accounted for: in (129) movement has occurred in two steps: firstly, the 56 The articulation of inflection in Jamaican Creole VP has moved to [Spec, Completive], then the entire AspPCompletive has moved on 24 to the specifier of a functional projection preceding TPAnterior : (130) … FP F‟ Spec F° TPAnterior T‟ Spec aredi T° AspCompletive Asp‟ Spec Asp° done VP nyam di whole a i’ Notice also that the successive movement illustrated in (130) accurately accounts for the scope facts of this sentence: Firstly the VP is interpreted as being in the scope of completive aspect although it is situated on its right at S-structure: what is completed is the act of eating something. Since done is generated in CompletiveAsp°, it c-commands the trace of the VP so that its taking scope over this VP is to be expected. Secondly, the entire CompletiveAspP is in the scope of [Spec,TPAnterior] realised by aredi, although this projection also surfaces to the right of [SpecTPAnterior]: what has already taken place is the completion of the act of eating something. 24 Notice that VP-movement past AspCompletive° could also potentially take place to the specifier of a functional projection FP situated to the left of AspCompletiveP. However if movement of the complement of CompletiveAsp° is triggered by the need to check a strong completive feature, then it is most plausible that this movement takes place to [Spec,AspCompletiveP] where the completive feature is located. Movement to the specifier of a FP such as that located to the left of Anterior Tense, which gives rise to a particular focus on the adverb crossed, could then be reserved for the purpose of focussing clause internal adverbs. 57 Stephanie Durrleman Again, since aredi c-commands the trace of CompletiveAspP, then it follows that the scope properties evoked here are derived under reconstruction. 4. Overt functional structure of JC: evidence for the framework in Cinque (1999) Cinque‟s (1999) structure given in (19) is repeated in (131) with bold letters applied to the evidence drawn from JC: (131) [Frankly Moodspeech act [fortunately Moodevaluative [allegedly Moodevidential [probably Modepistemic: shuda, wuda, maita, mosa, kuda [once T(Past):did [then T(Future):wi [perhaps Moodirrealis [necessarily Modnecessity:mos [possibly Modpossibility [willingly Modvolition [inevitably Modobligation:haffi [cleverly Modability/permission:kyan [usually Asphabitual [again Asprepetitive(I) [often Aspfrequentative(I) [quickly Aspcelerative(I) [aredi T(Anterior) done1 [no longer Aspterminative [still Aspcontinuative [always Aspperfect(?) [jus Aspretrospective [soon Aspproximative [briefly Aspdurative [characteristically (?) [? Aspgeneric/progressive: a [almost Aspprospective: go [completely Aspcompletive(I): done2 [tutto AspPlCompletive [well Voice [fast/early Aspcelerative(II) [completely AspSgCompletive(II) [again Asprepetitive(II) [often Aspfrequentative(II): reduplicated verb … JC fits harmoniously into Cinque‟s (1999) structure: (131) illustrates that one does not contradict the other. It must be noted, however, that this language does not 25 at first sight provide direct evidence for separating T past from T future , nor for separating Mod obligation from Mod ability/permission. If each of the members of these pairs of markers in JC were to be inserted under different heads such as that implied by (131), then their mutual exclusion could not be explained in terms of competition for the same position. In light of (131), therefore, the impossible cooccurrence of did/wi and haffi/kyan in JC remains to be explained. Markedness thoery may, once developed, provide a means for accounting for these facts. 25 58 It is not entirely clear that the data given in (Cinque 1999) from Guyanese Creole can really be taken as evidence for the structure T past > T future either: Jaan bin gu riid: J. PAST FUT read „J. would have read‟ (Cinque 1999, 59, taken from Gibson 1985, 585). Indeed gu here looks like JC go, the Asp [prosp] marker. Consider that wuda and [past] + [prosp] may be found to yield a similar interpretation in JC also: Jan did go riid/ Jan wuda riid … bot mi stap ‘im: „John was going to read/ would have read … but I stopped him‟. The articulation of inflection in Jamaican Creole 5. Conclusion This paper has concentrated on exploring the clausal structure of a basilectal variety of JC. In such varieties, there is an absence of morphological verbal inflection. Inflection is articulated by means of independent inflectional particles, or TMA markers. Manifestations of TMA markers, like that of morphological inflection and adverbs, may serve the generative linguist as a key source of evidence in identifying clausal functional projections (Pollock (1989), Belletti (1990), Cinque (1999)). A recent framework (Cinque (1999)) postulates a good 30 functional projections in the clausal domain. The data from JC discussed in this work give direct overt evidence for over a third of these projections. Future research of the notion of markedness may prove insightful in determining if the functional structure here attested for JC is entirely present in all clauses of the language through marked or default values. Pursuing this line of reasoning, it becomes conceivable that JC exploit the entire array of functional structure postulated by the framework (Cinque 1999). If this rich functional structure is ultimately proven present in every clause through default values, the articulation of inflection in this language, and all languages, would prove to be much richer than that which is overtly manifested. The framework here adopted (Cinque (1999)) is highly restrictive in that the numerous functional projections postulated are argued to universally respect a rigidly fixed order. The overt evidence for functional clause structure provided by JC proves directly hierarchically compatible with this rigid order: There is a transparent systematic match between markers in JC and the fixed hierarchy of functional projections postulated by the framework. Surface differences in the structures can be accounted for in terms of IP-internal movement. This work therefore upholds the universality of the architecture of the clause as provided by Cinque (1999). References ABOH, E.O. (1998) From the Syntax of Gungbe to the Grammar of Gbe, Doctoral Dissertation, University of Geneva. ADAMS, L.E. (1995) Understanding Jamaican Patois: An Introduction to Afro-Jamaican Grammar. Kingston: Kingston Publishers Limited. ARENDS, J., KOUWENBERG, S., & SMITH, N., (1995) „Theories focusing on the nonEuropean input‟, in Pidgins and Creoles: An introduction Eds. Arends, J., Muysken, P. & Smith, N. Amsterdam: John Benjamins. AVOLONTO, A. (1992) AspP et la catégorie INFL en fongbè. Journal of West African Languages 22 (1): 97-113. BAILEY, B. (1966) Jamaican Creole Syntax: A Tranformational Approach, Cambridge University Press. BAILEY, B. (1971) Can dialect boundaries be defined?, Pidginization & Creolization of Languages, Ed. Dell Hymes. Cambridge University Press. 59 Stephanie Durrleman BAKER, M. (1985) “The Mirror Principle and Morphosyntactic Explanation”, Linguistic Inquiry, 16: 373-415. BAKER, M. (1989) „Object sharing and projection in serial verb constructions‟, Linguistic Inquiry 20: 513-554. BAKKER, P. (1995) Pidgins, in Pidgins and Creoles: An introduction Eds. Arends, J., Muysken, P. & Smith, N. Amsterdam: John Benjamins. BELLETTI, A., (1990) Generalized Verb Movement. Turin: Rosenberg and Sellier. CASSIDY, F.G. (1961) Jamaica Talk: Three Hundred Years of the English Language in Jamaica, Macmillan & Co Ltd.: London. CINQUE, G. (1992) „On tout/tutto and the Syntax of Past Participles in French and Italian‟ unpublished ms., University of Venice. CINQUE, G. (1994) „Evidence for Partial N-movement in the Romance DP‟ in Cinque et al. (eds.) pp. 85-110. CINQUE, G. (1999) Adverbs and Functional Heads: A Cross-Linguistic Perspective, Oxford University Press, New York. COMRIE, B. (1976) Aspect. Cambridge University Press, Cambridge. DA CRUZ , M. (1995) „Aspectual verbs fó, vò „finish‟ in Fongbè‟. The Linguistic Review 12, pp. 361-380. DECAMP, D. (1971a) „The Study of Pidgin and Creole languages‟, in Pidginization and Croelization of Languages, Ed. Hymes, D., Cambridge University Press pp. 13-43. DECAMP, D. (1971b) „Towards a generative analysis of post-creole continuum‟, in Pidginization and Croelization of Languages, Ed. Hymes, D., Cambridge University Press pp. 349-70. DEGRAFF, M. (ed.) (To appear) Creolization, Language Acquisition & Language Change, MIT Press, Cambridge (Mass.) DELFOSSE, R. (1997) Viewpoint: Patois- The rights and wrongs. The Weekly Gleaner, June 1117, 1997, Kingston. DURRLEMAN, Stephanie (1999) „The architecture of the clause in Jamaican Creole‟, ms. University of Geneva, to appear in GG@G. EDWARDS, Walter F. (1991) “A comparative description of Guyanese Creole and Black English preverbal aspect don.” In Verb Phrase Patterns in Black English and Creole, Walter F. Edwards and Donald Winford (eds.) Detroit: Wayne State University Press. GADELII, K.E. (1997) Lesser Antillean French Creole and Universal Grammar, Doctoral Dissertation, Gothenburg Monographs in Linguistics 15. GIBSON, K. (1986) „The Ordering of Auxiliary Notions in Guyanese Creole‟, Language, 62. Pp. 571-586. JACKENDOFF, R. (1972) Semantic Interpretation in Generative Grammar. MIT Press, Cambridge (Mass.) JAKOBSON, R. (1971) [originally 1939] “Signe zéro”, in Selected Writings, vol.II, Mouton, The Hague, pp.211-219 KAYNE, R. (1994) The Antisymmetry of Syntax. MIT Press, Cambridge (Mass.) 60 The articulation of inflection in Jamaican Creole LAMIROY, B. (1987) „The complementation of aspectual verbs in French‟. Language 63: pp. 278-298. MAGLOIRE-HOLLY, H. (1982) „Les modaux: auxiliares ou verbes?‟, in Lefebvre, C. et al. (eds.), Syntaxe de l’Haïtien. Ann Arbor, Karoma. MAHAJAN, A. (1990) The A/A-Bar Distinction and Movement Theory. Doctoral Dissertation, MIT. MÜHLHÄUSLER, P. (1986) Pidgin and Creole linguistics. Oxford: Blackwell. PINKER, S. (1994) The Language Instinct. Penguin Books: England. POLLOCK, J.-Y. (1989) Verb movement, Universal Grammar and the structure of IP, Linguistic Inquiry, 20, pp. 365-424. RIZZI, L. (1990) Relativized Minimality, MIT Press, Cambridge (Mass.) ROMAINE, S. (1994) Pidgin and Creole Languages , Longman New York. RUSSEL, T. (1868) The Etymology of Jamaican Grammar, by a young gentleman, Kingston: De Cordova, McDougall. SEBBA, M. (1993) London Jamaican, Longman U.K. SEUREN, P. & WEKKER, H (1986) „Semantic Transparency as a factor in creole genesis‟, in Muysken, P., & Smith, N., (eds.), Universals vs. Substrata in Creole Genesis. Amsterdam: John Benjamins pp. 57-71. SPORTICHE (1993) „Adjuncts and Adjunction‟, ms., UCLA. 61 “L-TOUS”, RESTRUCTURING AND QUANTIFIER CLIMBING Marco NICOLIS 1. Introduction The aim of this paper is to investigate the distribution of the object quantifier tout/tutto in French and Italian. Although in both languages this quantifier can appear in a “low” derived position (which will be argued to be one and the same in both languages), French displays an additional option: in some biclausal structures, tout can optionally appear in a high derived position (e.g. (1)), a structure known as “Quantifier Climbing”. (1) Il a tout voulu manger The clausal structure adopted in this paper is the highly articulated structure recently put forth by Cinque (1999). In this work, the “spirit” of Pollock‟s (1989) “Split-Infl Hypothesis” reaches its most radical formulation. The node traditionally known as “IP” is not just split into two different Functional Projections (see Pollock (1989), Belletti (1990) and much related work), but is made up of about thirty FPs. Cinque‟s proposal relies on the individuation of a rigid, crosslinguistically consistent, hierarchical ordering of adverbs. Each FP hosts an adverbial class in its Spec and may allow (modulo the different “length” of V movement in different languages) Verb movement through its Head. The existence of a hierarchy of FPs rather than of a (multiple) adjunction structure (see Chomsky (1995), ch.4) is further confirmed by those languages expressing adverbial modification by means of “particles” (therefore, heads) incorporated into the verb: the ordering of these heads is (under Baker‟s Mirror Principle) exactly the same found for adverbs in “abverbial 1 languages” . We report in (2) the adverbial hierarchy which constitutes the lowest 1 Cinque‟s (1999) account of the adverbial ordering in terms of an FP hierarchy, rather than adjunction, is a welcome conclusion under a restrictive theory of syntax, such as Kayne‟s (1994) antisymmetric program. In this system, adjunction to XP is never an option, even though Specs are in general considered elements adjoined to single bar constituents. 63 Marco Nicolis part of “IP”, noting, as expected, its consistency in different languages (Italian (2a), French (2b), English (2c) respectively). (2) a. solitamente>mica>già>sempre>completamente>tutto>bene b. généralment>pas>déjà>plus>toujours>complètement>tout>bien 2 c. usually>…>already>always>completely>…>well The ordering in (2) has been obtained by simply juxtaposing adverbs belonging to different semantic classes and noting that they can only appear in just one of the n possible orderings. The mechanics of this process is illustrated by the paradigm (3): (3) a. Quando lo andiamo a trovare, Gianni ha solitamente già finito di mangiare When him-cl. go to meet, Gianni has usually already finished to eat b. *Quando lo andiamo a trovare, Gianni ha già solitamente finito di mangiare c. Gianni ha già spiegato bene la lezione a Maria Gianni has already explained well the lecture to Maria d. *Gianni ha bene spiegato già la lezione a Maria e. Gianni capisce solitamente bene la lezione Gianni understands usually well the lecture f. *Gianni capisce bene solitamente la lezione Examples (3 a-b) show that the adverb solitamente must precede the adverb già. (3 c-d) show that già precedes bene. (3e-f) confirm the validity of the ordering solitamente>già>bene, showing that transitivity holds: solitamente must in fact precede bene. The reiteration of this procedure with all the different adverbial 3 classes yields the highly articulated structure of IP proposed by Cinque (1999) . The adoption of this rich clausal architecture poses some preliminary problems of “translation”: given the pre-pollockian clausal structure adopted in classical works on floating quantifiers (such as Sportiche (1988)) or even the not-so-highly2 3 64 The clausal structure proposed by Cinque (1999) is actually even more fine grained than reported in (2). In particular, several adverbs can occupy two distinct positions yielding two different semantic interpretations. Cinque (1999, ch. 1 p. 30) argues that in (ia) “slowly qualifies the entire event (each test could well have been rapid), whereas in (ib) it qualifies each test individually”: (i) a. He has been slowly testing some bulbs b. He has been testing some bulbs slowly As the point is irrelevant to our argument, in order to avoid unnecessary complexity we will keep on referring to (2) as the lowest part of the IP. In (3) we tested three adverbs quite “distant” from one another. It is worth observing that the closer two adverbial classes are, the more nuanced the grammaticality judgement becomes. “L-Tous”, Restructuring and Quantifier Climbing articulated structure adopted by Cinque himself in works on leftward movement of tutto in Italian (see Cinque (1995, ch. 3, 9)), the positional characterisation of the landing site of these clause internal movements must be reconsidered; if “IP” is indeed constituted of about thirty adverbial FPs (and each one is probably “accompanied” by a “DP related” FP), the number of possible landing sites for movement dramatically increases. It is thus necessary to test the position of the moved material with respect to the fixed position occupied by adverbs in the relevant part of the clausal structure. The “positional analysis” will always be integrated by an analysis of the “typology of positions”: the individuation of a particular position in the structure will be constantly tied to the individuation of the typological class it belongs to. The matter is made more complex by the split between quantificational and modificational Ā positions recently proposed by Rizzi (1999), (2000). In Relativized Minimality (1990), Rizzi suggested that “the class of possible interveners triggering minimality effects is not coextensive to the class of target positions, but significantly wider” (Rizzi (2000)). In fact, it is well known that Negation, wh- elements and quantificational adverbs such as beaucoup/molto all pattern alike so far as their ability to block wh- extraction of non arguments in 4 “pseudo opacity” structures is concerned (Rizzi (1990)) , as (4 b, c, d) clearly show: (4) a. Combien i a-t-il consulté [ti de livres] ? b. *Combien i a-t-il beaucoup consulté [ti de livres]? c. *Combien i ne sais-tu pas résoudre [ti de problèmes] ? d. *Combien i sais-tu [comment résoudre [ti de prblèmes] ? On the other hand, if all adverbs sit in a [Spec, FP] and all such positions are (by assumption) uniformly Ā position, one should expect that all adverbs may trigger minimality effects with respect to wh- extraction of non arguments. The prediction anyway is not borne out, as (4) clearly shows: (5) a. *Combien a-t-il beaucoup consulté de livres? b. Combien a-t-il attentivement consulté de livres? On the basis of data like (5) Rizzi (2000) proposes that Ā positions are actually to be split into two classes: quantificational Ā positions (Ā-q henceforth) and modificational Ā positions (Ā-m). If this idea is correct, one should find cases in which a modificational adverb acts as an intervener wrt to the fronting of another modificational adverb. Consider (6): 4 These data were originally noted by Obenauer (1983). 65 Marco Nicolis (6) a. *Rapidamente, i tecnici hanno probabilmente risolto ___ il problema 5 b. RAPIDAMENTE i tecnici hanno probabilmente risolto ___ il problema (6a) shows that the fronting of a modificational adverb across an adverb of the same type triggers RM effects. Since Focalization, contrary to simple fronting, involves movement to a Left Peripheral Ā -q position (Cinque (1990), Rizzi (1997) and much related work), no RM effect is triggered in (6b), as Rizzi (2000)‟s theory correctly predicts (see also (44) and fn. 22). The same kind of effect is further illustrated by the fronting of adverbs for V2 reasons in Dutch. Given the ordering helaas (unfortunately) > waarschijnlijk (probably) (7), only the former can be fronted but not the latter: this movement would violate RM, since both the landing site and the intervener are Ā-m positions (8). (7) a. Het is zo dat hij helaas waarschijnlijk ziek is “It is so that he unfortunately probably sick is” b. *Het is zo dat hij waarschijnlijk helaas ziek is (8) a. Helaas is hij ___ waarschijnlijk ziek “Unfortunately is he probably sick” b. *Waarschijnlijk is hij helaas ___ ziek c. Waarschijnlijk is hij ___ ziek (helaas>waarschijnlijk) Koster (1978) Having spelled out the background theoretical assumptions underlying this work, we may now turn to the analysis of the movement of tout/tutto. 2. The movement of tout/tutto: the low position It is well known, at least since Belletti‟s (1990) influential work on Verb movement, that both French and Italian object quantifiers tout/tutto occupy a derived position at S-Structure (or at Spell Out) despite the fact that tutto follows the Past Participle and tout precedes it, as (9) shows. (9) a. Jean a {tout} mangé {*tout} b. Gianni ha {*tutto} mangiato {tutto} The contrast in (9) is amenable to the well known difference between the two languages concerning Past Participle movement; since Past Participle in Italian moves much higher that its French counterpart, the data in (9) are readily accounted 5 66 (i) shows that the correct hierarchical ordering of the two adverbs rapidamente and probabilmente is probabilmente>rapidamente. (i) a. I tecnici hanno probabilmente risolto rapidamente il problema b. *I tecnici hanno rapidamente risolto probabilmente il problema “L-Tous”, Restructuring and Quantifier Climbing for. This conclusion is further confirmed by the distribution of low adverbs; since both quantifiers show the same distribution with respect to the fixed positions occupied by low adverbs (they precede (see (10)), and follow (see (11)) the same adverbial classes), the obvious conclusion is that they occupy the same structural position in both languages: (10) a. b. c. d. Ha già detto tutto bene Gianni *Ha già detto bene tutto Gianni Elle a tout très mal compris *Elle a très mal tout compris (11) a. b. c. d. Ha rifatto già tutto bene Gianni *Ha rifatto tutto già bene Gianni Jean a déjà tout refait 6 *Jean a tout déjà refait Cinque (1999) Although the conclusion that both tout and tutto occupy the same derived position at S-Structure is fairly uncontroversial in the literature, there is to my knowledge no general agreement about the categorial status of this derived position. As a matter of fact, any of the three aforementioned structural classes for XPs (A, Ām, Ā-q) has been proposed in different works to be the class the landing site of the movement of tout/tutto belongs to. Let‟s now consider each of the three alternatives 2.1 Movement to an Ā-q position 7 Belletti (1990, pag. 78) argued that “[…] it can be assumed that rien/tout have the defining property of obligatorily undergoing a QR-type movement process in the syntax (presumably to be assigned scope already at this level of representation).” QR is typically an LF movement moving variables to a scope assigning position, therefore the landing site of this movement must be an Ā-q position, since scope assigning positions are by assumption quantificational. The position occupied by tout/tutto seems to be hardly characterizable as a scope assigning Ā-q position. First of all, claiming that being subject to a sort of pre-LF QR is a “defining property” of tout/tutto is a mere stipulation, unless independent empirical evidence (which is, to my knowledge, lacking) supporting this thesis is brought up. Even assuming that some evidence in the desired direction could be found, it still needs to 6 7 As in all the examples quoted in this paper, unless otherwise indicated, the sentences in (10), (11) are to be read with a flat intonation, namely, one which doesn‟t make any constituent more “salient”. The same argument is in a later passage extended to Italian as well. 67 Marco Nicolis be explained why the Quantifier can remain in base position when focalized, modified or coordinated, maybe with a very slight marginality (see (12)): (12) a. b. c. d. ?Il a repris TOUT ?Il a repris presque tout ?Il n‟a lu absolutement rien ?Il a lu tout ou presque tout Belletti (1990) Belletti (1990)‟s account of the data in (12) consists in simply proposing that “It could be assumed as in Kayne (1975) that the rule moving the quantifiers is suspended in conjunction with stress or heaviness” (Belletti (1990), page 138 fn. 67). Also in this case, there seems to be no independent empirical reason forcing such a conclusion. Moreover, since QR involves the interpretability of the moved elements, some kind of interpretive difference between the bare quantifier cases (such as (10)), in which QR would have applied, and the non-bare cases (such as (12)), where no QR is supposed to have taken place should be detectable. The prediction is clearly not borne out, though. Alongside these theoretical problems, the hypothesis according to which tout/tutto move to an Ā-q position seems to be empirically inadequate; in fact, these quantifiers can move across a quantificational adverb like beaucoup, without triggering any RM violation. In order to show that this argument goes through, it is first necessary to make sure that the position occupied by beaucoup/molto is indeed in between the moved quantifier and its trace and therefore qualifies as a potential intervener. Cinque (1999) observes that beaucoup/molto and bien/bene can be easily coordinated and 8 seem to surface in a quasi-complementary distribution , two (somewhat loose) indications that they may occupy the same position. However, this hypothesis does not seem very promising from a semantic viewpoint. In fact, the FPs constituting Cinque‟s hierarchy are each representative of a peculiar semantic class: beaucoup/molto and bien/bene, if anything, encode very different semantic properties and it is therefore highly implausible that they occupy the same position. Cinque (1999) further noticed that there are a few cases (like (13)) in which molto and bene can cooccur without requiring coordination. In all the relevant cases bene must necessarily follow molto. It is therefore plausible to assume that the two adverbs occupy two distinct, although contiguous positions. 8 Consider for example the sentences in (i), where coordination seems to be the only option ((ia) is grammatical in the irrelevant interpretation in which molto locally modifies bene): (i) a. *Gianni ha mangiato molto bene b. *Gianni ha mangiato bene molto c. Gianni ha mangiato molto e bene. 68 “L-Tous”, Restructuring and Quantifier Climbing (13) a. (?)Ballava molto assai bene anche il fratello 9 b. *Ballava assai bene molto anche il fratello (10) above shows that the derived position tutto moves to precedes the one filled by bene (and the one filled by completamente, assuming (2) is correct), therefore molto/beaucoup is indeed a potential intervener for the movement of tout/tutto. The examples in (10) jointly with those in (13) correctly predict that the ordering beaucoup/molto>tout/tutto is impossible for configurational reasons (see (13b)), as (14) shows: (14) a. *Il a beaucoup tout apprecié b. *Ha apprezzato molto tutto We may now turn to the crucial examples in which the quantifier moves to a position higher than the one occupied by beaucoup/molto: (15) a. Il a tout beaucoup apprecié b. Ha apprezzato tutto molto If the landing site of tout/tutto were an Ā-q position, an RM effect should arise, given the quantificational nature of the adverb beaucoup/molto. The absence of any such effect seems to suggest that the landing site of tout/tutto is not a quantificational position. A possible objection to this conclusion is that whatever class the position occupied by tout/tutto belongs to, these elements possess an inherent [+Quantificational] feature, which should emerge in any position they move to, being lexically determined. This idea seems to be operative in such domains as variable binding, as (16) shows. (16) a. Tuttoi, non dovrà vender(*lo) b. I suoi libri, non dovrà vender*(li) 9 The judgment is quite delicate in this case, because the presence of an inverted subject coocurring with a low adverb yields per se a marginal sentence as Rizzi (1996) showed (see (i)): (i) ?Ha giocato bene Gianni ??Ha fatto tutto bene Gianni The presence of the inverted subject is nonetheless very important in these cases: in fact, given the general possibility for any deaccented extraposed XP to appear in sentence in final position, it is necessary to show that the acceptability of a sentence like (13a) with a preverbal subject is not the result of the application of some “rescuing strategy”, like extraposition of the sentence final adverb. 69 Marco Nicolis Contrary to full DPs (16b), bare quantifiers such as tutto can bind a variable from a clearly non quantificational position (Topic in (16)) without requiring a resumptive clitic (16a). But not only don‟t they require a resumptive clitic, they actually do not tolerate it, plausibily because it would qualify as a closer potential binder for the variable in object position: this would lead to a case of vacuous quantification, since the operator-like item tutto would have no variable to bind. These facts however do not automatically extend to the domain of Locality, which on the contrary doesn‟t seem to be sensitive to intrinsic features. As a matter of fact, we can consider variable binding and the computation of Locality effects two independent phenomena that exploit different computational mechanisms. Rizzi (1999) proposes that a Spec position acquires the featural characterization relevant for the computation of Locality effects from its local head X°. Therefore the intrinsic features of a moved element are irrelevant for the computation of RM effects. Given the characterization of Relativized Minimality reported below and in particular (23 (i)), Rizzi (1999) proposes that “same structural type” is to be understood as “(i) head or Spec, and (ii) Spec licensed by features of same class”; the licenser of a given Spec is its local head. 2.2. Movement to an Ā-m position The hypothesis of a movement to Ā-m position had been explicitly put forth in various works by G. Cinque. Cinque (1995, ch.9) argued that “[…] only tout/tutto, among XPs, could move to what appears to be an adverbial-like Ā position, […] while retaining their ability to bind a variable”. The hallmark of bare quantifiers such as tout/tutto would be their being structurally reduced, “a complementless QP, unspecified for the features N V” (Cinque (1995), ch. 9, pag. 282). The structural peculiarity characterizing these QPs could be responsible for their ability to escape 10 the requirements of the Case Filter : in fact, being a complementless QP implies being not a N°‟s extended projection; therefore, the Case Filter can be ignored and the quantifier can move to an Ā-m position. The adverbial nature of tout/tutto is further discussed in Cinque (1999), where it is claimed that these Quantifiers encode a particular type of Completive Aspect. Cinque (1999), quoting previous work by Bybee, distinguishes two types of 10 70 I will not commit myself to the existence of the Case Filter; in particular, the ideas exposed in this paragraph will be rejected in a later one. Anyway, for consistence‟s sake, I will adopt a standard formulation of the Case Filter, such as that reported in (i) (taken from Chomsky (1995, ch. 1 page 111)): (i) Every phonetically realized NP must be assigned (abstract) Case Of course, the Case Filter extends also to all the nominal projections that can be considered extended projections of N° in Grimshaw‟s (1991) sense. “L-Tous”, Restructuring and Quantifier Climbing Completive Aspect in the case of a definite plural object, one expressing “that the plural set has been totally affected (i.e. each member of the set has been affected) and 2) that each member of the set has been totally affected”. Cinque proposes that, although several languages do not distinguish between the two (e.g. English (17), where the particle up ambiguously encodes both), others, like Polish (see (18)), do. (17) I ate up the sandwiches (18) Po-prze-czyt-yw-o am wszystkie jej ksiazki COMPL-COMPL-read-HAB-PAST all her books “I have read all her books occasionally one after the other and right through” Cinque (1999) Cinque calls the two aspects “plural completion” and “singular completion”. Back to Italian, Cinque (1999) observes that “it is temping to see completamente and tutto as the specifiers corresponding to „singular‟ and „plural‟ completion, respectively”; he further observes that in fact, tutto can only refer to a plurality of items (a singular object in not an appropriate answer to the question (19)) and that completamente obeys the same restrictions that the corresponding functional heads obey in other languages, namely it is “incompatible with situations lacking internal stages and a natural end point”. Thus, in the relevant respect, Hungarian, Chinese and Italian all pattern alike, modulo the realization of the head (Hungarian and Chinese) or of the Spec (Italian) of the same FP. (19) Hai trovato tutto? Cinque (1999) (20) a. *Karoly tejet I vott meg K. milk drank up „K. drank milk up‟ b. Ta xiao-(*wan)-le He smile-(COMPL)-PERF „He smiled completely‟ c. *Gianni ha riso completamente Cinque (1999) This hypothesis, although the point is not fully clarified by Cinque (1999), seems to imply that the quantifier tutto acquires its modificational „completive‟ nature by virtue of sitting in a modificational [Spec, FP] at SS or Spell-Out; in particular, the restriction concerning the possibility of referring only to singular referents should be the result of occupying the Spec of the AspectPluralcompletive projection. If this is correct, the prediction is that this restriction should be suspended in those contexts in which the modified or focalized quantifier tout/tutto remains in base position. (21) shows that the prediction is not borne out, in fact in both (21a) and (21b) the quantifier can only refer to a singular entity. 71 Marco Nicolis (21) a. Hai pulito tutto bene? b. Hai pulito bene quasi tutto? But the most important drawback of Cinque‟s analysis has to do with locality. If we adopt Rizzi‟s (2000) system, it is obvious that the quantifier tout/tutto can cross 11 low modificational adverbs, as (10a), (10c) show , without triggering any RM effect. A possible objection to this conclusion could be that tout/tutto are not inherently modificational elements (in the way most adverbs are), but acquire their modificational status only when they reach the relevant [Spec, FP] and therefore when the crossing takes place no RM effect arises because the moved element is not endowed with the relevant modificational feature the crossed element is. However, this objection does not go through. In fact, Rizzi‟s system (Rizzi (1990) as well as Rizzi (2000)) is essentially a representational system, namely one in which possible intervening effects are computed from the final representation rather than taking into account the whole derivation. Rizzi‟s (2000) basic idea of Relativized Minimality (see (22)) is in fact technically implemented making reference to the notion of Minimal Comfiguration (see (23)), which in turn relies on a representational notion of chain (see (24)), the typical locus where an MC is created. (22) In the configuration …X…Z…Y… “Y cannot be related to X if Z intervened and Z has certain characteristics in common with X. So, in order to be related to X, Y must be in a minimal configuration with X, where minimality is relativized to the nature of the structural relation to be established” Rizzi (2000) (23) Y is in a Minimal Configuration (MC) with X iff there is no such Z that (i) Z is of the same structural type as X, and (ii) Z intervenes between X and Y Rizzi (2000) (24) (A1,…An) is a chain iff, for 1 ≤ i < n 12 (i) Ai=Ai+1 (ii) Ai c-commands Ai+1 (iii) Ai+1 is in a MC with Ai Rizzi (2000) The combination of (22), (23), (24) implies that intervention effects are to be detected solely on the basis of the structural properties of the target position and of the intervener, therefore if they share the same feature (e.g. modificational in (10a), (10c), 11 12 72 10b, 10d show that bene>tutto is (expectedly) the only possible ordering between the unmodified quantifier tout/tutto and the bien/bene class of adverbs. See Chomsky (1995) for a theory of traces as copies. “L-Tous”, Restructuring and Quantifier Climbing as Cinque‟s theory seems to imply) a RM violation is expected, no matter which features the moved element is inherently endowed with. Given the lack of RM effects in the cases under examination, it is fair to conclude that the head of the FP to which the Quantifier moves in not endowed with the feature [+Modificational]. 2.3. Movement to an A position The option we are left with is movement to an A position. This hypothesis, defended in Cardinaletti & Starke (1994), (1994b), readily provides an explanation for exactly those aspects that were problematic for the hypotheses considered so far. In particular, it convincingly accounts for the possibility of leaving a non bare quantifier in base position and for the lack of RM effects in cases like (10), (15). The latter point is straightforward. In fact, if the target position is an A position, the intervention of an Ā element of whatever kind is not expected to trigger any RM effect. The case wuold be fully parallel to another case in which tout moves to an indisputable A position, namely subject position, crossing a modificational adverb (vraiment) without triggering any RM effect: (25) a. Tout a été vraiment apprécié par Jean Let‟s now consider (25b): (25) b. Gianni ha spiegato loro tutto bene Cases like (25b), in which the weak pronoun loro has moved to a derived A position, probably for Case reasons, (see below and Cardinaletti & Starke (1994)) across the quantifier tutto, cannot be considered counterexamples to the conclusion that tutto moves to an A position. In fact, (25b) is a member a of a larger set of cases in which a “short” A movement does not trigger RM violations; A movements of elements targeting positions within the verbal extended projection of the clause they were generated in, typically do not induce RM violations. The prototypical and most discussed cases of this sort are Object Shift and the (fully grammatical) crossing of 13 the [Spec, AgrOP] position by the subject on its way to [Spec, AgrS] . The typical 13 I do not really commit myself to any of the many proposal in the literature trying to account for the perfect acceptability of this movement, since the descriptive datum is all I really need in my argument. However, for completeness sake, it is worth noting that Chomsky (1995, ch. 3) argued for the well known device called equidistance, Roberts (1997) proposes that RM effects are voided if in the structural context …X…Z…Y… X and Z are non-distinct, where distinctness is defined as in (i): (i) α is a position nondistinct from β if α and β are of the same category and are in the extended projection of a single lexical category L. Since both AgrOP and AgrS are in the same extended projection of the verb V, the intervener and the target are non-distinct, therefore RM is not violate, given (i). 73 Marco Nicolis RM violation involving A positions is in fact superaising, which in fact involves an intervener and a landing site which are in two distinct extended projections, being 14 both positions required by EPP . As to the possibility of leaving a non bare quantifier in base position (such as modified, coordinated, focalized tout/tutto), Cardinaletti & Starke (1994) propose that what is at stake in these cases exactly corresponds to what is found with a large class of personal pronouns. Many personal pronouns have three distinct forms: a heavy, a weak and a clitic form (e.g. Italian dative plural pronoun a loro (Strong), loro (Weak), gli (clitic)). As the example in the parenthesis show, the strong form is often morphologically richer than the weak form, which is in turn richer than the 15 clitic form . This difference correlates to a distributional difference: strong elements occupy a position lower that those occupied by weak and clitic elements, as (26) clearly shows: (26) a. \Non gli metterò mai il cappuccio b. Non metterò mai loro il cappuccio c. Non metterò mai il cappuccio a loro The strong version of the pronoun can typically be coordinated, focalized and modified, while the weak (and a fortiori the clitic) forms cannot, as (27) vs. (28) and (29) shows: (27) a. Ho parlato [a loro] e [a loro] b. Ho parlato [a LORO], non [a LORO] c. Ho parlato solo [a loro] Finally, Rizzi (Univ. of Siena seminars, 1999) proposes that in the cases we are considering the intervener is not to be identified with just the head of the chain , but with the whole chain. Therefore, the crossing of [Spec, AgrOP] by the subject is not problematic since the tail of the chain formed by the movement of the object to AgrOP is not crossed by the subject, which is of course generated higher. 14 15 74 A case of superaising is illustrated in (i), where the subject of the most embedded clause in raised to the subject position of the matrix clause across the expletive subject it in [Spec, AgrS2]: (i) *[AgrS1 Johni seems that [AgrS2 it is likely [ AgrS3 t i to win]]] This is not necessary, though. In fact, in many cases the different classes are homophonous (German strong and weak forms sie, French weak and clitic forms il) , but it is never the case that a stronger form is morphologically poorer than a weaker counterpart. “L-Tous”, Restructuring and Quantifier Climbing (28) a. *Ho parlato [loro] e [loro] b. *Ho palato [LORO], non [LORO] c. *Ho parlato solo [loro] (29) a. *[gli] e [gli] ho parlato b. *[GLI] ho parlato, non [GLI] c. *Solo [gli] ho parlato Cardinaletti & Starke (1994) further show that when the structural conditions for the appearance of functionally equivalent pronouns are met, the weaker form is always preferred over the strong one. In other words, if nothing forces a strong form to occur (prominence, by and large, as in (27), (28), (29)), the weak form is chosen. (30) illustrates the point in different languages (from Cardinaletti & Starke (1994): (30) a. b. c. d. e. Jean {laC} regarde {*elleS} Gianni {laC} guarda {*leiS} Ich habe {ihnW} gestern {*ihn} eingeladen …, dass {zC/ *esW} toire isch …, dat {zeW/*zjjS} niet will komen (FRENCH) (ITALIAN) (GERMAN) (OLANG TIROLESE) (DUTCH) The properties observed so far for pronouns exactly mirror those of the quantifier tout/tutto. On the assumption that the quantifier tout/tutto has both a weak and a strong omophonous form (see fn. 12), all the data straightforwardly fall into place. (10) shows on a par with (30) that the weak form, which occupies a higher position (as (26) shows for pronouns), is preferred over the strong one when nothing forces the latter to appear. (12) shows for French that the contexts in which the quantifier is left in base position are exactly the same in which a strong pronoun shows up, namely modification (27c), (12 b, c), coordination (27a), (12d) and focalization (27b), (12a). (28) shows the same point in Italian: (31) a. Gianni ha capito bene assolutamente tutto (*assolutamente tutto bene) (modification) b. Gianni ha capito bene tutto o quasi (*tutto o quasi bene) (coordination) c. Gianni ha capito bene TUTTO, non solo il primo capitolo (*?TUTTO bene, non…) (focalization) d. *Ha fatto bene tutto anche Gianni (Ha fatto tutto bene anche Gianni) (bare quantifier) Cardinaletti & Starke (1994) account for the properties of weak elements in terms of structural deficiency: since their projection is structurally reduced (it 75 Marco Nicolis 16 contains less structural layers) , they need to compensate for such a deficiency by moving to an adequate [Spec, FP], in order to recover the properties they are not structurally endowed with. For example, it is argued in Cardinaletti & Starke (1994) that the property pronouns need to recover is case: the dummy marker a appearing in 17 (26c) would be the morphological realization of a case assigning head . Weak elements lack the relevant projection DP internally, therefore they must move to a “Case assigning” projection in the IP space, plausibly an Agreement projection (AgrOP or a similar projection if Agreement projections in the low IP space are more than traditionally assumed). This case based account does not automatically extend to all the weak elements, since Cardinaletti & Starke (1994) observe that some languages such as Greek have 18 weak adverbs , which of course do not need case. Determining the kind of structural deficiency affecting quantifiers is far from straightforward: in fact, it could only be assessed on the basis of a much deeper knowledge of the fine structure of QPs. If Cinque is right in claiming that bare quantifiers are complementless QPs, then a case based account of their structural deficiency would prove implausible, since they should be able to escape the Case Filter. If Cardinaletti & Starke‟s (1994) approach (lack of the highest layer(s)) is on the right track, then bare quantifiers could still perhaps be considered extended projections of some empty N°, thus being subject to the Case Filter. Their deficiency could then be due to Case reasons on a par with full DPs. 16 17 18 76 Cardinaletti & Starke (1994) propose that these projection lack the uppermost layer (CNP), which, just like CP would be able to host a prepositional dummy marker, such as a in (26). I do not really commit myself to all the technical details of this proposal, given the many recent works showing that both the sentential CP and the DP are probably to share the same fate of the node IP, namely split into many FPs. The overt morphological realization of this head is of course not a necessary condition for case assignment: the necessary condition is the structural presence of such a case assigning head. The Greek adverb sigo is considered by Cardinaletti & Starke (1994) the weak counterpart of sigà, as their distribution in (i) confirms: (i) a. To {*sigà} évrasa {sigà} b. To {sigo} évrasa {*sigo} It slowly I boiled slowly As predictable, only the strong form sigà can be grammatically coordinated: (ii) a. To évrasa sigà ke kalà b. *To sigo ke kalo évrasa It slowly and well I boiled “L-Tous”, Restructuring and Quantifier Climbing 3. The movement of tout: the high position As anticipated in the Introduction, in some, but not all biclausal structures, the French quantifier tout can occupy a high derived position, an option unavailable to its Italian counterpart tutto. Consider (32) vs. (33), adapted from Kayne (1975 ch. 1) and Kayne (1984 ch. 3): (32) a. b. c. d. e. Il a tout voulu manger Elle n‟a rien pu boire Vous n‟auriez rien osé dire (de plus) Tu vas tout devoir apprendre par cœur Il a tout failli rater Kayne (1975) (33) a. *Elle va tout avouer mépriser (Elle va avouer tout mepriser) b. *Elle est tout montée mettre à la poubelle (Elle est montée tout mettre à la poubelle) c. *Elle va tout courir mettre dehors (Elle va courir tout mettre dehors) The class of verbs allowing Quantifier Climbing by and large corresponds to the class of Restructuring verbs in restructuring languages. This datum is surprising, since Modern French lacks Restructuring, which is shown by the ungrammaticality of Long NP movement with mediopassive “se” (34a), Auxiliary Change (34b) and Clitic Climbing (34c), the typical hallmarks of Restructuring. (34) a. *Les nouvelles maisons se commenceront à construire b. *Pierre est voulu venir avec nous c. *Jean le veut faire Although Modern French lacks Restructuring, Old French was a restructuring languge, as (35) shows: (35) a. Nuls om mortals no.l pod penser No man mortal not.it(cl.) can think b. Elle la commença a desirer She it(cl.) begins to desire c. Vous estes volue apparoir You are wanted appear d. car amors ne se puet celer as loves not “se”(mediopassive) can hide (clitic climbing) (clitic climbing) (auxiliary change) (long NP movement) (35) could lead one to hypothesize that the occurrence of Quantifier Climbing with the class of Restructuring verbs in French is a residual phenomenon, since French was a restructuring language. The hypothesis seems however hard to 77 Marco Nicolis maintain, since Italian, a full fledged Restructuring Language, strongly disallows Quantifier Climbing: (36) **Gianni tutto vuole mangiare We now face a somewhat paradoxical situation: French, a non restructuring language allows Quantifier Climbing in sentences involving restructuring verbs, whereas Italian, a restructuring language, simply disallows Quantifier Climbing. Before coming back to this intricate problem, let‟s first of all determine the structural class of the position the “climbed Quantifier” moves to. 3.1. Quantifier climbing: the landing site 19 Although Modern French does not allow Restructuring, Clitic Climbing is grammatical in some causative structures, plausibly a residue of earlier stages of the language, when Restructuring was an option: (37) Jean la fait manger par/à Paul Given Shlonsky (1991) analysis of the structure of QP, according to which tout is generated in Q°, Quantifier Climbing could in principle be though of as head movement, on a par with (37). But a clear difference between Clitic Climbing and Quantifier Climbing emerges with respect to RM effects. Clitic Climbing clearly IS head movement, given the impossibility in both Italian and French for the clitic to cross an intervening head: (38) a. *Gianni li vuole non vedere (OK… non li vuole vedere) (OK… vuole non vederli) b. *Gianni non li vuole che Maria veda c. *Jean/Cela lui a fait ne pas manger a l‟enfant (??Jean/Cela a fait ne pas manger sa soupe a l‟enfant) (Kayne 1989) On the other hand, the climbed Quantifier can cross a head (C° in (36)): (39) a. b. c. d. e. 19 78 ?Jean veut tout qu‟elle refasse ?Je veux tout que tu leur enlèves ?Je ne veux rien que tu fasses (d‟autre) ?Il faut tout que je leur enlève ?Il ne faut rien que tu fasses I adopt Kayne‟s (1989) account of cliticization as composite movement, namely XP movement to [Spec,AgrOP] followed by X° movement up to the cliticization site. “L-Tous”, Restructuring and Quantifier Climbing In all the cases in (39) the Quantifier crosses the Subject position of the embedded clause, thus showing that the position it targets is not an A position. The Quantifier tout appears to be able to move out of wh- islands (40), thus showing that this position is not an Ā-q position: (40) a. ?Il a tout su ou mettre Kayne (1989) On the contrary, the intervention of a modificational adverb yields ungrammaticality: (41) a. *?Il a tout voulu obstinément voir en même temps I take the evidence in (39), (40), (41) as a clear indication that the climbed Quantifier moves to an Ā-m position. This conclusion seems to be inconsistent with the grammaticality of the 20 extraction of tout from Inner Islands, as (42b) shows : (42) a. Il aurait voulu ne tout dire qu‟à son advocat b. *Il aurait tout voulu ne dire qu‟à son advocat Kayne (1975) It is well known, at least since Ross (1983), that wh- islands are stronger than Inner Islands; therefore, the asymmetry (40) vs. (42) is unexpected: the mild deviance of (40) should imply the perfect grammaticality of (42), contrary to fact. Negation has long been considered an Ā-q element; Rizzi (1990) observed that “…negation patterns on a par with other uncontroversial Ā binders such as wh21 elements and adverbial QPs”, as (43) shows : (43) a. Combien a-t-il lu [e de livres] b. Il a beaucoup lu [e de livres] c. Il n‟a pas lu [e de livres] 20 21 Rizzi (1990) The intervener in (41) is plausibly not ne, which is standardly assumed to be the head of a NegP, but rather a null negative operator (the covert counterpart of pas) sitting in the Spec of the projection headed by ne. The Ā-q nature of Negation is further shown by the systematic lack of ambiguity in sentences like (ib), contrary to structures like (ia): (i) a. Il n‟a [pas [résolu [beaucuop de problèmes]]] b. Il n‟a [pas [beaucoup résolu [e de problèmes]]] In (ia) the whole object can be grammatically Q-Raised at LF (whatever the reason allowing this structure: θ-marking of the object in a disjunctively formulated ECP, proper index assignment, D-Linking…), giving beaucoup wide scope. Beaucoup can only have narrow scope in (ib), because its raising across pas would violate RM, as (ii), the LF representation of (ib) with beaucoup assigned wide scope, shows: (ii) *beaucoup il n‟a [pas [t résolu [de problèmes]]]. 79 Marco Nicolis On the other hand, there seems to be good reason to believe that Negation also carries a [+Modificational] feature, as (44) shows: (44) *Rapidamente, Gianni non ha risolto il problema 22 This sentence is ungrammatical if the adverb is interpreted as simply fronted in a “Modificational” position in the CP space, plausibly beacause the crossing of pas 23 violates RM . Rizzi‟s (1999) theory accounts for the double featural specification of Negation ([+Modificational] (44), [+Quantificational] (43) and fn.21) by simply proposing that different positions can inherit more than one feature by their head. The table (45), taken from Rizzi (1999), is a characterization of all the possible featural combinations of [Q], [Mod], [Arg]. (45) ARG Q MOD EXAMPLES + + + + - + + + + - + + + + - John (non-quantificational subjects) * Wh, Foc Beaucoup, pas Noone (quantificational subjects) * Carefully Topic Given (45), the asymmetry (40) vs. (42) is no longer surprising: the Wh word ou in (40) is specified as [+Q], therefore it does not qualify as an intervener for the 22 23 80 This sentence is grammatical if uttered in particular pragmatic contexts in which the adverb can be interpreted as a real Topic. Rizzi (1999b) reports the following convers ational context: A: Pare che Gianni abbia risolto rapidamente il primo problema B: Mi sembra impossibile: rapidamente, ha probabilmente risolto IL SECONDO As noted earlier, the Top° head is [-Q], [-Mod] and [-Arg], therefore the lack of RM effects in B is expected. Rizzi (1999b), proposes, in a way consistent with other recent works ( see Benincà & Poletto (2000) among others), that the fine structure of the Left Periphery is more fine grained than it was assumed in Rizzi (1997). In particular, examples like (44), involving a non-topicalized adverb in the CP space, are taken to be a clear indication of the existence of a (recursive) Mod projection. Therefore, the CP structure proposed by Rizzi (1999b) is as in (i): (i) Force Top* Focus Mod* Top* Fin “L-Tous”, Restructuring and Quantifier Climbing movement of tout, which targets a [+Mod] position. On the contrary, since it is specified as both [+Q] and [+Mod], Negation blocks Quantifier Climbing. Hence, we can conclude that Quantifier Climbing targets a [+Mod] position. 3.2. Quantifier Climbing: Restructuring and CP reduction What has still to be assessed is the relation between Restructuring and Quantifier Climbing. I adopt the basic guidelines of Roberts (1997) analysis of Restructuring, in particular the proposal that Rizzi‟s (1982) idea of Restructuring as clause union can be implemented in more modern terms making reference to a process of incorporation of the embedded verb onto the restructuring verb, a process that yields a unitary extended projection. Turkish data overtly support the hypothesis that an incorporation process underlies Restructuring; in fact in this language “the modal and aspectual predicates that trigger restructuring are realized as affixes” (Roberts (1997)), as (46) shows: (46) a. O Adam el aç-iyordu The man hand open-PROG “The man is begging” b. Yika-u-ma-mali-yim Wash-REFL-NEG-NEC-1SG “I shouldn‟t wash myself” However, Romance Restructuring clearly does not rely on a process of morphological incorporation; in fact, the Restructuring verb and the embedded verb can surface under different heads, as the (grammatical) intervention of XPs such as Adverbs, FQs and wh- words between the two shows: (47) a. Questi libri si volevano proprio leggere b. Gianni li vuole tutti leggere c. ?Certe risposte non si sanno mai come dare In order to make the data in (47) consistent with the idea that an incorporation process underlies Restructuring, Roberts (1997) proposes (48): (48) a. Head movement is copying b. *[X° W1 W2], where Wn are morphological words c. A head is spelled out in the highest position of its chain, subject to (48b) The morphological constraint in (48b) limits the “quantity” of morphological material that can be spelled out under a single head. Since two verbs are of course two distinct lexical items, their morphological incorporation would violate (48b), therefore, according to Roberts‟ proposal, they have to be spelled out under distinct 81 Marco Nicolis heads (48c). Roberts concludes that Restructuring in Romance can be considered a sort of covert incorporation, which cannot surface morphologically, due to (48b). This idea along with the proposal that RM is suspended if the intervener and the landing site both belong to the same extended projection (see fn. 13) provide an explanatory tool for the emergence of the typical phenomena associated with 24 Restructuring, namely Clitic Climbing, Aux change, Long NP movement . Consider for example (49), a case of Long NP movement: (49) [AgrS1 [Queste case]i si vogliono [AgrS2 PRO venderev [AgrOP ti Agr° [VP tv ti a caro prezzo]]]] The embedded verb vendere is “virtually incorporated” onto the Restructuring verb volere, but they are spelled out under different heads, due to (45b). The incorporation process yields a single extended projection: therefore, the movement of the DP queste case to [Spec, AgrS1] does not violate RM, since PRO in [Spec, Agrs2] does not qualify as a potential intervener, under the assumption that RM does not hold if landing site and potential intervener are in the same extended projection. So, Roberts‟ account of Restructuring relates the availability of Restructuring to long Verb movement, an empirically correct conclusion. Hence, so far as French is concerned, it does not allow Restructuring, since Infinitival Verbs do not move long enough to (virtually) incorporate onto the Restructuring verb, therefore in cases comparable to (49), PRO would count as an intervener. As Roberts (1997) himself observes, his analysis does not easily extend to Quantifier Climbing. In fact, if anything, French should be more “opaque” than Italian, given the shortness of [-fin] Verb movement. The necessary conclusion seems to be that Roberts‟ analysis must be somehow integrated in order to account for Quantifier Climbing. I would like to propose that Restructuring verbs, alongside their ability to trigger the virtual incorporation of the embedded verb, also select a reduced CP. I take this (these?) missing layer(s) to be responsible for the ungrammatical climbing of tout with non Restructuring verbs in (33). This is just a speculation, though, since I have no strong independent evidence for such a claim. However, it is well known that Restructuring verbs and their counterpart in V2 languages allow for clause bound movements to happen in biclausal structures. Consider (50), from Giusti (1993): (50) a. weil die berühmte Friedlandia [CP dieses Lied in Wien zu singen] versuchte because the famous Friedlandia this Lied in Wien to sing tried b. weil [dieses Lied] i die berühmte Friedlandia [CP ti in Wien zu singen] 24 82 See Rizzi (1982) for the original account of these phenomena and Burzio (1986) for the non full coextensivity of the three phenomena. “L-Tous”, Restructuring and Quantifier Climbing (51) a. weil die berühmte Friedlandia [ CP dieses Lied in Wien zu singen] verlangte because the famous Friedlandia this Lied in Wien to sing b. *weil [dieses Lied]i die berühmte Friedlandia [CP ti in Wien zu singen] versuchte pretended With a particular class of verbs, which corresponds by and large to Restructuring verbs, Scrambling, a typically clause bound movement, can cross a CP projection ((50) vs. (51)). This is of course no compelling evidence for the hypothesis that Restructuring verbs indeed select a reduced CP. However, the contrast (50) vs. (51) strongly resembles the one in (32) vs. (33) above. It seems to me fair to conclude that whatever lexical property characterises Restructuring verbs (be it selecting a reduced CP or not), allowing for (50), the sam property should also be operative in French allowing (32). But if this conclusion is correct, how come Italian strongly disallows Quantifier Climbing? It is first of all important to notice that Quantifier Climbing is a marked option in French, that seems to satisfy no core UG requirement; in fact in all the cases in (32) the Quantifier can remain in between the main verb and the embedded verb; therefore it is plausible that an existing structural option is not exploited in Italian, given the peripheral status of the construction. Moreover, I have shown that Quantifier Climbing targets an Ā-m position, therefore it must be attractable by a [+Mod] head. If it was possible to show that French tout is in some sense “more adverbial” than its Italian counterpart tutto, the absence of Quantifier Climbing in Italian could be due the impossibility of attraction of tutto by a [+Mod] head. Some evidence in the desired direction seems to exist. In fact, French tout, but not Italian tutto, can modify some adverbial classes. (52) a. tout bêtement b. tout simplement c. tout doucement a‟. *tutto stupidamente b‟. *tutto semplicemente c‟. *tutto dolcemente Therefore, Italian tutto could not be attractable by a [+Mod] head. A last point is worth mentioning. Given the optional possibility of Quantifier Climbing in French, one may wonder whether the Quantifier moves directly from the embedded object position, or moves through the derived low position. An answer to this question relies on the precise individuation of the kind of structural deficiency the Quantifier has to recover for. If we adopt Cardinaletti & Starke (1994) hypothesis that the movement to the low position is mandatory for Case reasons, the natural conclusion would be that Quantifier Climbing is a two step movement, since otherwise the moved Quantifier would violate the Case Filter. 83 Marco Nicolis References Belletti, A. 1990. Generalized Verb Movement. Rosenberg & Sellier, Torino. Baltin, M. & Ch. Collins, eds. 2000 A Handbook of Syntactic Theory, Blackwell, Oxford. Burzio, L. 1986. Italian Syntax. Kluwer, Dordrecht. Cardinaletti, A. & M. Starke. 1994. “The typology of structural deficiency. On the Three Grammatical Classes”, University of Venice Working Papers in Linguistics, 4 (2).41-109. Chomsky, N. 1995. The Minimalist Program. MIT Press, Cambridge, Mass. Cinque, G. 1990. Types of A’ dependencies, MIT Press, Cambridge, Mass. Cinque, G. 1995a “Bare quantifiers, quantified NPs, and the notion of operator at S-structure” (ch. 3 in Cinque (1995b)). Cinque, G. 1995b “On leftward movement of tutto in Italian” (ch. 9 in Cinque (1995c)). Cinque, G. 1995c Italian Syntax and Universal Grammar, Cambridge University Press. Cinque, G. 1998. Adverbs and Functional Heads. Oxford University Press. Giusti, G. 1993. La sintassi dei determinanti. Unipress, Padova. Grimshaw, J. 1991. “Extended Projections”. Manoscritto, Brandeis University, Waltham, Mass. Kayne, R. 1975. French syntax. MIT Press, Cambridge, Mass. Kayne, R. 1984. Connectedness and binary branching. Dordrecht: Foris. Kayne, R. 1989. “Null subjects and clitic climbing”. In Jaeggli, O. & K. Safir (eds.), The Null Subject Parameter: 239-261. Kayne, R. 1994. The antisymmetry of syntax. MIT Press, Cambridge, Mass. Koster, J. 1978. Locality Principles in Syntax, Foris, Dordrecht. Obenauer, H. 1983 “On the Identification of Empty Categories”, The Linguistic Review, 4, 153-202. Pollock, J.-Y. 1989. “Verb Movement, Universal Grammar, and the structure of IP”, Linguistic Inquiry 20, 365-424. Rizzi, L. 1982. Issues in Italian Syntax, Foris, Dordrecht. Rizzi, L. 1990. Relativized Minimality. MIT Press, Cambridge, Mass. Rizzi, L. 1996. “Residual Verb Second and the Wh-Criterion”, in Belletti, A. & L. Rizzi (eds), Parameters and Functional Heads, Oxford University Press, 63-90. Rizzi, L. 1997. “The Fine Structure of Left Periphery”, in L. Haegeman (ed.), Elements of Grammar, Kluwer, Dordrecht. Rizzi, L. 1999. “Some Issues in the Theory of Locality”, talk given at the Workshop on the Cartography of Syntactic Positions and Semantic Types, Certosa di Pontignano, Nov. 1999. Rizzi, L. 2000. “Relativized Minimality Effects”, in Baltin & Collins, eds. 2000. Roberts, I. 1997. “Restructuring, Head Movement and Locality”, Linguistic Inquiry 28: 423-460. Shlonsky, U. 1991. “Quantifiers as functional heads: a study of quantifier float in Hebrew”, Lingua, 84.159-180. Sportiche, D. 1988. “A theory of floating quantifiers and its corollaries for constituent structure”. Linguistic Inquiry 19, 425-449. 84 BILINGUISMO E ACQUISIZIONE INFANTILE DI L2: ALCUNE OSSERVAZIONI SULL’ACQUISIZIONE SIMULTANEA DI ITALIANO E TEDESCO DA PARTE DI BAMBINI IN ETÀ PRESCOLARE* Manola SALUSTRI 1. Introduzione Ricerche recenti hanno mostrato che l’acquisizione del linguaggio nei bambini bilingui non presenta differenze rilevanti rispetto all’acquisizione nei bambini 1 monolingui . Negli studi svolti da J. Meisel nell’ambito del progetto D.U.F.D.E. su bambini bilingui francese-tedesco di età prescolare sono state infatti riscontrate le stesse fasi presenti nello sviluppo linguistico dei bambini monolingui francesi e tedeschi. In particolare non sono stati riscontrati errori nella struttura della frase, 2 anche se le due lingue acquisite presentano un diverso ordine di base. Questo lavoro riporta i risultati di uno studio longitudinale sull’acquisizione simultanea di italiano e tedesco da parte di tre bambini bilingui di circa tre anni, Renzo, Samuele e Marianna, residenti in Italia. Lo studio è basato sull’analisi quantitativa di un corpus di produzione spontanea raccolto nel corso di circa un anno e trascritto nel formato CHAT di CHILDES. In particolare verrà considerata la posizione del verbo nella struttura della frase tedesca. * 1 2 Desidero ringraziare Adriana Belletti, Maria Teresa Guasti e Luigi Rizzi per i preziosi suggerimenti e per aver pazientemente letto e commentato questo lavoro. Vorrei ringraziare inoltre Imke Kruse, Claudia Perlitius, Claudia Ruff e David Nolan per l’ aiuto nella raccolta ed interpretazione dei dati. Infine un ringraziamento particolare ai bambini, alle loro famiglie e alla Scuola Materna di Gaiole in Chianti per la disponibilità e la collaborazione. Cfr. Meisel (1986/1990), Meisel & Müller (1992). Nel corso del progetto DUFDE (Deutsch und Franzosisch, Doppelter Erstspracherwerb) diretto dal 1986 al 1990 da J. Meisel all’Università di Amburgo è stata studiata longitudinalmente l’acquisizione simultanea di tedesco e francese in bambini di età prescolare. Cfr. Meisel (1994). 85 Manola Salustri Indicherò con il termine “bilinguismo” solo l’acquisizione simultanea di due lingue dalla nascita (Bilingual First Language Acquisition) o dai primi mesi di vita ( 3 Bilingual Second Language Acquisition) . Nel caso in cui i bambini siano venuti in contatto con una seconda lingua solo dopo i due anni di età, è infatti più opportuno il 4 termine di “acquisizione infantile di L2”. Dalla ricerca emerge chiaramente che i bambini hanno acquisito le regole grammaticali della lingua tedesca, il verbo viene posto infatti correttamente nella seconda posizione nelle frasi principali e nell’ultima nelle frasi subordinate. La presenza di frasi principali V1, cioè con verbo flesso in prima posizione, può essere analizzata come il risultato dell’omissione del primo elemento della frase dovuta a due distinti fenomeni: la Fase del Soggetto Nullo Infantile ed il Topic-drop. Quest’ultimo fenomeno, che caratterizza il tedesco parlato, viene ipergeneralizzato da uno dei bambini, Renzo, anche agli avverbi. Nel corpus di questo bambino sono presenti, seppur in percentuale molto limitata, anche frasi a V1 non analizzabili come il risultato dei suddetti fenomeni, le quali sembrano indicare una vera e propria fase nello sviluppo linguistico. Dai risultati emerge la tendenza riscontrata anche nell’input e sempre più frequente nel tedesco contemporaneo, a posizionare il verbo flesso in seconda posizione nelle subordinate introdotte da “weil”, come già notato da Schönenberg (1998) e Hamann (1997). Quest’uso non viene quindi considerato un’interferenza con l’italiano, ma è dovuto ad un influsso dell’input, infatti nella lingua parlata “weil” permette il Verb-Second (V2). Verranno riportati infine alcuni dati riguardanti l’acquisizione dell’italiano come L2 da parte di Angela, una bambina di madrelingua tedesca di tre anni residente in Italia, i quali mostrano differenze evidenti rispetto ai corpus dei bambini italiani nelle stesse fasi dello sviluppo linguistico (cfr. Guasti (1993/1994)). Pur essendo limitati, questi dati sembrano mostrare che l’acquisizione infantile di L2 segue, almeno inizialmente, processi diversi rispetto all’acquisizione nei bambini bilingui e monolingui. L’articolo è organizzato come segue: dopo aver mostrato le diverse posizioni in cui può trovarsi il verbo nella frase tedesca, considerando anche il fenomeno del 3 4 86 Per distinguere questi due casi di bilinguismo (BFLA e BSLA), spesso difficili da identificare, vengono usate a volte terminologie diverse. Bohnacker (1998) ad esempio, usa il termine “successive bilingualism” per indicare un tipo di acquisizione linguistica in cui l’esposizione alle due lingue ha avuto luogo entro il primo anno di vita ma non fin dalla nascita: “Successive bilingualism takes place when children are exposed to a second language not from birth but later, tough well before they have mastered the essentials of their first language. […] ... successive bilingualism in childhood is sometimes not consider “proper” bilingualism or true second language acquisition.”. Cfr. De Houwer (1996). Bilinguismo e acquisizione infantile di L2 Topic-drop, verranno analizzati alcuni aspetti della grammatica infantile. Verrà quindi presentato il corpus raccolto, mentre la sezione 7 avrà come oggetto i risultati dell’analisi quantitativa. Nell’ultima parte si discuteranno i risultati ottenuti alla luce della teoria linguistica adottata. 2. La struttura della frase tedesca 2.1. Asimmetria principale-subordinata: il fenomeno del Verb-Second Nella lingua tedesca gli elementi nominali hanno una distribuzione piuttosto libera, dato che, nonostante la presenza di un ordine stabilito nella frase, sono presenti posizioni cosiddette di scrambling, cioè posizioni dove il costituente 5 nominale può trovarsi liberamente. La posizione del verbo è invece molto più rigida: questo si trova necessariamente nella seconda posizione nelle frasi principali come in (1)-(2) (Verb- second) mentre nelle subordinate, come in (3)-(4), il verbo 6 occupa l’ultima posizione (Verb-final) . (1) Der Professor las ein Buch. Il professore leggeva un libro S V O (2) Ein Buch las der Professor. Un libro leggeva il professore O V S (3) , daß der Professor ein Buch las. , che il professore un libro leggeva S O V (4) , ob der Professor ein Buch las. , se il professore un libro leggeva. S O V 5 6 Cfr. Cardinaletti & Giusti (1996). Questa asimmetria tra la frase principale e la frase subordinata ha causato una notevole incertezza ai fini di un inquadramento tipologico della lingua tedesca, caratterizzata da un possibile ordine SVO nelle principali e da un ordine SOV nelle subordinate. La prima e meglio conosciuta analisi in ambito generativo del fenomeno del V2 (v. Thiersch (1978) e Den Besten (1983)) considera il tedesco come una lingua SOV e spiega l’ordine della principale come il risultato di un movimento testa a testa del verbo flesso da INFL a C° (cfr. tra gli altri, la discussione in Tomaselli (1990), v. paragrafo 3.2.1). 87 Manola Salustri Nel caso in cui la prima posizione nelle frasi principali sia occupata dall’oggetto o da un altro costituente, come un avverbio, il soggetto segue immediatamente il verbo 7 flesso nell’ordine frasale, come in (1), fenomeno sintattico designato come Inversione. Questo schema è generalmente rispettato in tedesco, ma esistono contesti in cui il verbo può trovarsi in posizioni diverse: questo può infatti occupare la prima e la seconda posizione nelle frasi principali, la prima, la seconda e l’ultima nelle subordinate. Le frasi principali a Verb-First sono costituite dalle Interrogative SI/NO, dalle imperative e dalle frasi caratterizzate da Topic-drop. (v. par. successivo). Le frasi subordinate invece, possono avere il verbo in prima posizione nelle frasi ipotetiche 8 non introdotte da un complementatore. Il V2 nelle subordinate è invece possibile con i verbi cosiddetti “bridge” (come sagen, meinen, glauben, denken, wissen) con C nullo, e nelle subordinate introdotte da “weil” nel registro parlato. 2.2. Il fenomeno del Topic-drop in tedesco Nelle frasi principali in (5) e (6) il soggetto e l’oggetto, rispettivamente, sono omessi. Tale processo, che è possibile nella lingua parlata quando il soggetto e l’oggetto sono noti e recuperabili dal contesto, viene chiamato Topic-drop. Non è possibile omettere altri tipi di complementi, né marcati con il caso dativo o genitivo, 7 8 88 Tutte le lingue germaniche, tranne l’inglese moderno, presentano il fenomeno del V2, così come le lingue romanze medievali. Studi recenti hanno mostrato però, che non tutte le lingue a V2 presentano un’asimmetria principale-subordinata. L’Islandese e l’Yiddish, ad esempio, presentano questo fenomeno, ma con caratteristiche parzialmente diverse rispetto al tedesco. In queste lingue infatti il V2 è caratterizzato dalla restrizione alla seconda posizione del verbo flesso e dall’inversione del soggetto, qualora la prima posizione sia occupata da un altro costituente, non è presente però l’asimmetria tra frasi principali e frasi subordinate, il V2 in queste lingue è un fenomeno generale, come si nota nei seguenti esempi: (i) Jonas tsveyfelt az morgen vet Miriam fri oyfshteyn. (Yiddish) Fr. Sub. John doubts (on) that-tomorrow will M. early stand up. (ii) Jón efast um að á morgun fari María snemma á fætur. (Islandese) Fr. Sub. John doubts (on) that-tomorrow will M. early stand up. (per una discussione più dettagliata cfr. Vikner (1994)). L’Yiddish, in particolare, non ha sempre permesso il V2 nelle frasi subordinate come accade adesso, troviamo infatti la stessa asimmetria principale/subordinata in alcuni testi dell’Yiddish medievale. (cfr. Santorini (1995)). Questa lingua presenta quindi il processo inverso rispetto all’inglese, il quale pur permettendo il V2 nel periodo medievale, ha perso questo fenomeno in seguito, conservandone solo alcune tracce (cfr. Belletti e Rizzi (1996)). Questo tipo di costruzione si ha anche in italiano in frasi come la seguente: (i) Avessi il tempo, lo farei. Bilinguismo e acquisizione infantile di L2 né introdotti da una preposizione e, come notato da Huang (1984), in una frase può essere omesso un solo argomento . Il Topic-drop è un fenomeno “radice”, non si può avere nelle subordinate, in contesti post-verbali o post-wh. Gli esempi seguenti riassumono questo aspetto della distribuzione: (5) __ist gekommen (lui) è arrivato. omissione del soggetto (6) __hat er gekauft. (esso) ha lui comprato. omissione dell’oggetto (7) *, dass__gegessen hat , che (lui) mangiato ha *omissione in una subordinata (8) *Was hat__gegessen? cosa ha (lui) mangiato *omissione in contesto post-verbale (9) A: Braucht er Hilfe? Ha bisogno lui di aiuto B: __habe ich schon geholfen (lui +dat.) ho io già aiutato. *omissione di un complemento marcato con il caso genitivo (ihm) (10) A: Ist Hans gekommen? è Hans arrivato? B: __habe ich lange gewartet (a lui +acc.) ho io a lungo aspettato *omissione di un complemento introdotto da una preposizione (auf ihn) Seguendo Rizzi (1994), il Topic-drop è possibile solo se il costituente è omesso dalla prima posizione della frase, cioè dallo Spec/CP, una chiara manifestazione di quello che Rizzi definisce “Privilege of the Root”. Gli espletivi, non costituendo un 9 elemento topicalizzato, cioè dato, non sottostanno a questo processo. Si assume che le frasi a Topic-drop coinvolgano il movimento di un operatore nullo (OP), generato nella posizione soggetto o nella posizione oggetto, dalla posizione di base allo Spec/CP. Tale movimento lascia come traccia una variabile legata dall’operatore. Quest’ultimo verrà identificato dal discorso tramite un elemento topic introdotto da una frase precedente. Consideriamo ad esempio la rappresentazione in (11). (11) [CP OP i kommt [IP ti morgen]] arriva domani 9 omissione del soggetto in tedesco L’omissione di un espletivo dalla prima posizione della frase, non accettabile in tedesco, è invece possibile nello svedese parlato in frasi come la seguente (cfr. Rizzi (2000)): (i) (det) verkar som om… (esso) sembra come se.... 89 Manola Salustri Anche se apparentemente il verbo finito in queste costruzioni sembra occupare la prima posizione, si assume che nelle frasi a Topic-drop il verbo sia preceduto da un operatore nullo. Nelle lingue come il cinese ed il tedesco, quindi, la possibilità di omettere soggetti ed oggetti è dovuta alla disponibilità di un operatore nullo legato al discorso. Consideriamo adesso la categoria vuota lasciata dal movimento dell’operatore nella posizione di base. Essa non può essere pro, altrimenti avrebbe la stessa distribuzione che si trova nelle frasi a soggetto nullo ad esempio in italiano, non può essere PRO perché questa categoria, identificata dai tratti [+ana; +pro] si trova solo in contesti non flessi, in particolare come soggetto non esplicito delle infinitive e non può essere una traccia di NP perché si trova in una posizione a cui viene dato caso. Rizzi propone che la categoria vuota nei contesti di Topic-drop sia una costante 10 nulla , identificata dai tratti [-a, -p, -v] e legata all’operatore nullo, che si trova sempre in prima posizione ed è legato al discorso. Questo tipo di categoria si ha in frasi come in (12)-(14). (12) This book [OP I really like NC] Topicalizzazione in inglese (13) [OP habe [NC es schon gesehen]] ho ciò già visto Topic-drop in tedesco (14) [OP[NC kanjian ta le]] ha visto lui-ASP Topic-drop in cinese Il parametro responsabile delle frasi a soggetto ed oggetto nullo nel tedesco ed 11 anche nel cinese, è il parametro del Topic-drop (vedi Huang (1984)): . Parametro del Topic-drop È disponibile un operatore legato al discorso? (si/no) Nelle lingue come l’italiano e lo spagnolo, cosiddette a pro-drop, i soggetti nulli vengono invece interpretati come pro, elementi pronominali non realizzati foneticamente i cui tratti di accordo, seguendo Rizzi (1986), possono essere recuperati grazie ai tratti di accordo presenti in I. 10 11 90 Lasnik e Stowell (1991) hanno proposto di differenziare le categorie vuote [-ana; +pro] in vere variabili ed epiteti nulli, osservando che certe costruzioni ad operatore nullo hanno un’interpretazione ed un’analisi diversa dalle costruzioni ad operatore quantificazionale e variabile in riferimento al fenomeno di Weak crossover. Il Topic drop in tedesco presenta delle differenze rispetto al Topic-drop in cinese. In quest’ultima lingua infatti l’omissione del soggetto e dell’oggetto non è limitata alla prima posizione della frase ma può trovarsi anche in frasi subordinate ed interrogative wh-. Tuttavia la possibilità di omettere soggetti ed oggetti viene riportata in entrambe le lingue alla disponibilità di un operatore nullo legato al discorso (cfr. Guasti, in stampa). Bilinguismo e acquisizione infantile di L2 La distribuzione dei soggetti nulli nelle lingue è quindi governata dai seguenti parametri: (i) il parametro del pro-drop. (ii) il parametro del Topic-drop. (italiano, spagnolo, catalano) (cinese, coreano, tedesco) Come vedremo nel paragrafo 3., i soggetti nulli riscontrati nella grammatica infantile presentano delle caratteristiche diverse rispetto a quelli presenti nella grammatica adulta delle lingue a soggetto nullo e necessitano di una diversa interpretazione. 3. Alcune osservazioni sulla grammatica infantile Numerosi studi hanno confermato che fino all’età di circa tre anni, i bambini hanno l’opzione di omettere il soggetto anche se la loro lingua target non è una lingua a soggetto nullo. Gli esempi in (15)-(17) mostrano questo aspetto della grammatica infantile in tedesco (15), francese (16) e danese (17). (15) __bin wieder lieb am again good (I) am good again (Elisa 2; 10) (16) __a tout tout tout mangé has all all all eaten (he) has eaten all (Augustin 2; 0) (17) __ikke køre traktor (Jens, 2; 0) not drive traktor (I, you, etc.don’t drive the tractor) (da Hamann (1997)) Negli anni ottanta la fase del Soggetto nullo infantile era stata interpretata come il risultato di una scelta parametrica sbagliata da parte del bambino: seguendo Hyams (1986), il parametro del pro-drop è fissato inizialmente al valore positivo e solo se il bambino viene a contatto con una lingua non a pro-drop il parametro viene “rifissato” al valore negativo. Tuttavia studi più approfonditi hanno mostrato delle chiare discontinuità tra questo fenomeno e l’omissione del soggetto nelle lingue a pro-drop. Mentre in queste ultime l’omissione del soggetto è possibile anche in frasi subordinate e nelle interrogative wh-, come in (18) e (19), i soggetti nulli non sono presenti in questi contesti strutturali nei bambini che acquisiscono lingue non a pro-drop. (18) Gianni canta quando pro è contento. (19) Chi hai pro visto? Come mostrano le frasi in (20)-(21) l’omissione del soggetto in questi contesti è invece presente nei bambini che acquisiscono l’italiano. Sembra quindi che il parametro del pro-drop venga fissato molto presto. 91 Manola Salustri (20) Ov’è? (Martina,1;8) (21) Pecchè piangi? (Martina, 2;3) (da Cipriani et al. (1993)) Seguendo Rizzi (1994), i bambini hanno l’opzione di omettere il soggetto solo se esso si trova nello Specificatore della Radice, il fenomeno del Soggetto nullo infantile è infatti limitato alla prima posizione assoluta nelle frasi con verbo flesso. La presenza di soggetti nulli in posizioni non iniziali, come in (22), è limitata alla frasi caratterizzate da un altro fenomeno della grammatica infantile, i Root 12 Infinitives, o infiniti principali, la differenza cruciale è che tale presenza è limitata 13 a contesti non flessi (v. par. 3.1.). (22) Where ____go(ing) La restrizione del fenomeno alla prima posizione assoluta nelle frasi con verbo flesso ha portato a trovare degli elementi di continuità tra questo ed altri “fenomeniradice” presenti nella lingua adulta, come il Topic-drop (cfr. tra gli altri Bromberg e Wexler (1995)), ed il Diary-drop, osservato in alcuni registri particolari (per una discussione più dettagliata di questo tipo di omissione cfr. Haegeman (2000)). Come si è visto nel paragrafo precedente, anche nelle lingue germaniche caratterizzate da Topic-drop, i soggetti nulli non possono trovarsi in posizione postverbale o post-wh, lasciando supporre che questo fenomeno e la fase del Soggetto nullo infantile siano interpretabili con un’analisi simile, mettendo in gioco una categoria vuota diversa da pro, specificamente una costante nulla legata ad un operatore nullo (vedi par. 2.2.), come in (23). (23) [CP OPi [IP ti wan(t) do (a)gain] (Sarah, 2;7, da Guasti, in stampa) Come notato da Hyams e Wexler (1993) è tuttavia presente una significativa differenza tra la fase del Soggetto nullo infantile ed il fenomeno del Topic-drop. Dato che il fenomeno del Topic-drop permette anche l’omissione di oggetti, dovremmo riscontrare nella grammatica infantile dell’inglese anche la presenza di oggetti nulli. Gli studi di Hyams e Wexler mostrano invece una notevole asimmetria tra l’omissione di soggetti ed oggetti nei bambini inglesi, mentre nei bambini cinesi questa non è riscontrata (cfr. Wang, Lillo Martin, Bast e Levitt (1992)). Questa osservazione lascia supporre che il fenomeno del soggetto nullo in inglese non sia interpretabile con l’analisi in (23). 12 13 92 Cfr. Wexler (1992), cfr. anche Rizzi (1993/94). Dal momento che un ambiente non flesso può licenziare un soggetto non realizzato foneticamente, la presenza dell’uso di forme non flesse nelle grammatiche infantili potrebbe essere la diretta causa della non realizzazione del soggetto in questi casi (cfr. Rizzi (2000)). Bilinguismo e acquisizione infantile di L2 In particolare, verificando la curva estensiva dell’omissione del soggetto nelle lingue germaniche a V-2 si osservano delle differenze, indicate di seguito, tra questo fenomeno nella grammatica infantile ed il fenomeno del Topic-drop nella lingua adulta (le osservazioni si riferiscono a studi di Haegeman (1995) sull’olandese): (i) Mentre la grammatica infantile permette l’omissione degli espletivi, la 14 grammatica adulta non presenta questa opzione. (ii) In tutti i bambini esaminati c’è una fase in cui gli infiniti principali ed i soggetti nulli decrescono parallelamente, il che lascia supporre una correlazione tra i due fenomeni. Basandosi su queste osservazioni Haegeman (1995) sostiene che i soggetti nulli presenti nelle grammatiche infantili di lingue come olandese e tedesco, caratterizzate dal Topic-drop, debbano essere considerate non come il risultato di questo fenomeno, bensì come il risultato di troncamenti nella struttura, responsabili anche della presenza di Root Infinitives, così come proposto da Rizzi (1994) per i bambini francesi ed inglesi (vedi paragrafo successivo). Nel paragrafo seguente verrà presentata la teoria del Troncamento di Rizzi (1994). Questa teoria, tenendo conto delle osservazioni di Haegeman (1995), verrà generalizzata anche alle lingue a V2, come il tedesco. Seguendo Rizzi (1997) verrà adottata una struttura del CP più articolata in modo da mantenere un’analisi simmetrica del fenomeno del V2, pur ipotizzando la possibilità di troncamenti. Si assumerà infatti che in queste lingue il troncamento ha luogo ad un livello più alto della struttura rispetto a lingue come il francese o l’inglese, in modo da mantenere una struttura sufficiente all’analisi del V2. (vedi par. 3.2.) È plausibile pensare, tuttavia, che perlomeno una parte dei soggetti nulli infantili in queste lingue sia dovuta al fenomeno del Topic-drop, come confermano i dati della ricerca, in cui non solo soggetti, ma anche oggetti vengono omessi dai bambini. Questo aspetto verrà discusso nel paragrafo 3.2.2. 3.1. L’ipotesi del Troncamento Seguendo Rizzi (1994) la grammatica infantile, pur possedendo la stessa struttura frasale della grammatica target, quindi l’intera gamma delle categorie funzionali, non comprende vincoli per quanto riguarda la radice delle frasi dichiarative. Seguendo questa ipotesi, il principio in (24), secondo il quale il sistema 14 L’omissione di espletivi nella prima posizione nella lingua parlata è comunque accettabile in altre lingue a Topic-drop come lo svedese (vedi nota 9) non sarebbe irragionevole pensare, quindi, che i bambini ipergeneralizzino una regola della grammatica adulta. 93 Manola Salustri del complementatore viene sempre proiettato nella struttura della grammatica adulta, 15 viene acquisito dai bambini solo verso i tre anni di età. (24) The Root Principle (Rizzi 1994) CP = Root Ne deriva che i bambini possono proiettare delle strutture “troncate”, ad esempio non comprendenti CP all’inizio del loro sviluppo linguistico. Il troncamento in un punto della struttura implica che tutte le categorie al di sotto di tale punto siano incluse nella rappresentazione, mentre le categorie al di sopra ne siano escluse. Se la struttura viene troncata ad Agr, ad esempio, come in (25), viene proiettato solo TP, di modo che l’accordo non risulterà specificato. (25) CP C’ spec C Agr P Agr’ spec neg P Agr neg’ spec neg TP spec T’ T VP spec V’ V 15 94 Cfr. anche Rizzi (2000). Bilinguismo e acquisizione infantile di L2 Seguendo Rizzi l’opzione di troncamento nella struttura è responsabile della presenza di Root Infinitives (o infiniti principali) e Soggetti nulli nella grammatica dei bambini. Come si è visto sopra, l’omissione del soggetto nella grammatica infantile di lingue non a pro-drop è stata riscontrata prevalentemente in due casi: (i) con infiniti non flessi. (ii) nella prima posizione assoluta. Seguendo Rizzi, possiamo assumere che il soggetto nullo in contesti non flessi sia PRO, presente anche nella grammatica adulta, per cui vale la seguente restrizione di occorrenza: PRO è legittimato in contesti non flessi. Quest’ultimo sarebbe legittimato dalla presenza dei cosiddetti Root Infinitives (RI), presenti nella grammatica infantile e dovuti, seguendo Rizzi (1998), alla presenza di strutture troncate in cui Agr non è specificato. Il soggetto nullo legittimato nello specificatore della radice, sempre seguendo Rizzi, sarebbe invece una costante nulla. Ipotizzando che i bambini di due-tre anni abbiano l’opzione di “troncare” la struttura a diversi livelli, ad esempio IP, lo Spec/IP diverrebbe lo Spec. della radice. Quest’ultimo, assumendo la riformulazione dell’ECP in (26), proposta da Rizzi (1994) risulta libero dal principio di identificazione, non essendo c-comandato da nessun’altra posizione. (26) ECP. Una categoria vuota non pronominale deve essere identificata da un antecedente che la c-comanda solo se può esserlo. Una costante nulla (NC) generata nello specificatore della radice non ha bisogno di essere identificata e ciò permette un’analisi del soggetto nullo infantile come in (27). (27) [NC mange du pain] eat-3SG some bread (Grégoire, 2;1, da Guasti, in stampa) 3.1.1. Osservazioni sul fenomeno del Troncamento ed il fenomeno del Topic-drop Nel paragrafo 3. abbiamo visto che il fenomeno del soggetto nullo infantile non può essere interamente ridotto ad un’interpretazione come in (23). in quanto sono presenti degli elementi di discontinuità tra questo fenomeno ed il Topic-drop che caratterizza la grammatica adulta di lingue come l’olandese o il tedesco (vedi Haegeman (1995)). Seguendo Rizzi (1994/2000) abbiamo quindi interpretato i 95 Manola Salustri soggetti nulli infantili come costanti nulle generate nello specificatore della radice, dovute all’opzione di troncamenti della struttura. Contrariamente all’ipotesi di Haegeman (1995) (v. par. 3), sembra però plausibile pensare che nelle lingue caratterizzate dal Topic-drop come il tedesco e l’olandese, la presenza di soggetti nulli nella grammatica infantile sia dovuta almeno in parte alla presenza di questo fenomeno. I dati relativi al cinese ci mostrano che i bambini che imparano questa lingua hanno fissato il parametro già molto presto (v. Wang et al. (1992)) sarebbe quindi improbabile che nei bambini tedeschi ed olandesi non avvenga lo stesso. I dati della ricerca relativi al corpus di Renzo, inoltre, (v. pag. 26) mostrano che sono presenti anche omissioni di oggetti. Questo lascia supporre che il bambino abbia fissato il parametro del Topic-drop. Nei bambini che acquisiscono l’inglese, che non presenta questo fenomeno, questo tipo di omissione non è infatti riscontrata. Hamann (1997), analizzando il corpus di due bambini tedeschi di circa tre anni, ha notato una diminuzione dei soggetti nulli in prima posizione dal 40% al 5% nel corso del loro sviluppo linguistico. Questa diminuzione, come quella di Renzo, indica che il bambino ha perso un’opzione, quella di troncare la struttura, mentre è ancora presente un meccanismo di legittimazione dei soggetti nulli identificabile proprio nel fenomeno del Topic-drop che resterà operativo anche nella lingua adulta. 3.2. L’acquisizione della posizione del verbo L’acquisizione della posizione del verbo è stata il centro di interesse di molte ricerche sull’acquisizione del linguaggio. Il tedesco, in particolare, come si è visto sopra, presenta un ordine verbale piuttosto complesso (SVO nelle principali con soggetto in prima posizione, SOV nelle subordinate), ma dalle ricerche emerge che i 16 bambini non commettono errori nel posizionare il verbo. Nelle principali viene rispettato il parametro del V2, e non appena i bambini iniziano ad usare le subordinate il verbo viene subito posto in ultima posizione (Rothweiler (1993)). Seguendo Clahsen (1982), superata la fase di una sola parola, gli ordini verbali prevalenti sono V2 e Verb-final e non si hanno casi in cui il verbo appare prima del soggetto nelle frasi principali dichiarative (VS). Gli ordini dominanti nelle frasi principali, con differenze a seconda del bambino, sono quindi SVO e SOV. Come notato da Poeppel e Wexler (1993) il verbo si trova correttamente nella seconda posizione se flesso, mentre i verbi all’infinito o senza flessione sono collocati in posizione finale (si veda anche Clahsen (1990)). 16 96 Cfr. tra gli altri Meisel (1992), cfr. anche Poeppel & Wexler (1993). Bilinguismo e acquisizione infantile di L2 (28) ich hab ein dossen Ball. I have a big ball Verbo flesso (29) du das haben. you that have Verbo non flesso (da Poeppel eWexler (1993)) Gli oggetti e gli avverbi vengono topicalizzati, cioè posti all’inizio della frase già molto presto. Inizialmente il soggetto è spesso omesso in prima posizione, raramente in posizione post-verbale e nelle frasi subordinate, e quasi mai nelle interrogative Wh. Per quanto riguarda i bambini bilingui, non sono state riscontrate differenze rilevanti rispetto ai bambini monolingui, tranne una minore variabilità nell’ordine 17 dei costituenti frasali all’inizio della fase delle due parole. 3.2.1. Osservazioni sull’ipotesi del Troncamento d il fenomeno del Verb-second Seguendo Poeppel e Wexler (1993), come si è visto sopra, già dalle prime fasi dell’acquisizione del tedesco i bambini non commettono errori per quanto riguarda il parametro del V2: il verbo flesso si trova cioè correttamente in seconda posizione nelle frasi principali, mentre il soggetto od un altro costituente frasale occupa la prima posizione. 18 Seguendo l’analisi tradizionale, detta simmetrica , il fenomeno del V2 può essere spiegato sostenendo che il tedesco ha un ordine di base SOV e che l’ordine presente nelle frasi principali è dovuto all’interazione di due movimenti sintattici: (i) un movimento sintattico del verbo flesso dalla posizione di base a C° (VIC°). (ii) un movimento sintattico di un primo costituente della frase dalla posizione di base a Spec/C. 17 18 Basandosi sui dati di tre bambini bilingui francese-tedesco, Meisel e Müller hanno riscontrato alcuni errori nelle frasi subordinate. Uno dei bambini, Ivar, utilizza lo stesso ordine presente nelle frasi principali, con il verbo flesso che segue direttamente il primo costituente della frase. Gli altri due bambini bilingui usano lo schema del verb-final non appena iniziano ad usare frasi subordinate e non commettono errori. Müller sostiene che anche i bambini che acquisiscono tedesco e inglese (v. Leopold (1949)) e italiano e tedesco (v. Taeschner (1983)) commettono occasionalmente errori nelle subordinate, posizionando il verbo nella prima, nella seconda o nella terza posizione. Durante questo periodo i bambini usano allo stesso tempo anche lo schema del verb-final nelle subordinate. (cfr. Meisel e Müller (1992)). L’analisi simmetrica tratta uniformemente le frasi principali a soggetto iniziale (SVO) e le frasi principali non a soggetto iniziale (XVS), mentre nell’analisi asimmetrica, proposta da Travis e ripresa da Zwart nel 1990 le prime vengono considerate con una struttura più piccola rispetto alle ultime con l’ordine V2 espresso a livello dell’IP. Per una discussione più approfondita del fenomeno del V2 vedi tra gli altri i lavori raccolti in Belletti & Rizzi (1996). 97 Manola Salustri La periferia sinistra della frase si rende quindi necessaria come posizione di arrivo del verbo flesso (C°) e del primo costituente frasale (Spec/C). Come abbiamo mostrato nel paragrafo 3.1. la presenza di RI e di Soggetti nulli nella grammatica infantile viene interpretata, seguendo Rizzi (1994/2000), con l’opzione di troncamenti nella struttura. D’altra parte, ipotizzando un troncamento ad Agr, questo verrebbe ad eliminare la periferia sinistra della frase, verrebbe dunque a mancare una struttura sufficiente a spiegare il movimento del verbo flesso, non essendo più disponibile una posizione di arrivo per il primo costituente della frase (Spec/C) ed una per il verbo (C°), come in (30). (30) CP C’ spec C Agr P Agr’ spec neg P Agr neg’ spec neg TP spec T’ T VP spec V’ V Questo tipo di analisi porterebbe quindi a dover dare un’interpretazione non uniforme della conoscenza del fenomeno del V2, applicata in alcune frasi, quelle 98 Bilinguismo e acquisizione infantile di L2 cioè con soggetto realizzato fonologicamente, in cui la periferia sinistra della frase viene attivata, ma non in altre, quelle a soggetto nullo, costituite da strutture “troncate” in cui viene proiettato solo IP. Si porrebbe inoltre il problema di considerare IP a testa mediale nella grammatica infantile, in modo da rendere conto dell’ordine dei costituenti frasali (VO) e a testa finale nella grammatica adulta, ipotizzando un cambiamento ad un certo punto dello sviluppo linguistico. Se vogliamo mantenere l’ipotesi che anche nelle frasi a soggetto nullo i bambini applicano la conoscenza che sottende il V2, dobbiamo adottare un’analisi più complessa della periferia sinistra della frase. Questo ci permetterà di analizzare il fenomeno del V2 ipotizzando anche la possibilità di troncamenti nella struttura nelle prime fasi di acquisizione del linguaggio ad un certo punto dello 19 sviluppo linguistico. Una proposta che permette un’analisi soddisfacente di entrambi i fenomeni consiste nell’adottare la teoria di Rizzi (1997) riguardo allo split-CP. Seguendo Rizzi, il livello del CP dovrebbe essere suddiviso in un numero maggiore di 20 proiezioni funzionali . Basandosi su dati di italiano, inglese, francese, Rizzi propone la struttura in (31), in cui la testa funzionale indicata con Fin contiene le informazioni sulla finitezza della frase. La testa funzionale Focus ospita i costituenti focalizzati, ed i costituenti interrogativi, mentre la testa Topic ospita gli elementi topicalizzati. Topic, contrariamente a Focus è ricorsivo, come indicato da *. 19 20 Seguendo Haegeman (1995) la generalizzazione della proposta di Rizzi (1994) alle lingue a V2 implica una riconsiderazione della grammatica di queste ultime. Per mantenere un’analisi del V2 nella grammatica infantile ipotizzando la possibilità di troncamenti nella struttura, Haegeman propone di adottare un’analisi asimmetrica del V2 come proposta da Travis e ripresa da Zwart (1990). In questo caso, le frasi con soggetto iniziale verrebbero analizzate come IP (con I a testa mediale) mentre le frasi XPVS verrebbero analizzate come CP. Nelle frasi con soggetto iniziale quindi, la parte periferica della struttura non verrebbe attivata ed un’eventuale troncamento, responsabile della presenza di soggetti nulli infantili, permetterebbe tuttavia un’analisi del V2 all’interno di IP. Haegeman propone che la periferia sinistra venga invece attivata in entrambi i casi nella grammatica adulta, lasciando però questo aspetto come un problema aperto (per una discussione più dettagliata cfr. Haegeman (1995)). Cfr. Rizzi (1997). 99 Manola Salustri (31) Force P Force’ SPEC Force Top P* Top’ SPEC Top Foc P Foc’ SPEC Foc Top P* Top’ SPEC Top Fin P SPEC Fin’ Fin IP Con un sistema del CP più articolato, il fenomeno del Verb-Second non può ricevere un’interpretazione uniforme in cui il verbo sale a C° ed il primo costituente della frase sale allo Spec/C. Dobbiamo infatti considerare a quale testa, nel dominio di CP, salga il verbo, ed in quale Spec salga il primo costituente. Partendo da questo presupposto Haegeman (1996) ha ipotizzato una nuova analisi del fenomeno del V2, basandosi sull’olandese, proponendo la rappresentazione in (32). Per frasi del tipo SVO, il soggetto muove allo Spec/Fin ed il verbo flesso alla testa Fin, mentre i costituenti topicalizzati (oggetti, avverbi, ecc.) 21 si muoverebbero, seguendo Haegeman, a Spec/Top . 21 Cfr. Haegeman (1996). 100 Bilinguismo e acquisizione infantile di L2 (32) Force Force’ SPEC Force SPEC het Fin P Fin’ Fin regende IP de hele dag Contrariamente all’ipotesi tradizionale, quindi, in cui il costituente che precede il verbo flesso muove sempre a Spec/C, nel sistema split-CP il costituente può occupare, a seconda delle caratteristiche di quest’ultimo, diverse posizioni nel 22 dominio di CP. Quest’ultima analisi del fenomeno del Verb-second permette di esprimere certe proprietà interessanti, come il diverso comportamento di soggetti pronominali tonici ed atoni, (v. discussione in Tomaselli (1990)) presupponendo diversi “luoghi di arrivo” a seconda degli elementi considerati, all’interno del sistema del CP. Allo stesso tempo è possibile mantenere un’analisi unitaria del V2, in quanto il fenomeno risulta operativo sia nelle frasi SVO, nelle quali il soggetto si muove a Spec/FinP, che nelle frasi non a soggetto iniziale, in cui il primo costituente si muove a posizioni diverse a seconda delle sue proprietà. Presupponendo una struttura più articolata del CP possiamo ipotizzare che il troncamento avvenga ad un certo livello del sistema del complementatore. Tale 23 troncamento, responsabile della presenza di Soggetti nulli infantili , sarebbe tuttavia più alto della posizione di arrivo del verbo e dell’elemento topicalizzato, in modo da rendere disponibile una struttura sufficiente per l’analisi del V2. 22 23 Questa caratteristica avvicina l’ipotesi dello split CP all’analisi asimmetrica di Travis e Zwart, secondo la quale i due casi, SVO e XVS, vanno trattati in maniera distinta. Il troncamento responsabile della presenza dei Root Infinitives ha luogo in un punto più basso della struttura sotto la specificazione temporale. 101 Manola Salustri Seguendo questa ipotesi, il livello del troncamento sarà più alto nei bambini che acquisiscono una lingua a V2, rispetto ai bambini che acquisiscono una lingua in cui questo fenomeno non è presente, come il francese, in cui la posizione di arrivo fuori da IP per il verbo flesso e per il primo costituente della frase non si rende necessaria. Queste osservazioni sulla Teoria del Troncamento implicano necessariamente un maggiore approfondimento, legato alla studio della periferia sinistra della frase ed al ruolo dei processi maturativi nell’acquisizione del linguaggio. Gli aspetti della teoria discussi sopra rimangono quindi un problema aperto, destinato ad ulteriori ricerche. 4. Il Corpus Tre bambini bilingui italiano-tedesco di circa tre anni, residenti in Italia, Renzo, 24 Samuele e Marianna, sono stati registrati per un periodo di circa un anno . Le registrazioni di produzione spontanea, effettuate in un contesto di gioco, sono state trascritte nel formato CHAT e controllate da due parlanti di madrelingua tedesca ed una di madrelingua italiana (io stessa). Il corpus comprende inoltre tre registrazioni di Angela, una bambina di madrelingua tedesca residente in Italia che ha acquisito l’italiano solo dopo i tre anni e che quindi rappresenta un caso di acquisizione infantile di L2. Tav. 1. Corpus BAMBINO TIPO_DI_ACQUISIZIONE ETÀ ITA Samuele BFLA 3;9-4;0 Si Renzo BFLA 3;5-4;1 Si Marianna BFLA 3;8-4;2 Si Anna L2 3;0-3;1 Si * BFLA= Bilingual First Language Acquisition (v. Introduzione) L2= Acquisizione (infantile) di L2 TED ING Si Si Si Si (Si) No No No 4.1. Renzo Il corpus principale è costituito dalle registrazioni di Renzo, che è stato seguito per un periodo di circa nove mesi. Le registrazioni sono infatti iniziate nel Gennaio ’98, quando il bambino aveva 3;5 anni, e sono terminate nel Settembre ’98. Il bambino parla tedesco con la madre, con la quale trascorre la maggior parte della giornata, ed italiano con il padre. Renzo ha iniziato a frequentare l’asilo italiano all’età di 4;1 anni, ed ha una sorella Isa, di circa un anno e mezzo. Il tempo di 24 Le registrazioni sono state effettuate con un registratore SONYTMC-59V. 102 Bilinguismo e acquisizione infantile di L2 esposizione al tedesco è stato maggiore rispetto all’italiano, il padre parla comunque spesso al bambino in questa lingua. I genitori si sono rivolti a lui nelle due lingue fin dalla nascita, ed il bambino ha iniziato a formulare le prime frasi all’età di circa un anno e mezzo. Renzo è molto precoce ed il livello raggiunto sia in italiano che in tedesco è molto buono per la sua età. Le registrazioni sono state eseguite a casa del bambino, a Firenze, alla presenza della madre. Quando questa non era presente, un’altra persona di madrelingua tedesca o parlante L2 di tedesco si è rivolta a lui in questa lingua. In questo modo il bambino ha sempre avuto un referente per l’italiano (io stessa) ed uno per il tedesco. Tav. 2. Durata di esposizione durante il periodo delle registrazioni Renzo (BFLA) L. materna L. paterna L. ambiente Ted., (Ita.) Tutta la giornata. Ita. Alcune ore al giorno. Ita. Limitato (il bambino non frequenta ancora l’asilo). 4.2. Samuele Samuele ha iniziato a frequentare l’asilo italiano a circa tre anni, ma già prima parlava italiano con i figli dei vicini e gli amici dei genitori. La lingua di famiglia è il tedesco (la madre è di madrelingua tedesca) ma il padre, di madrelingua inglese, a volte cerca di rivolgersi al bambino in questa lingua, quando giocano insieme. Samuele, che ha un fratellino di circa un anno e mezzo, si esprime molto bene sia in italiano che in tedesco, mentre il vocabolario inglese è limitato, anche se il bambino riesce a comprendere questa lingua. Le registrazioni sono state effettuate all’asilo o a casa del bambino. In quest’ultimo caso Samuele ha sempre avuto un referente di madrelingua italiana (io stessa), uno di madrelingua inglese e uno di madrelingua tedesca (la madre). Il corpus è costituito prevalentemente dalle lingue che riguardano questa ricerca, italiano e tedesco. Tav. 3. Durata di esposizione durante il periodo delle registrazioni Samuele (BFLA) L. materna L. paterna L. ambiente Ted. Ing. Ita. Circa metà giornata. Limitato (circa un’ora al giorno). Circa metà giornata (il bambino frequenta l’asilo da circa un anno*). * prima dell’inizio dell’asilo Samuele era esposto all’italiano per alcune ore al giorno 103 Manola Salustri 4.3. Marianna Marianna ha iniziato a parlare italiano fin dalla nascita con la baby-sitter e la sorellina Valery, tre anni più grande. La madre si rivolge alle bambine sempre in tedesco mentre il padre usa a volte anche l’italiano. Marianna si esprime correttamente sia in tedesco sia in italiano e presenta una forte inflessione toscana. Oltre a vocaboli dialettali notiamo la presenza della gorgia toscana. Le registrazioni sono state effettuate a casa della bambina, dalla madre, oppure all’asilo. Nella registrazione di Marianna sono presenti molti esempi di Code-Switching la bambina 25 cioè “mischia “ i termini delle due lingue. Tav. 4. Durata di esposizione durante il periodo delle registrazioni. Marianna (BFLA) L. materna L. paterna L. ambiente Ted. Ted. Ita. Circa metà giornata. Alcune ore al giorno. circa metà giornata (la bambina frequenta l’asilo da circa un anno*). * prima dell’inizio dell’asilo Marianna era esposta all’italiano per alcune ore al giorno (baby-sitter italiana) 4.4. Angela Lo sviluppo linguistico di Angela , una bambina di tre anni di madrelingua tedesca, è stato seguito per due mesi. La madre di Angela è tedesca mentre il padre è italiano; Angela ha anche un fratellino, Andrea, di circa cinque anni che come lei ha imparato l’italiano all’asilo, in quanto in famiglia i due bambini parlano solo tedesco. La prima registrazione è stata eseguita a Settembre, solo una settimana dopo l’apertura dell’asilo. Angela non riusciva ad esprimersi in italiano e giocava prevalentemente con gli altri bambini che parlano tedesco. La seconda registrazione è stata effettuata ad Ottobre, sempre all’asilo frequentato dalla bambina. 25 Alcuni esempi di Code-Switching dal corpus di Marianna (MAR- R1.): a. Ich bin auch eine incavolata. Io sono anche una incavolata b. Und das war lustig als du das in den insalata gemacht hast. E questo era divertente quando tu questo nell’insalata messo hai. (Marianna, 3; 8) 104 Bilinguismo e acquisizione infantile di L2 Tav. 5. Durata di esposizione durante il periodo delle registrazioni. Angela (L2) L. materna L. paterna L. ambiente Ted. Ita. Ita. Circa metà giornata. Limitato.* Circa metà giornata ( la bambina frequenta l’asilo da circa una settimana). * prima dell’inizio dell’asilo Angela era esposta quasi esclusivamente al tedesco 5. Dati utilizzati nella ricerca In questa ricerca verranno utilizzati i dati relativi alla lingua tedesca nel corpus di Renzo, Samuele e Marianna, al fine di svolgere un’analisi quantitativa degli errori presenti rispetto alla posizione del verbo nelle frasi tedesche principali e subordinate. Le registrazioni contrassegnate da un asterisco non sono state utilizzate in quanto costituite prevalentemente da frasi in italiano. Verranno quindi considerati i dati relativi alla produzione spontanea di Angela, al fine di paragonare la sua acquisizione dell’italiano come L2 e l’acquisizione dei bambini italiani monolingui nelle stesse fasi dell’apprendimento. Tav. 6. Registrazioni utilizzate nella ricerca. Registrazioni Renzo R1 R2* R3 R4 R5 Samuele R1* R2 R3 R4* Marianna R1 R2* R3* Angela R1 R2 (a.& b.) Età del bambino Data 3;5 3;7 3;9 3;11 4;1 18.01.1998 29.03.1998 04.05.1998 08.07.1998 3;9 3;10 3;11 4;0 24.06.1998 28.07.1998 10.08.1998 24.09.1998 3;8 3;9 4;2 20.12.1997 12.01.1998 24.06.1998 3;0 3;1 24.09.1998 10.10.1998 105 Manola Salustri 6. Errori nella posizione del verbo nei bambini bilingui 6.1. Metodo Nel paragrafo seguente verranno riportate tutte le frasi presenti nel corpus di Renzo con un ordine verbale scorretto. Ogni registrazione sarà analizzata singolarmente prendendo in considerazione prima le frasi principali, quindi le secondarie. Gli errori sono stati controllati da due parlanti di madrelingua tedesca che hanno letto l’intero corpus. Se è stata ipotizzata un’omissione, verrà indicato tra parentesi l’elemento omesso. I risultati dell’elaborazione verranno quindi confrontati con i dati di altri due bambini, Samuele e Marianna. 6.2. Analisi delle frasi principali nel corpus di Renzo: i casi di Verb-First Considerando gli errori presenti nelle frasi principali del corpus di Renzo è emerso che, tranne alcuni rari casi di V3 e V-Final, la maggioranza delle frasi con verbo flesso non in seconda posizione è costituita da casi di Verb-First. Queste frasi, in cui il verbo flesso occupa la prima posizione, vengono analizzate, nella maggior parte dei casi, come il risultato dell’omissione di un elemento dalla prima posizione della frase, il Vorfeld, dovuto ai fenomeni del Topic-drop e del Soggetto nullo infantile, come in (33)-(34). (gli elementi tra parentesi sono omessi). (33) (Du) kommst auch von Fenster rein. (tu) entri anche da finestra omissione del soggetto (Renzo 3; 5) (34) (Das) kannst du ruhig reinwerfen . (questo) puoi tu tranquillamente buttare. omissione dell’oggetto (Renzo 3; 5) Per quanto l’omissione dell’oggetto sia difficile da valutare, in quanto spesso si creano contesti ambigui, le frasi considerate non avrebbero senso e non sarebbero grammaticalmente accettabili senza considerare l’omissione dell’oggetto, identificabile dal contesto (i giudizi sull’omissione degli oggetti sono stati dati da parlanti nativi che hanno letto non solo le frasi ma anche l’intero contesto delle registrazioni). A giudizio di parlanti nativi tedeschi, inoltre, in alcune frasi sembrerebbero essere omessi dalla prima posizione anche altri elementi oltre ai soggetti ed agli oggetti. In alcune frasi il bambino sembrerebbe omettere avverbi come jetzt (= adesso), dann (= poi), da (= qui). Prendiamo in considerazione alcune frasi ed il loro contesto come in (35) in cui l’avverbio jetzt è stato omesso: (35) *REN: 106 und jetzt hab ich das Wasser wieder ausgestellt. E adesso ho io l’acqua di nuovo “chiuso” Bilinguismo e acquisizione infantile di L2 *RUF: mhm *REN: ___ hab ich das Wasser wieder ausgestellt. ___ho io l’acqua di nuovo “chiuso” In questo esempio possiamo facilmente ipotizzare un’omissione, trattandosi della ripetizione della stessa frase. I casi di Verb-First sono schematizzati nelle tabelle seguenti (gli elementi tra parentesi sono omessi). Tav. 7. Registrazione REN1. Data: 18.01.’98. a. (S) b. (S) c. (O) d. (Avv) e. f. Vfnt Vfnt Vfnt Vfnt Vfnt Vfnt Vp S S S S Vp Vp 14 2 5 1 5 4 Nella registrazione REN1. 31 frasi principali su 164 (circa il 20%) presentano 26 il verbo finito in prima posizione. Di queste, 16 sono interpretabili con l’omissione del soggetto dalla prima posizione della frase, 5 frasi presentano invece l’omissione dell’oggetto, mentre una sola frase è interpretabile con l’omissione di un avverbio. Nella prima registrazione sono presenti anche 9 frasi a V1 non interpretabili con l’omissione di un elemento dalla prima posizione, (come in (36)). Queste costituiscono un fenomeno limitato ma non trascurabile (circa il 5% del totale) e, come vedremo, tendono a diminuire fino a scomparire completamente. (36) Kann der Zug dann wieder fahren. Può il treno poi di nuovo correre. (Renzo 3; 5) Tav. 8. Registrazione REN.3. Data: 04.05.’98. a. (S) b. (S) c. (O) d. (O) e. f. g. 26 Vfnt Vfnt Vfnt Vfnt Vfnt Vfnt Vfnt Vp S S S S S Vp O Vp 3 1 1 1 1 1 1 In questa registrazione è presente ache un caso di Verb-final, in 1. 1. Ah, man zurückschauen kann. (Renzo, 3; 5) 107 Manola Salustri Delle 136 frasi principali presenti nella registrazione REN.3., 9 presentano il 27 verbo flesso in prima posizione, quindi il 6,6,%. Possiamo considerare 6 delle 9 frasi principali con Vfnt in prima posizione come il risultato dell’omissione del primo elemento della frase (in particolare, in 4 frasi è stato omesso il soggetto, in 2 l’oggetto), mentre gli ultimi tre casi sembrano non essere causati dall’omissione di elementi dal Vorfeld. Tav. 9. Registrazione REN.4. Data: 07.04.’98. a. (O) b. (S) c.(Avv) d. Vfnt Vfnt Vfnt Vfnt S O S S Vp Vp 11 21 10 4 Nelle frasi principali presenti nella registrazione REN.4. il verbo flesso si trova in 46 frasi su 420 in prima posizione (circa l’11%). In 42 casi ciò può essere attribuito all’omissione di elementi topic nel Vorfeld (in 11 casi si ha l’omissione dell’oggetto, in 21 l’omissione del soggetto ed in 10 l’omissione di un avverbio), 28 mentre quattro frasi sembrano non essere interpretabili con un’omissione. Tav. 10. Registrazione REN5. Data: 19.10.’98. a. (S) b. (S) c. (O) d. (Avv) Vfnt Vfnt Vfnt Vfnt O S S 2 1 2 2 Delle 164 frasi principali presenti nella registrazione REN5., 8 hanno un ordine verbale scorretto, quindi il 4,8% circa. È presente solo una frase con il verbo flesso in ultima posizione mentre le rimanenti 7 frasi sono tutti casi di Verb-First, interpretabili con l’omissione di un elemento dal Vorfeld (in 4 casi si ha l’omissione del soggetto, in 2 casi l’omissione dell’oggetto ed in 2 casi l’omissione di un avverbio. 6.2.2. Conclusione Dall’analisi dei dati emerge che Renzo ha acquisito il fenomeno del V2. Gli errori presenti nelle frasi principali sono, nella maggior parte dei casi, frasi a V1 interpretabili come il risultato del fenomeno del Topic-drop e della fase del Soggetto 27 28 Oltre ai casi di V1, sono presenti in questa registrazione anche 4 casi di V3. Sono invece assenti casi di Verb-final. In questa registrazione sono presenti, oltre ai casi di V1, un caso di Verb-final ed uno di V3. 108 Bilinguismo e acquisizione infantile di L2 nullo infantile. Come si nota dal grafico questi casi diminuiscono progressivamente, passando da un 20% nella prima registrazione ad un 5% circa nell’ultima. Il residuo riscontrato nell’ultima registrazione è attribuibile al fenomeno presente nel tedesco parlato, che persiste anche nella lingua adulta, cioè il Topic-drop. Grafico 1. La linea unita indica le frasi con il verbo flesso in prima posizione interpretabili nella maggior parte dei casi con l’omissione di un elemento dal Vorfeld, risultato del fenomeno del Soggetto nullo infantile e del Topicdrop ( il residuo è dovuto alla permanenza, nella grammatica adulta, di quest’ultimo fenomeno). La linea tratteggiata indica il totale delle frasi con ordine verbale scorretto. Le due linee tendono a coincidere, indicando che errori come V3 e Verb-final sono quasi assenti. Errori frasi principali 25 % Errori 20 15 10 5 0 Gennaio V1 Maggio Luglio Settembre Percentuale totale di errori Mesi Nelle prime registrazioni erano tuttavia presenti delle frasi V1 non interpretabili come il risultato dell’omissione di un costituente dalla prima posizione della frase le quali lasciano ipotizzare una fase V1 in cui il bambino posiziona il verbo flesso in prima posizione, forse interpretabile con un transfer dall’italiano. Queste frasi, pur costituendo il 5,4% delle principali nella prima registrazione, scomparendo completamente nell’ultima registrazione, costituiscono un fenomeno non trascurabile nello sviluppo linguistico di Renzo. 109 Manola Salustri Grafico 2. Come si vede dal grafico il V1 costituisce una vera e propria fase nello sviluppo linguistico di Renzo, che scompare nell’ultima registrazione, quando il bambino ha 4,1 anni. VERB-FIRST non interpretabile con l'omissione di elementi dal Vorfeld 6 5 % V1 4 VERB-FIRST 3 2 1 0 Gennaio Maggio Luglio Settembre Mesi Nel corso dello sviluppo linguistico di Renzo si può notare una graduale diminuzione dei soggetti non realizzati foneticamente, che scendono da circa il 9,2% (rispetto al totale delle frasi) nella prima registrazione ad un 1,8% nell’ultima, 29 mostrando la presenza di una fase del Soggetto nullo infantile. 29 Nel corpus di Renzo sono presenti anche soggetti nulli in posizione non iniziale (circa il 4% rispetto al totale delle frasi nella prima registrazione), in frasi principali del tipo XP V (S), in subordinate ed in interrogative si/no, come di seguito, dove l’elemento tra parentesi è stato omesso: i) Hier ist (das) so. Qui è (ciò) così ii) Wenn (ich) es so gebaut habe. Se (io) ciò così costruito ho. iii) Hast (du) gesehen? (Renzo 3; 5) Hai (tu) visto? Nelle interrogative si/no ( che costituiscono la maggioranza dei casi) l’omissione risulta però ambigua in quanto nella stessa registrazione si nota anche la presenza di forme contratte come in iv.: iv) Haste [=hast du] gesehen? (Renzo 3, 5) 110 Bilinguismo e acquisizione infantile di L2 Grafico 3. Il grafico mostra lo sviluppo della fase del Soggetto nullo infantile: la linea indica i soggetti nulli in posizione topic, che diminuiscono progressivamente. Il “residuo” presente nell’ultima registrazione è attribuibile al fenomeno del topic-drop, che è presente nel tedesco parlato e persiste anche nell’età adulta. Soggetti Nulli % Soggetti Nulli 10 8 6 4 2 0 Gennaio Soggetti Nulli in Posizione Topic Maggio Luglio Settembre Mesi 6.2.3. Analisi delle frasi subordinate nel corpus di Renzo Per quanto riguarda le frasi subordinate è emersa la tendenza da parte del bambino ad usare il verbo flesso in seconda posizione nelle frasi introdotte da “weil” , uso sempre più frequente nel tedesco parlato. Questa congiunzione sembrerebbe aver assunto negli ultimi decenni uno statuto particolare, come confermano le frasi delle mamme e dell’investigatrice tedesca presenti nel corpus. L’uso del V2 introdotto da “weil” sembra non essere legato ad un fattore dialettale. Sono presenti anche frasi subordinate con ordine verbale scorretto introdotte da altre congiunzioni, che, come notiamo dal grafico, scompaiono nelle ultime registrazioni. Anche se queste frasi costituiscono una percentuale molto bassa, esse Considerato che già nella seconda registrazione l’omissione in contesti non iniziali è scesa al 1,4% fino a scomparire totalmente nell’ultima registrazione, questo fenomeno non verrà considerato come rilevante (ma vedi Hamann (1997) per uno studio dei soggetti nulli post-verbali in una fase tarda dello sviluppo del tedesco (Late argument drop)). 111 Manola Salustri costituiscono una differenza rispetto ai corpus dei bambini monolingui dove non si 30 riscontra nessun errore della posizione del verbo nelle frasi subordinate. Grafico 4. Dal grafico si nota che la percentuale delle frasi introdotte dalla congiunzione “weil” con ordine verbale non finale rimangono costanti nel corso dello sviluppo, indicando che non si tratta di una fase ma di un particolare uso riscontrato anche negli adulti. Sono presenti anche frasi subordinate con ordine scorretto che costituiscono circa il 5% delle frasi subordinate con errori e scompaiono nell’ultima registrazione. Errori frasi subordinate 25 % errori 20 15 10 5 0 Gen Errori totali frasi subordinate Mag Lug mesi Sett V2 introdotto da "weil" Dai risultati della ricerca possiamo comunque concludere che il bambino ha acquisito il parametro d’ordine OV/VO. La percentuale degli errori nelle frasi subordinate è infatti molto bassa (circa il 5%) mentre le frasi a V2 introdotte da “weil” non costituiscono un errore ma una caratteristica dell’input. 6.2.4. Confronto con il corpus di Samuele Nella registrazione di luglio 11 frasi su 157, quindi il 7% delle frasi principali presentano un ordine verbale scorretto. Tutte le frasi con un ordine verbale scorretto presentano il verbo flesso in prima posizione. Di queste, 10 sono interpretabili con l’omissione di un elemento dal Vorfeld (soggetto o oggetto), mentre una sola risulta ambigua, in (37). 30 Schönenberg (2000) ha notato, in uno studio sull’acquisizione di una varietà del tedesco (Swiss German Lucernese) che i bambini da lei considerati muovono il verbo nelle frasi subordinate anche in contesti in cui ciò non è permesso nella grammatica adulta. 112 Bilinguismo e acquisizione infantile di L2 (37) Habe ich noch nett [= nicht] fertig. ho io non ancora finito (Samuele 3; 11) Tav. 11. Registrazione SAM 2. Data : 28.07.’98. a. (S) b. (O) V V S 8 2 Delle 6 frasi subordinate presenti, 2 non presentano il verbo flesso in ultima posizione. Una di queste è introdotta da “weil”, l’altra è una interrogativa indiretta. Questi dati mostrano che Samuele ha acquisito il fenomeno del V2 e le frasi con posizione del verbo scorretta possono essere attribuite all’omissione di elementi dal Vorfeld. La percentuale di frasi con omissione di un costituente dalla prima posizione della frase è circa il 6,3%, la percentuale di soggetti omessi in prima posizione è il 5%. Nella registrazione SAM3. del 18.08.’98 è presente un solo caso di frase principale con verbo flesso in posizione scorretta. Ciò potrebbe essere attribuito alla prevalenza in questa registrazione di frasi interrogative, nelle quali il verbo flesso si trova correttamente nella prima posizione. Dall’analisi delle frasi principali risulta che circa l’80% di queste sono frasi interrogative SI/NO. L’unico errore riscontrato è la frase in (38), un’interrogativa. (38) Du hast alles gesehen? tu hai tutto visto (Samuele 4; 0) Per quanto riguarda le subordinate, è stato riscontrato un caso di V2 introdotto da “weil”. (39) weil ich bin schwach, ok? perché io sono debole, ok (Samuele 4; 0) 6.2.4. Confronto con il corpus di Marianna Per quanto riguarda l’ordine verbale, nella prima registrazione di Marianna sono presenti una subordinata introdotta da “weil” con il verbo flesso in terza posizione ed una principale con un ordine verbale V1 dovuta all’omissione del soggetto in prima posizione. Dato il numero limitato di errori, possiamo affermare che anche Marianna posiziona correttamente il verbo flesso nelle principali e nelle subordinate. Anche in questo caso si osserva l’uso del V2 con le subordinate introdotte da “weil” riscontrabile, in questa registrazione, anche nella produzione linguistica della mamma e della sorellina di otto anni. 113 Manola Salustri 7. Acquisizione infantile di L2: alcune osservazioni Nel corso della ricerca sono stati registrati anche bambini venuti in contatto con l’italiano solo dopo i due anni di età. In questo caso, come anticipato nell’introduzione, è più opportuno parlare di acquisizione infantile di L2, seppur precoce, e non di vero e proprio bilinguismo. D’altra parte le difficoltà nel tracciare i limiti tra Bilinguismo ed Acquisizione precoce di L2 rendono forse più opportuno parlare di una fase transitoria attraversata dal bambino nel corso dell’acquisizione, durante la quale si riscontrano gli stessi fenomeni di transfer tipici dell’acquisizione di L2. Nonostante non siano disponibili registrazioni riguardanti lo sviluppo successivo del caso qui considerato, è ragionevole pensare che l’acquisizione della seconda lingua in bambini di tre anni 31 raggiunga livelli linguistici propri dei bambini monolingui e bilingui. Tra i dati raccolti, il corpus più interessante riguarda una bambina di tre anni, Angela, che è stata seguita per circa due mesi. La prima registrazione è stata eseguita a settembre, solo una settimana dopo l’apertura dell’asilo. In poche frasi Angela cerca di esprimersi in italiano usando parole di questa lingua e del tedesco, come in (40)-(41), dove sono presenti anche errori di accordo. (40) Io vuole guck mal man. (Angela 3; 0) (41) Io vuole questo angucken. Nella registrazione di Ottobre Angela ha già migliorato il suo italiano, e nonostante la presenza di errori, si nota una maggiore fluenza, come in (42)-(43). (42) Vuole un foglio. (43) Io vuo un attro disegno. (Angela 3; 1) Sono ancora presenti, tuttavia, errori di accordo, persona e numero, che non si riscontrano invece nei bambini che acquisiscono l’italiano come L1, sia monolingui che bilingui. Nel corpus studiato da Guasti, ad esempio, i bambini presentano un 32 accordo corretto già dalle prime fasi dell’acquisizione linguistica. Possiamo osservare, come dimostrano le frasi in (44)-(45), che in questa fase dell’acquisizione Anna non ha ancora acquisito la persona ed il numero. (44) io vuoi un attro disegno. (45) Io vuo un attro disegno. 31 32 A questo proposito cfr. Belletti & Hamann (1999). Cfr. Guasti (1993/94). 114 Bilinguismo e acquisizione infantile di L2 (46) Vuole un foglio. Sembra inoltre che la bambina provi a formare delle frasi in italiano traducendo direttamente dal tedesco, applicando all’italiano le regole della grammatica di questa lingua, come in (47)-(49). (47) Isegno vuole io. Inversione (48) Quetto vuole io. Inversione (49) È questo mio? Interrogativa con verbo flesso in prima posizione La bambina utilizza i movimenti sintattici tipici del tedesco anche per l’italiano, come l’Inversione, in cui il verbo flesso ed il primo elemento della frase (in questo caso l’oggetto) si spostano dalla posizione di base ad una posizione più alta della struttura frasale. I dati di Angela sembrano quindi essere in accordo con l’ipotesi Full Transfer/Full Access (FT/FA) di Schwartz e Sprouse (1994). Seguendo questa ipotesi la grammatica di L1 costituisce “lo stato iniziale” della grammatica di L2: le proprietà sintattiche di L1 vengono trasferite ad L2 e solo in seguito la grammatica, sottostando ai principi di 33 GU viene ristrutturata secondo le proprietà della lingua target. Conclusioni (i) Nel corpus di Renzo non è stata riscontrata la presenza di infiniti principali, sembra quindi che la fase dei RI sia terminata già dalla prima registrazione (3;5 anni) in accordo con i dati di Wexler (1990) e Rizzi (1994). La fase del Soggetto nullo infantile continua invece fino circa 4 anni. A questa età la percentuale di frasi a V1 diviene la stessa riscontrata negli adulti (circa il 5%). Questo residuo è attribuibile al fenomeno del Topic-drop che caratterizza la lingua tedesca. (ii) Nei termini dell’ipotesi del Troncamento, ciò significa che in questa fase il bambino non ha più l’opzione di troncamenti al livello di TP, responsabili della presenza di infiniti principali, mentre sono ancora possibili troncamenti ad un livello più alto della struttura frasale, responsabili della fase del Soggetto nullo infantile. Solo intorno ai quattro anni il bambino acquisisce completamente il principio: CP = Root (Rizzi 1994) (iii) Il parametro d’ordine OV/VO risulta fissato già dalla prima registrazione (3; 5 anni). Il bambino non commette errori nella struttura frasale, posizionando il verbo flesso nella seconda posizione nelle frasi principali e nell’ultima nelle 33 Seguendo Schwartz (1998), nonostante i diversi “punti di inizio” e di arrivo, i processi cognitivi responsabili dell’acquisizione di L1 ed L2 sarebbero comunque gli stessi. 115 Manola Salustri 34 subordinate, mostrando così di aver acquisito il fenomeno del V2 . La presenza di frasi V1 viene interpretata come il risultato del fenomeno del Soggetto nullo infantile e del fenomeno del Topic-drop. Solo una piccola percentuale di frasi V1 non è attribuibile a questi due fenomeni e necessita una diversa interpretazione. (iv) Il bambino ha fissato negativamente il parametro del pro-drop, non sono presenti infatti soggetti nulli in posizione post-verbale o post-wh (ma vedi anche nota 29), mentre il parametro del Topic-drop è stato fissato positivamente, il bambino infatti omette non solo soggetti ma anche oggetti. Quest’ultimo 35 fenomeno viene ipergeneralizzato dal bambino anche agli avverbi . (v) Nelle frasi subordinate è stato riscontrato un numero limitato di errori (circa il 5%) mentre è emersa la tendenza da parte del bambino ad usare il V2 nelle frasi introdotte dalla congiunzione “weil”, uso riscontrato anche nell’input e già notato in precedenti lavori (cfr. tra gli altri Hamann (1997)). (vi) I dati di Samuele e Marianna confermano i risultati riscontrati in Renzo. I bambini posizionano correttamente il verbo flesso ed hanno fissato i parametri del pro-drop e del Topic-drop. In entrambi i casi sono presenti frasi subordinate introdotte da “weil” con verbo flesso in seconda posizione. La percentuale di soggetti nulli omessi in prima posizione da Samuele a 3;10 anni è confrontabile con la percentuale presente in Renzo all’età di 3;11 anni (entrambe circa il 5%). Sembra tuttavia che la fase del soggetto nullo sia terminata in Marianna già dalla prima registrazione (3;2 anni). (vii) Nei dati relativi all’acquisizione dell’italiano come L2 da parte di Angela, una bambina tedesca di tre anni, sono presenti interferenze con il tedesco e differenze rispetto ai dati relativi all’acquisizione dell’italiano in bambini monolingui. La presenza di frasi con oggetto topicalizzato e frasi con inversione del soggetto nei dati di Angela indicano la presenza di perlomeno una categoria 34 Un’ulteriore prova per quanto riguarda l’acquisizione del parametro d’ordine OV/VO è costituita dal fatto che Renzo non commette errori nelle frasi del tipo in i. (i) ich habe das gesehen. S Vfin O Vp Io ho ciò visto. Mentre il verbo finito in queste frasi si sposta nella periferia sinistra della frase, in un punto più alto della struttura, il verbo all’infinito resta nella posizione di base. Trovandosi l’oggetto sempre alla sinistra del verbo non flesso nelle costruzioni perifrastiche, possiamo dedurre che il bambino ha fissato correttamente il parametro d’ordine OV/VO. 35 Peraltro, l’omissione di un avverbio dalla prima posizione della frase, specialmente della particella “da”sembra essere accettata anche da parlanti adulti del tedesco nel registro parlato (Rizzi, L., p.c.). 116 Bilinguismo e acquisizione infantile di L2 funzionale fuori da VP. D’altra parte i numerosi errori di accordo (circa il 28%) sembrerebbero mostrare l’assenza di Agr. L’assenza di errori o interferenze nell’acquisizione della posizione del verbo flesso nella frase tedesca sembra indicare che Renzo, Samuele e Marianna, esposti all’italiano ed al tedesco sin dalla nascita, quindi casi di Bilingual First Language Acquisition, sono in grado di separare le due grammatiche e presentano le stesse fasi nell’acquisizione riscontrate nei bambini monolingui riguardo a questo aspetto della 36 grammatica tedesca. I dati di Angela, invece, riguardanti l’Acquisizione infantile di L2 mostrano che la bambina, esposta alle due lingue solo dopo i due anni di età, tende a confondere le due grammatiche e l’acquisizione segue, almeno inizialmente, processi diversi. Bibliografia Belletti A. & L. Rizzi (1996), Parameters and functional heads, Oxford University Press. Belletti A. & C. Hamann (1999), “Ça on fait pas! On the L2-Acquisition of French by two your children with different source languages”, BUCLD 24 Proceedings. Bohnaker U. (1998), Iceland plus English: language differentiation and functional categories in a successively bilingual child, tesi di dottorato, University of Durham, UK. Bromberg H. & K. Wexler (1995), “Null subjects in wh-questions”, in C.T. Schutze et al. (eds.) Papers on language processing and acquisition, 221-48. Cardinaletti A. & G. Giusti (1996), Problemi di sintassi tedesca, Unipress, Padova. Cipriani P., A.M. Chilosi, P. Bottari & L. Pfanner (1993), L’acquisizione della morfosintassi in italiano: fasi e processi, Unipress, Padova. Chlasen H. (1982), Spracherweb in der Kindheit. Eine Untersuchung zur Entwicklung der Syntax bei Kleinkindern, Narr, Tübingen. Clahsen H. (1990), “Constraints on parameter setting: a grammatical analysis of some stages in German child language”, Language Acquisition, 1: 297-335. De Houwer A. (1996), “Bilingual first language acquisition”, in The Handbook of Child Language, Fletcher P. & B. MacWhinney (eds.), Blackwell Publisher Inc. Cambridge. Besten H. den (1983), “On the interaction of Root Transformations and lexical deletive rules”, in W. Abraham (ed.), On the formal Syntax of the Westgermania, 47-131, John Benjamins, Amsterdam. Friedemann M.-A. & L. Rizzi (2000), The acquisition of syntax, Longman. Guasti M.T. (1993/94), “Verb syntax in Italian child grammar: finite and non finite verbs”, Language acquisition, 3 (1): 1-40. 36 L’analisi dei dati in italiano, rimandata ad uno studio futuro, permetterà una visione più completa dello sviluppo linguistico di questi bambini bilingui, che non presentano differenze rilevanti per quanto riguarda il tedesco. 117 Manola Salustri Guasti M.T., Introduction to language acquisition, MIT Press (in stampa). Hamann C. (1997), From syntax to discourse: children’s use of pronominal clitics, nullarguments, infinitives and operators, Habilitationsschrift, Universität Genf Sommersemester 1997. Haegeman L. (1995), Root infinitives and initial root null subject in early Dutch, Proceedings of GALA. Haegeman L. (1996), “Verb-second, the split-CP and null subject in early Dutch Finite Clauses”, GenGenP 4 (2). Haegeman L. (2000), “Adult null subjects in non pro-drop languages”, in M.-A. Friedemann & L. Rizzi. Hyams N. (1986), Language acquisition and the theory of parameters, Reidel, Dordrecht. Hyams N. & K. Wexler (1993), “On the grammatical basis of null subjects in child language”, Linguistic inquiry 24: 421-453. Wang Qi, D. Lillo-martin, C.T. Best & A. Levitt (1992), “Null subject versus null object: some evidence from the acquisition of Chinese and English”, Language Acquisition, 2: 221-254. Huang C.-T.J. (1984), “On the distribution and reference of empty pronouns”, Linguistic Inquiry, 15: 531-574. Lasnik H. & T. Stowell (1991), “Weakest crossover”, Linguistic Inquiry, 22: 687-720. Meisel J. (1986), “Word order and case marking in early child language. Evidence from simultaneous acquisition of two first languages: French and German”, Linguistic, 24: 123-183. Meisel J. (1990), Two first languages, early grammatical development in bilingual children, Foris Publications, Dordrecht-Holland. Meisel J. & N. Müller (1992), “Finitess and verb placement in early child grammar: evidence for simultaneous acquisition of French and German bilinguals”, in J. Meisel (ed.), The acquisition of verb placement: functional categories and V2 phenomena in language development, Kluwer, Dordrecht, 109-138. Meisel J. (1992), The acquisition of verb placement, Kluwer Academic publishers, Holland. Meisel J. (1994), Bilingual first language acquisition: French and German grammar development, John Benjamins, Amsterdam. Poeppel D. & K. Wexler (1993), “The full competence hypothesis of clausal structure in early German”, Language, 69: 1-33. Rizzi L. (1986), “Null objects and the theory of pro”, Linguistic inquiry 17, ripubblicato in Rizzi (2000b). Rizzi L. (1991), “Residual verb –second and the Wh-criterion”, Technical reports on formal and computational linguistics N. 2, Geneva University, ripubblicato in Rizzi (2000b). Rizzi L. (1994), “Some notes on linguistic theory and language development: the case of root infinitives”, Language Acquisition 3: 371-393, ripubblicato in Rizzi (2000b). Rizzi L. (1994), “Early null subject and root null subject”, T. Hoekstra & B. Schwarzt (eds.), Language acquisition studies in Generative Grammar, Benjamins, Amsterdam, ripubblicato in Rizzi (2000b). 118 Bilinguismo e acquisizione infantile di L2 Rizzi L. (1997), “The fine structure of the left periphery”, in L. Haegemann (ed.), Elements of grammar. Handbook of generative syntax, Kluwer, Dordrecht, the Netherlands, ripubblicato in Rizzi (2000b). Rizzi L. (2000a), Remarks on early null subject and root infinitives, in M.-A. Friedemann & L. Rizzi (2000). Rizzi L. (2000b), Comparative syntax and language acquisition, Routledge, London. Rothweiler M. (1993), Der Erverb von Nebensatzen in Deutschen, Niemeyer, Tübingen. Santorini B. (1995), “Two types of verb second in the history of yddisch”, in A. Battye & I. Roberts (eds.), Clause structure and language change, Oxford University Press, New York. Schönenberg M. (1998), The acquisition of verb placement in Swiss German, Thése de doctorat, Université de Geneve. Schönenberg M. (2000), “The acquisition of verb placement in lucernese Swiss German” in M.A. Friedemann & L. Rizzi (2000). Schwartz B.D. (1998), “The second language instinct”, in Lingua, 106: 133-160. Thiersch C. (1978), Topics in German syntax, PhD Dissertation, MIT. Tomaselli A. (1989), La sintassi del verbo finito nelle lingue germaniche, tesi di dottorato, Università di Pavia. Valian V. (1991), “Sintactic subjects in the early speech of American and Italian children”, Cognition, 40: 21-81. Vikner S. (1994), “Finite verb movement in Scandinavian embedded clauses”, in N. Horstein & D. Lightfoot (eds.), Verb Movement, Cambridge University Press. Vikner S. (1995), Verb movement and expletive subjects in the germanic languages, Oxford University Press, New York. Wang Q., D. Lillo-Martin, C.T. Best & A. Levitt (1992), “Null subject versus null object: some evidence from the acquisition of Chinese and English”, Language Acquisition. Wexler K. (1992), “Optional infinitives, head movement, and the economy of derivations in child grammar”, MIT, Occasional Paper #14. 119 ON THE RELATIVE POSITION OF BEAUCOUP, GUÈRE, PEU, RIEN AND TROP IN FRENCH 1 Sara VECCHIATO 1. Introduction Adverbs and quantifiers have been studied extensively in linguistics and several accounts of their distributional properties have been provided under different approaches. In the generative framework, their nature and behaviour have been considered as possible clues for the internal structure of the sentence. Since Pollock (1989), word order variation concerning adverbs has been dealt with by assuming that it is the verb that moves around the adverb, which always remains in situ. Instead, quantifiers are taken to move leftward in the sentence, towards a „scope‟ position, either in visible syntax or in Logical Form. What we have seen is an attempt to provide simple and elegant accounts of apparently puzzling facts. Among the various hypotheses made, I will be mainly concerned with Cinque (1999)‟s proposal that there is a single universal hierarchy of adverbs, where each adverb occupies the SPEC position of a functional projection marked with a certain feature (aspect, tense, modality, etc). The idea is that adverbs are licensed by the relevant feature associated with their corresponding functional head. My purpose in this article is to examine a few French adverbs and quantifiers, not considered in Cinque ‟99 and to try to determine their position in the hierarchy proposed there. Most of them actually belong to both categories. In fact, while rien „nothing‟ can be used only as a negative quantifier, beaucoup „a lot‟, peu „little‟, trop „too much‟, guère „not much‟ can be used either as quantifiers or as adverbs. It would be difficult to establish whether they are actually the same words being used in two ways - as bare quantifiers and as quantificational adverbs - or distinct homonymic words. 1 I wish to thank Paolo Acquaviva, Maria Teresa Biason, Guglielmo Cinque, MarieChristine Jamet, Dominique Sportiche and Michal Starke for native judgments and/or helpful comments and suggestions. 121 Sara Vecchiato The testing method is the same as Cinque‟s (‟99) - namely, a comparison between minimal pairs of sentences in which two adverbs appear in opposite orders. (1) a. Pierre a tout beaucoup aimé. Pierre has everything a-lot liked. Pierre liked everything a lot. b. *Pierre a beaucoup tout aimé. The order accepted by French native speakers is taken as relevant to establish the position of the examined items in the hierarchy. It seems that each couple adverbquantifier (i.e. guère, trop, beaucoup / peu both as QPs and as AdvPs) occupies exactly the same position in the hierarchy. Despite some difficulties in testing some couples of adverbs due to their semantic incompatibility (for example, beaucoup „a lot‟ and complètement „completely‟), the data point decisively to this direction. This implies the existence, in the hierarchy, of aspectual heads licensing both adverbs and quantifiers by checking the same features. The identified positions are presumably the scope positions to which quantifiers move in overt syntax. I will give some suggestions concerning the names of some heads‟ features. This article is organized as follows: in the first part I will present the full hierarchy of Cinque (‟99), which is the basis for my analysis; then, I will examine the distributional properties of the QPs/quantificational AdvPs with respect to the auxiliary and lexical verbs in active and passive sentences. The third part is devoted to the crucial data concerning the location of the examined items, followed by an addictional section meant to show that, if beaucoup, guère, peu, rien and trop are tested with “lower” adverbs other than those presented in the key-sentences, their mutual order is consistent with the identified positions. 2. The hierarchy The AdvP hierarchy is given in English, since this is the original version of Cinque‟s work. However, it is positively confirmed by French data. There are three French items; tout, corresponding to English everything, and two duration adverbs, longtemps and longuement, which can be both roughly translated with for long. The location of tout was identified in Cinque (‟99), while I have added longtemps and longuement (see Vecchiato (‟99)). These two adverbs exactly cover the position of brièvement „briefly‟, formerly established as the only duration adverb in the hierarchy. [frankly MOODspeech act [fortunately MOODevaluative [allegedly MOODevidential [probably MOODepistemic [once T(Past) [then T(Future) [perhaps MOOD irrealis [necessarily MOODnecessity [possibly MODpossibility [usually ASPhabitual [again ASPrepetitive (I) [often ASPfrequentative(I) [intentionally MODvolitional [quickly ASPcelerative (I) [already T (Anterior) [no longer ASPterminative [still ASPcontinuative [always ASPperfect [just ASPretrospective [soon ASPproximative [longtemps ASPdurative (I) [longuement ASPdurative (II) [characteristically (?) ASP122 On the relative position of beaucoup, guère, peu, rien and trop in French [almost ASPprospective [completely ASPSgCompletive (I) [tout ASPPlCompletive [well Voice [fast/early ASPcelerative (II) [completely ASPSgCompletive (II) [again ASPrepetitive (II) [often ASPfrequentative (II) ]]]]]]]]]]]]]]]]] generic/progressive It will be seen that all the examined items occupy “lower” positions in this hierarchy, namely, between presque „almost‟ and bien „well‟. Adverbs directly occupy their own position in the hierarchy, whereas quantifiers move leftward in the sentence, from the canonical object position [V, NP] towards their „scope‟ position, where their fea2 tures are checked . The scope position A‟-binds the canonical object position. (2) Il a [touti [compris ti ]]. He has everything understood He understood everything 3. Distributional properties In order to establish the distribution of French QPs/quantificational AdvPs in the sentence, three positions have been taken into account: (i) the position which immediately follows the auxiliary verb; (ii) the position which immediately follows the active past participle or the passive auxiliary été (been); (iii) the position between the active past participle and the direct object or between the auxiliary été and the passive participle. Sentences (3a-c) illustrate this pattern. When French bare QPs are the subjects of passive sentences, they are allowed not to be in [Spec, IP] if the expletive pronoun il (it) is used (sentence (4)). In this configuration, they are thought to occupy their base position. We will see that guère „not much‟, trop „too much‟ and rien „nothing‟ cannot follow the past participle, either as adverbs or as quantifiers, except if they have a focus reading. Instead, for some reasons, beaucoup „a lot‟ and peu „little‟, as quantifiers, can also occur in the (post-participial) object position: [V‟, NP]. We will also see that the patterns of distribution vary considerably with respect to the passive auxiliary été (been). I will argue that this is not due to AdvP-movement, but that the verbal head moves further leftwards than the past participle through being an auxiliary. 2 “ Lower” adverbs and quantifiers are delimited on the left by the past participle in Italian presumably because the Italian past participle is allowed to move further to the left in the sentence than its French equivalent. Then, the apparent word order of an Italian sentence looks as if tutto (everything) has remained in its base position, but in fact it has not: (i) Ha capitok [tuttoj [tk tj]] pro has understood everything He understood everything 123 Sara Vecchiato (3) a. Marie a soigneusement peigné François. Marie has carefully combed François. Marie combed François carefully. b. Marie a peigné soigneusement François. c. Marie a peigné François soigneusement. (4) Il a beaucoup été fait pour eux. Itexpl has a-lot been done for them A lot was done for them. 3 Guère and trop The distribution of guère and trop varies with respect to the passive auxiliary été „been‟ according to their being quantifiers or adverbs. Namely, adverbs can appear in position (ii) in passive sentences, whereas quantifiers cannot. Guère-QP: Michel n‟a guère mangé. Michel notcl has not-much eaten Michel didn’t eat much b. *Michel n‟a mangé guère. (5) a. (6) a. Il n‟a guère été fait pour les sauver. Itexpl notcl has not-much been done to themcl save Not much was done to save them. b. *Il n‟a été guère fait pour les sauver. c. *Il n‟a été fait guère pour les sauver. 3 As a verb modifier, guère is currently used with the negative marker ne „not‟ in Modern French. These two negative expressions are interpreted as just one semantic negation (“Negative Concord”): Je n’aime guère ce quartier „I don‟t like this district much‟. As a bare quantifier it is very formal: Le nom d’Alain ne me disait guère „The name of Alain didn‟t tell me much‟. Guère was used without ne, with a positive meaning, in Middle French: Si nature ne prête un peu, il est malaisé que l’art et l’industrie aillent guiere avant (Montaigne) „If Nature doesn‟t help a bit, it is difficult that industry and the arts make a lot of progress‟. In Modern French, it can actually be used without ne in elliptical answers or as an adverb modifier, though it keeps its current negative meaning: Tu connais l’opium? - Guère (Malraux) „Do you know opium? - Not much‟; Une autre possibilité, guère moins irritante, était qu’il s’en fichait peut-être „Another, not much less sad, possibility was that he didn‟t give a damn about it‟. For an analysis of guère in a hierarchy of negative projections, see Zanuttini (‟97). 124 On the relative position of beaucoup, guère, peu, rien and trop in French Trop-QP: (7) a. Maurice a trop mangé. Maurice has too much eaten Maurice ate too much. b. *Maurice a mangé trop. (8) a. Il a trop été fait pour le repas; on a gaspillé la nourriture. Itexpl has too much been done for the lunch PRON impers has wasted the food Too much was done for lunch; we wasted the food. b. *Il a été trop fait pour le repas; on a gaspillé la nourriture. c. *Il a été fait trop pour le repas; on a gaspillé la nourriture. Guère-AdvP: (9) a. Michel n‟a guère changé l‟ameublement. M. notcl has not much changed the furniture Michel hasn’t changed the furniture much. b. *Michel n‟a changé guère l‟ameublement. c. *Michel n‟a changé l‟ameublement guère. (10) a. Ce livre n‟a guère été lu l‟année dernière. This book not cl has not-much been read the year last This book wasn’t read much last year. b. Ce livre n‟a été guère lu l‟année dernière. c. *Ce livre n‟a été lu guère l‟année dernière Trop-AdvP: (11) a. Maurice a trop étudié la partition. M. has too-much studied the score Maurice studied the score too much. b. *Maurice a étudié trop la partition. c. *Maurice a étudié la partition trop. (12) a. La partition a trop été changée. The score has too-much been changed The score was changed too much. b. La partition a été trop changée. c. *La partition a été changée trop. 125 Sara Vecchiato Beaucoup and peu Beaucoup and peu, as quantifiers, enjoy more possibilities than guère and trop. In fact, they can appear also in the post-participial positions. Sentences (13)-(14) and (15)-(16) actually differ in their intonational contour, though the post-participial position is not necessarily focussed. Beaucoup-QP: (13) a. Kay a beaucoup fait pour sa famille. Kay has a-lot done for her family Kay did a lot for her family. b. Kay a fait beaucoup pour sa famille. (14) a. Il a beaucoup été fait pour les sauver, mais sans résultat. Itexpl has a lot been done to themcl save, but without result A lot has been done to save them, but unsuccessfully. b. Il a été beaucoup fait pour les sauver, mais sans résultat. c. Il a été fait beaucoup pour les sauver, mais sans résultat. Peu-QP: (15) a. Lola a peu fait pour sa famille. L. has little done for her family Lola did little for her family. b. Lola a fait peu pour sa famille. (16) a. Il a peu été fait pour les sauver, et tout le monde a du remords. Itexpl has little done to themcl save, and all the world has artpartitive remorse Little was done to save them, and everybody feels remorse. b. Il a été peu fait pour les sauver, et tout le monde a du remords. c. Il a été fait peu pour les sauver, et tout le monde a du remords. 4 As intensity adverbs , beaucoup and peu can appear in position (ii) only in passive sentences, like trop and guère. Then, in (18b) and (20b) respectively, beaucoup and 4 Beaucoup and peu have not only an intensive value, but also a (temporal) frequentative value. These two readings are apparently associated with two different positions in the hierarchy, since beaucoup can occur twice in a sentence: La pièce a beaucoup été beaucoup changée. „The play has been very changed many times‟. The frequentative beaucoup apparently enjoys one position more than its intensive equivalent both in active sentences and in passive sentences. On the contrary, the frequentative peu does not differ in its distribution from peu-intensive-AdvP. 126 On the relative position of beaucoup, guère, peu, rien and trop in French peu modify the past participle discuté „discussed‟ and frappé „struck‟. Since beaucoup 5 is incompatible with adjectives , it can be claimed that the past participle is a verbal head, here, and not an adjective. We will see the relevance of this point below. Beaucoup-intensity AdvP: (17) a. On a beaucoup discuté ce projet à la réunion. PRONimpers has a lot discussed this project at the meeting. We discussed this project a lot at the meeting. b. *On a discuté beaucoup ce projet à la réunion. c. *On a discuté ce projet beaucoup à la réunion. (18) a. Ce projet a beaucoup été discuté à la réunion. This project has a lot been discussed at the meeting This project was discussed a lot at the meeting b. Ce projet a été beaucoup discuté à la réunion. c. *Ce projet a été discuté beaucoup à la réunion. Peu-intensity AdvP: (ii) a. On a beaucoup discuté ce projet ces derniers jours. PRONimpers has a lot discussed this project these latest days. We have discussed this project a lot of times the latest days. b. On a discuté beaucoup ce projet ces derniers jours. c. *On a discuté ce projet beaucoup ces derniers jours. (iii) a. Ce projet a beaucoup été discuté ces derniers jours. This project has a lot been discussed these latest days This project has been discussed a lot of times the latest days. b. Ce projet a été beaucoup discuté ces derniers jours. c. ?Ce projet a été discuté beaucoup ces derniers jours. (iv) a. On a peu discuté ce projet ces derniers jours. PRONimpers has little discussed this project these latest days We have discussed this project few times the latest days. b. *On a discuté peu ce projet ces derniers jours. c. *On a discuté ce projet peu ces derniers jours. (v) a. Il a peu été frappé par la violence à la télé. He has little been struck by violence on TV He has been struck few times by violence on TV b. Il a été peu frappé par la violence à la télé. c. *Il a été frappé peu par la violence à la télé. 5 In French, the modifier of APs (and AdvPs) is très „very‟: Il est très aimable „He is very lovable‟ vs *Il est beaucoup aimable. 127 Sara Vecchiato (19) a. On a peu discuté ce projet à la réunion. PRONimpers has little discussed this project at the meeting We little discussed this project at the meeting. b. *On a discuté peu ce projet à la réunion. c. *On a discuté ce projet peu à la réunion. (20) a. Il a peu été frappé par cette prédiction. He has little been struck by this prediction He was little struck by this prediction. b. Il a été peu frappé par cette prédiction. c. *Il a été frappé peu par cette prédiction. 6 Rien (21) a. Paul n‟a rien fait. Paul notcl has nothing done Paul didn’t do anything. b. *Paul n‟a fait rien. (22) a. Il n‟a rien été fait. Itexpl notcl has nothing been done Nothing was done. b. Il n‟a été rien fait. c. *Il n‟a été fait rien. The reader has easily noticed the lack of symmetry in the distribution of the examined items with respect to the passive auxiliary été „been‟. As Pollock (‟89) argued, it is much costlier to postulate that AdvPs move in the sentence than to assume that only verbal heads move, creating the illusion of adverbial „transportability‟ (see 6 The distribution of its semantic opposite tout „everything‟ is slightly different, since it can never appear in the second position: (vi) a. André a tout compris. André has everything understood. André understood everything. b. *André a compris tout. (vii) a. Il a tout été entrepris pour les sauver, mais sans résultat. Itexpl has everything been undertaken to them cl save, but without result Everything was undertaken to save them, but unsuccessfully. b. *Il a été tout entrepris pour les sauver, mais sans résultat. c. *Il a été entrepris tout pour les sauver, mais sans résultat. 128 On the relative position of beaucoup, guère, peu, rien and trop in French Travis (‟88)). Giving up with the idea that adverbs move, then, it may be argued that quantificational adverbs gain the position (ii) because the past participle is, in fact, an adjective, as traditional grammars claim. Thus, été „been‟ would be a copula and the analysed adverbs would specify the participial adjective. This is actually the unmarked choice in French, since the adjective modifier très „very‟ is currently employed in passive sentences: (23) Ce livre a été très apprécié. This book has been very appreciated This book was very appreciated. However, it is definitely impossible to claim this in the case of beaucoup, because beaucoup is not an adjective modifier, as said above. Then, it is less costly to assume a unique derivation for all the cases, namely that été, being an auxiliary, simply moves further leftward than the active past participle. The fact that bien shares this pattern as well (see ftn. 9) also supports this hypothesis. 4. Location within the hierarchy In this section, I try to identify the position of guère, trop, beaucoup, peu and rien in the hierarchy, both as quantifiers and intensity adverbs. In the data given below, the reader will notice few “gaps” due to the fact that some couples of adverbs are impossible to test since they are semantically incompatible. Unfortunately, one of these gaps 7 concerns the preverbal adverb complètement „completely‟ , which is often the keyadverb to establish the position of some items. We will see if and when we can make up for this difficulty. My hypothesis, to which the data seem to point, is that each couple adverb-quantifier occupies exactly the same position in the hierarchy. By the way, we can observe that the behaviour of both longtemps and longuement is the same as that we would expect form brièvement „briefly‟. Trop Trop-AdvP and trop-QP are both found after presque „almost‟ and before complètement „completely‟. I chose to test these adverbs in passive sentences to make 7 “Completamente can occupy two distinct positions; a preverbal and a post-object one, associated with two distinct interpretations, which likely depend on their different scope” (Cinque (‟99), p. 172): (viii) a. John completely forgot her instructions b. John forgot her instructions completely The first sentence can only mean that her instruction didn‟t occur at the appropriate moment, while the second sentence can also mean that John forgot every part of her instruction. 129 Sara Vecchiato sure a given order was not established because the first adverb was the specifier of the second one. Since the participle été „been‟ can be found between the two adverbs, we are certain that they do not form a constituent. Presque „almost‟ > trop „too much‟ 8 (24) a. Ma mère a presque été trop touchée par cette nouvelle. My mother has almost been too much touched by this piece of news. My mother has almost been touched too much by this piece of news. b. *Ma mère a trop été presque touchée par cette nouvelle. Trop „too much‟ > complètement „completely‟: (25) a. Les employés ont trop été complètement exploités. The employees have too much been completely exploited The employees have been completely exploited too much b. *Les employés ont complètement été trop exploités. Presque „almost‟ > trop „too much‟ (26) a. Il a presque été trop fait pour moi, et rien pour toi! Itexpl has almost been too much done for me, and nothing for you Too much was almost done for me, and nothing for you! b. *Il a trop été presque fait pour moi, et rien pour toi! Trop „too much‟ > complètement „completely‟ (27) a. Il a trop été complètement refait, ça semble artificiel. Itexpl has too much been completely redone, it looks artificial Too much has been completely redone, it looks artificial b. *Il a complètement été trop refait, ça semble artificiel. Guère Guère is found after presque „almost‟ and before trop „too much‟, both as an adverb (28-31) and as a quantifier (32-35). Actually, presque can occur with guèreAdvP only in the active sentence (28), whereas in the passive sentence (29) both the 8 It could be argued that presque is found before trop through a kind of modification by transitivity - that is, presque modify the VP and, by transitivity, trop. In fact, in Italian, È quasi stata troppo commossa da questa notizia can mean both „She risked being too touched by this piece of news‟ and „She was too touched by this piece of news‟. However, in the appendix the reader can find evidence that trop actually follows all the adverbs higher than presque. 130 On the relative position of beaucoup, guère, peu, rien and trop in French orders seem to be impossible. It could be claimed that (28) is possible only because presque is the specifier of guère. However, in the Appendix we can see that guère actually follows all the adverbs before presque. Moreover, guère-QP clearly follows presque both in the active and in the passive sentence. Then, I would think it is the semantic awkwardness of (29) that causes its rejection. As to the objection that guère forms a constituent with trop, it is refuted by the fact that in a passive sentence the participle été „been‟ separates the two adverbs. Presque „almost‟ > guère „not much‟ (28) a. ?Les manifestants n‟ont presque guère gâché l‟exposition. The demonstrators notcl have almost not very much spoiled the exhibition The demonstrators haven’t almost spoiled the exhibition very much. b. *Les manifestants n‟ont guère presque gâché l‟exposition. (29) a. *L‟exposition n‟a presque été guère gâchée par les manifestants. The exhibition notcl has not very much been almost spoiled by the demonstrators b. *L‟exposition n‟a guère été presque gâchée par les manifestants Guère „not much‟ > trop „too much‟ (30) a. Les révolutionnaires n‟ont guère trop changé la morale publique. The revolutionaries notcl have not-much too-much changed the morals public Revolutionaries haven’t much changed public morals too much. b. *Les révolutionnaires n‟ont trop guère changé la morale publique. (31) a. La morale publique n‟a guère été trop changée par les révolutionnaires. The morals public notcl has not-much been too-much changed by the revolutionaries. Public morals haven’t been much changed too much by revolutionaries. b. *La morale publique n‟a trop été guère changée par les révolutionnaires. Presque „almost‟ > guère „not much‟ (32) a. Les manifestants n‟ont presque guère fait contre la mairie, le soir, mais le lendemain ils ont tout détruit. The demonstrators notcl not-much almost done against the town hall, the evening, but the following day they have everything destroyed. The demonstrators almost didn’t much against the town hall in the evening, but the following day they destroyed everything. b. *Les manifestants n‟ont guère presque fait contre la mairie, le soir, mais le lendemain ils ont tout détruit. 131 Sara Vecchiato (33) a. Il n‟a presque été guère gâché par les manifestants, le soir, mais le lendemain la mairie a été mise en sac. Itexpl notcl almost been not-much spoiled by the demonstrators, the evening, but the following day the city hall has been put in sack Not much was almost spoiled by the demonstrators, in the evening, but the following day the city hall was sacked. b. *Il n‟a guère été presque gâché par les manifestants, le soir, mais le lendemain la mairie a été mise en sac. Guère „not much‟ > trop „too much‟ (34) a. Les révolutionnaires n‟ont guère trop changé dans le monde. The revolutionaries notcl have not-much too-much changed in the world Revolutionaries haven’t much changed too much in the world. b. *Les révolutionnaires n‟ont trop guère changé dans le monde. (35) a. Il n‟a guère été trop repeint dans l‟église, presque tous les originaux sont encore là. Itexpl notcl has not-much been too-much repainted in the church, almost all the originals are still there Not much was repainted too much in the church almost all, the originals are still there. b. *Il n‟a trop été guère repeint dans l‟église, presque tous les originaux sont encore là. Beaucoup It is perfectly clear that beaucoup-AdvP is between tout „everything‟ and bien „well‟. Bien can be employed as an IP-adverb, meaning in fact, definitely (Le prof a bien analysé le théâtre ‘In fact, the teacher has analysed theater‟), which obviously 9 produces ambiguity. Thus, I chose to use its modified form très bien „very well‟ . 9 Très can be adjoined to bien without making its distribution wider nor narrower. The pattern is the same as that of beaucoup/peu/trop/guère intensive AdvPs and rien. (ix) a. Liliane a (très) bien compris la question. Liliane has very well understood the question Liliane understood the question (very) well. b. *Liliane a compris (très) bien la question. c. *Liliane a compris la question (très) bien. 132 On the relative position of beaucoup, guère, peu, rien and trop in French Tout „everything‟ > beaucoup „a lot‟ 10 (36) a. Pierre a tout beaucoup aimé. Pierre has everything a-lot loved Pierre liked everything a lot. b. *Pierre a beaucoup tout aimé. Beaucoup „a lot‟ > bien „well‟. 11 (37) a. ?Le prof a beaucoup très bien analysé la pièce de théâtre . The teacher has a-lot very well analysed the theater play. The teacher analysed the play a lot and very well. b. *Le prof a très bien beaucoup analysé le théâtre. As to beaucoup-QP, there are two difficulties in establishing its position: first, it is impossible to test its relative order with tout, because they would compete for the position of subject or direct object in the same sentence; second, beaucoup-QP can12 not appear with complètement „completely‟ . (x) 10 11 12 a. On a (très) bien été réchauffés par le feu du camping. PRONimpers has well been warmed by the camp-fire. We have been warmed well by the camp-fire. b. On a été (très) bien réchauffés par le feu du camping. c. *On a été réchauffés (très) bien par le feu du camping. Compare the passive sentence: Il a tout été beaucoup apprécié „Everything was much appreciated‟ vs *Il a beaucoup été tout apprécié. French speakers show some resistance to beaucoup and (très) bien appearing in the same sentence without being coordinated. The form beaucoup et (très) bien is much preferred. However, the passive sentence is judged as perfect: La pièce a beaucoup été très bien analysée par notre professeur „The play was analysed a lot very well by our teacher vs *La pièce a très bien été beaucoup analysée par notre professeur. If beaucoup is the „dislocated‟ quantifer of a DP, it follows complètement. By dislocation I do not mean Left Dislocation, but a typical French configuration which Obenauer (‟94) defines Quantification à Distance (Quantification from the distance, QAD). QAD allows quantifiers to precede the past participle while the quantified DPs follow it: (xi) J‟ai beaucoup lu de livres I have a lot read of books I read a lot of books Obenauer (‟94) supposes, following Kayne (‟81), that from its base structure [Q [NP]], the quantifier is allowed to move towards its scope position, which is identified with [Spec, VP]. Moving to [Spec, VP], the quantifier gains a frequentative value: “I often read books”. (xii) J‟ai beaucoupk lu [tk de livres]. 133 Sara Vecchiato (38) a. *Fanny a complètement beaucoup caché. Fanny has completely a lot hidden *Fanny a beaucoup complètement caché. However, beaucoup-QP follows presque „almost‟ and precedes bien, which delimits the range of its possible locations. Presque „almost‟ > beaucoup „a lot‟ (39) a. Il avait presque été beaucoup fait, mais on nous a communiqué d‟arrêter. Itexpl had almost been a-lot done, but PRONimpers uscl has told to stop A lot had almost been done, but we were told to stop. b. *Il avait beaucoup été presque fait, mais on nous a communiqué d‟arrêter. Beaucoup „a lot‟ > bien „well‟ (40) a. ?Lucie a beaucoup très bien produit à l‟usine. L. has a-lot very well produced at the factory. L. has produced a lot at the factory very well. b. *Lucie a très bien beaucoup produit à l‟usine. In conclusion, it seems to me that the data point to the direction that beaucoupQP is in the same position as beaucoup-AdvP, and that beaucoup-QP‟s location after tout simply cannot be seen. Peu The behaviour of peu is identical to that of beaucoup: as an adverb, it follows tout „everything‟ and it precedes bien „well‟. Although the hypothesis of beaucoup‟s location must be updated, the quantifier‟s frequency reading is certainly worth analysing further. In our case, supposing beaucoup is „dislocated‟ (i.e. moved) to its scope position, then it should move to the position in the hierarchy in which it is checked by its corresponding aspectual head - in other words, in the same position as beaucoup bare quantifier. (xiii) a. Fanny a complètement beaucoup caché de choses. Fanny has completely a lot hidden of things Fanny has completely hidden a lot of things b. *Fanny a beaucoup complètement caché de choses. However, this tempting hypothesis needs verification. Alternatively, one should check if this „dislocated‟ position, having a frequentative meaning, coincides with the location of frequentative quantificational adverbs (i.e. beaucoup meaning beaucoup de fois „a lot of times‟). 134 On the relative position of beaucoup, guère, peu, rien and trop in French Tout „everything‟ > peu „little‟ (41) a. Jacques a tout peu révisé. J. has everything little revised. J. revised everything little. b. *Jacques a peu tout révisé. Peu „little‟ > bien „well‟ (42) a. ?Il a peu très bien marché. He has little very well walked. He has little walked very well. b. *Il a très bien peu marché. Peu-QP follows presque „almost‟ and precedes bien „well‟. Presque „almost‟ together with peu is felt as somewhat unnatural by native speakers. Consequently, (43) has been given an ironic sense. 13 Presque „almost‟ > peu „little‟ (43) a. Il a presque été peu fait pour le Kosovo. Quelle honte! Itexpl has almost been little done for Kosovo. What a shame Little has almost been done for Kosovo. What a shame! b. *Il a peu été presque fait pour le Kosovo. Quelle honte! Bien „well‟ > peu „little‟ (44) a. Il a peu très bien mangé. He has little very well eaten He has eaten little very well b. *Il a très bien peu mangé. 14 Peu-QP cannot appear with complètement . The reading where peu modifies complètement has obviously been excluded. 13 14 Compare: Yves a presque peu fait aujourd’hui…quel garçon paresseux! 'Yves has almost done little today…what a lazy boy!‟ vs *Yves a peu presque fait aujourd’hui…quel garçon paresseux! However, just like its antonym beaucoup, peu can actually follow complètement if it is a „dislocated‟ quantifier: Fanny a complètement peu caché de choses „Fanny has completely hidden few things vs *Fanny a peu complètement caché de choses. 135 Sara Vecchiato (45) a. *Fanny a peu complètement changé dans le théâtre moderne Fanny has completely little changed in the theatre modern b. *Fanny a complètement peu changé dans le théâtre moderne. Rien 15 Rien seems to be located between presque and complètement . 16 Presque „almost‟ > rien „nothing‟ (46) a. Il n‟a presque été rien changé. Itexpl not cl has almost been nothing changed Nothing has almost been changed. b. *Il n‟a rien été presque changé. Rien „nothing‟ > complètement „completely‟ (47) a. Daniel n‟a rien complètement changé. Daniel notcl has nothing completely changed. Daniel hasn’t completely changed anything. b. *Daniel n‟a complètement rien changé. 5. Conclusion We have seen quite clear-cut data for the location of the examined quantifiers in the general hierarchy. As to their corresponding aspectual heads, one of them had already been identified in Cinque (1999) - namely, guère (Neg4). However, no reference is made there to the other four. Here, I would like to label the aspectual projections corresponding to beaucoup, peu, trop and rien, at least in a tentative way. 15 16 It could be argued that rien cannot be found after complètement because of semantic reasons. In fact, though complètement normally precedes tout, the antonym of rien, it can also follow it with some lexical choices (see Cinque (‟99), p. 10): (xiv) a. Tu as complètement tout refait? b. Tu as tout complètement refait? In (a) complètement modifies both the verb and the object (the natural answer could be Non! Je n’ai fait que la cuisine). In (b) the adverb modifies just the verb (the answer could be: Non! Je n’ai fait que la peinture). Being rien a negative quantifier, it would be impossible for it to be under the scope of a „completion‟ adverb. Then, it would exploit only the second option, where complètement is focussed on the verb. I leave the question open. Compare: Jean n‟a presque rien changé dans son milieu de travail. ‟Jean has changed almost nothing in his place of work‟ vs *Jean n‟a rien presque changé dans son milieu de travail. 136 On the relative position of beaucoup, guère, peu, rien and trop in French Of course, it will be necessary to check if aspectual heads corresponding to these adverbs exist in some languages. I would like to suggest ASPexcess for trop, ASPdegree for beaucoup/peu and ASPabsence for rien. Since beaucoup and peu seem to occupy the same specifier position, I would like to propose that their aspectual heads have two features, a positive one, [+ASPdegree], associated with beaucoup and a negative one, [-ASPdegree], associated with peu. [frankly MOODspeech-act [fortunately MOODevaluative [allegedly MOODevidential [probably MOODepistemic [once T(Past) [then T(Future) [perhaps MOODirrealis [necessarily MOODnecessity [possibly MODpossibility [usually ASPhabitual [again ASPrepetitive(I) [often ASPfrequentative(I) [intentionally MODvolitional [quickly ASPcelerative(I) [already T(Anterior) [no longer ASPterminative [still ASPcontinuative [always ASPperfect [just ASPretrospective [soon ASPproximative [briefly ASPdurative [characteristically (?) ASP generic/progressive [almost ASPprospective [guère Neg4 [trop ASP excess [rien ASP absence] [completely ASPSgCompletive(I) [tout ASPPlCompletive [beaucoup / peu ASPdegree [well Voice [fast/early ASPcelerative(II) [completely ASPSgCompletive(II) [again ASPrepetitive (II) [often ASPfrequentative (II)]]]]]]]]]]]]]]]]] 6. Appendix: there are no contradictions In order to find their locations, all the examined items were tested to determine their position with respect to every lower adverbs, starting from toujours (always). The data are consistent with the key sentences provided above. Trop-AdvP (49) a. Les films ont toujours trop impressionné David. b. *Les films ont trop toujours impressionné David. (50) a. On a récemment trop contesté le député b. *On a trop récemment contesté le député. (51) a. Ses louanges t‟ont bientôt trop flatté. b. *Ses louanges t‟ont trop bientôt flatté. (52) a. Les médecins ont longtemps trop ignoré la dépression. b. *Les médecins ont trop longtemps ignoré la dépression. (53) a. Il a brièvement trop utilisé son ordinateur, mais après tout a été arrangé. b. *Il a trop brièvement utilisé son ordinateur, mais après tout a été arrangé. (54) a. On a trop tout analysé. b. *On a tout trop analysé. 137 Sara Vecchiato (55) a. On a trop très bien dansé et on est crevés. b. *On a très bien trop sauté et on est crevés. (56) a. On a de nouveau trop poussé à l‟extrême le réacteur. b. *On a trop de nouveau poussé à l‟extrême le réacteur. (57) a. Les gardiens du zoo ont vite trop nourri les lions. b. *Les gardiens du zoo ont trop vite nourri les lions. (58) a. Tu l‟as rarement trop embarrassé par tes remarques b. *Tu l‟as trop rarement embarrassé par tes remarques Trop-QP: (59) a. David a toujours trop mangé. b. *David a trop toujours mangé. (60) a. On a récemment trop fait pour influencer sa décision. b. *On a trop récemment fait pour influencer sa décision, il nous en veut beaucoup. (61) a. ?Tu as bientôt trop acheté, il fallait que tu attendes d‟avoir l‟argent. b. *Tu as trop bientôt acheté, il fallait que tu attendes d‟avoir l‟argent. (62) a. Les médecins ont longtemps trop ignoré pour pouvoir soigner réellement. b. *Les médecins ont trop longtemps ignoré pour pouvoir soigner réellement. (63) a. ??Yvonne a trop très bien produit, je l‟envie, franchement. b. *Yvonne a très bien trop produit, je l‟envie, franchement. (64) a. *Sa famille n‟a guère toujours influencé Claude. b. Sa famille n‟a toujours guère influencé Claude (65) a. *Les spectateurs n‟ont guère récemment écouté son discours. b. Les spectateurs n‟ont récemment guère écouté son discours. (66) a. *Les gens âgés du village n‟ont guère bientôt apprécié la musique techno b. Les gens âgés du village n‟ont bientôt guère apprécié la musique techno (67) a. *Paul n‟a guère brièvement utilisé son ordinateur. b. Paul n‟a brièvement # guère utilisé son ordinateur. (68) a. *Je n‟ai guère longtemps soigné le jardin. b. Je n‟ai longtemps guère soigné le jardin. 138 On the relative position of beaucoup, guère, peu, rien and trop in French (69) a. Tom n‟a longuement guère examiné le problème. b. *Tom n‟a guère longuement examiné le problème. (70) a. *L‟incendie n‟a guère complètement touché la bibliothèque b. *L‟incendie n‟a complètement guère touché la bibliothèque (71) a. Les révolutionnaires n‟ont guère trop changé la morale publique. b. *Les révolutionnaires n‟ont trop guère changé la morale publique. (72) a. Tu n‟as guère tout envisagé. b. *Tu n‟as tout guère envisagé. (73) a. Gilles n‟a guère beaucoup estimé sa mère. b. *Gilles n‟a beaucoup guère estimé sa mère. (74) a. Les citoyens n‟ont guère bien compris son emprisonnement b. *Les citoyens n‟ont bien guère compris son emprisonnement (75) a. ?David n‟a guère vite affaibli le géant b. David n‟a vite guère affaibli le géant. (76) a. *Ils n‟ont guère de nouveau amélioré le jus de fruit. b. Ils n‟ont de nouveau guère amélioré le jus de fruit. (77) a. Cette thérapie n‟a rarement guère soigné un claustrophobe. b. Cette thérapie n‟a guère rarement soigné un claustrophobe. (78) a. *Sa famille n‟a guère toujours fait pour Claude b. Sa famille n‟a toujours guère fait pour Claude (79) a. *Les spectateurs n‟ont guère récemment vu, les lumières ne marchaient pas. b. Les spectateurs n‟ont récemment guère vu, les lumières ne marchaient pas. (80) a. *Les gens âgés du village n‟ont guère bientôt mangé, ils ont perdu l‟appétit. b. Les gens âgés du village n‟ont bientôt guère mangé, ils ont perdu l‟appétit. (81) a. *Je n‟ai guère longtemps acheté dans ce magasin, il était déguelasse, vraiement. b. Je n‟ai longtemps guère acheté dans ce magasin, il était déguelasse, vraiement. (82) a. *Tom n‟a longuement guère écouté, il en avait marre de la radio. b. *Tom n‟a guère longuement écouté, il en avait marre de la radio. (83) a. Il n‟a longuement été guère fait pour l‟hôpital, il est presque ruiné. b. *Il n‟a guère été longuement fait pour l‟hôpital, il est presque ruiné. (84) a. *L‟incendie n‟a guère complètement détruit, il en reste assez de choses. b. *L‟incendie n‟a complètement guère détruit, il en reste assez de choses 139 Sara Vecchiato (85) a. Les citoyens n‟ont guère très bien fait pour leur ville. b. *Les citoyens n‟ont très bien guère fait pour leur ville. Beaucoup-AdvP (86) a. Marguerite Duras a toujours beaucoup lu. b. *Marguerite Duras a beaucoup toujours lu. (87) a. On a récemment beaucoup voyagé en Angleterre. b. *On a beaucoup récemment voyagé en Angleterre. (88) a. Yves a bientôt beaucoup aimé sa fille adoptive. b. *Yves a beaucoup bientôt aimé sa fille adoptive. (89) a. Jeanne a longtemps beaucoup ignoré son fils. b. *Jeanne a beaucoup longtemps ignoré son fils. (90) a. Pascal a longuement beaucoup ignoré sa sœur. b. *Pascal a beaucoup longuement ignoré sa sœur. (91) a. Nous avons brièvement beaucoup souhaité la victoire de ce candidat. b. *Nous avons beaucoup brièvement souhaité la victoire de ce candidat. 17 (92) a. Nicolas a beaucoup trop surchargé ses collègues. b. *Nicolas a trop beaucoup surchargé ses collègues. (93) a. Gilles a de nouveau beaucoup travaillé. b. *Gilles a beaucoup de nouveau travaillé. (94) a. Jean a rarement beaucoup apprécié le travail des autres. b. *Jean a beaucoup rarement apprécié le travail des autres. (95) a. Vous avez vite beaucoup couru dehors. b. *Vous avez beaucoup vite couru dehors. Peu-AdvP (96) a. Les lecteurs moyens ont toujours peu apprécié son livre. b. *Les lecteurs moyens ont peu toujours apprécié son livre. 17 Beaucoup (a lot) seems to follow trop (too much). However, it is difficult to interpret the data, because trop and beaucoup are apparently incompatible from a semantic point of view: They can appear together in an active sentence because beaucoup is trop‟s specifier: “far too much”. Instead, in passive sentences, the higher beaucoup and trop have a frequency reading. Consequently, we are obliged to establish beaucoup‟s position with respect to trop by transitivity. 140 On the relative position of beaucoup, guère, peu, rien and trop in French (97) a. Gilles m‟a récemment peu écouté. b. *Gilles m‟a peu récemment écouté. (98) a. Ses camarades ont bientôt peu estimé Céline. b. *Ses camarades ont peu bientôt estimé Céline. (99) a. Les soldes ont longtemps peu attiré les gens. b. *Les soldes ont peu longtemps attiré les gens. (100) a. On a longuement peu laissé les animaux du zoo en liberté. b. *On a peu longuement laissé les animaux du zoo en liberté. (101) a. Les armes à feu ont brièvement peu effrayé Geneviève. b. *Les armes à feu ont peu brièvement effrayé Geneviève. (102) a. Les critiques musicaux ont presque peu considéré Jimi Hendrix b. *Les critiques musicaux ont peu presque considéré Jimi Hendrix (103) a. *On a peu trop exercé sa conscience morale. b. *On a trop peu exercé sa conscience morale. (104) a. On a de nouveau peu reconnu Agnès comme un auteur important. b. *On a peu de nouveau reconnu Agnès comme un auteur cela. (105) a. Ton patron t‟a vite peu chargé de travail. b. *Ton patron t‟a peu vite chargé de travail. (106) a. Le professeur a rarement peu blâmé Guitry. b. *Le professeur a peu rarement blâmé Guitry. Rien (107) a. Valentine n‟a rien très bien fait. b. *Valentine n‟a très bien rien fait. (108) a. Il n‟a rien été très bien fait. b. *Il n‟a très bien été rien fait. References Acquaviva, Paolo: “The Logical Form of Negative Concord”, in “Empirical Issues in formal syntax and semantics”, European Academic Publishers, Berne 1997. Alexiadou, Artemis: “Issues in the Syntax of Adverbs”, Phd Dissertation, Universität Potsdam, 1994. Bat-Zeev Shyldkrot, Hava: “Tout: Polysémie, Grammaticalisation et sens prototypique”, Langue Française, vol.107, Larousse, Paris 1995. Belletti, Adriana: “Generalised Verb Movement”, Rosenberg & Sellier, Torino, 1990. 141 Sara Vecchiato Borillo, Andrée.: “La quantification temporelle: durée et itérativité en français”, Cahiers de Grammaire, 11, 117-156, Université de Toulouse - Le Mirail, Toulouse 1986. Cinque, Guglielmo: “Adverbs and Functional Heads”, Oxford Studies in Comparative Syntax, Oxford University Press, New York , Oxford, 1999. Cinque, Guglielmo, “On leftward movement of tutto in Italian”, University of Venice Working Papers in Linguistics, Venice, 1992. Cinque, Guglielmo: “Types of A-bar Dependencies”, Linguistic Inquiry Monographs, 17, 56-97, The MIT Press, Cambridge, Massachussetts, 1990. Cinque, Guglielmo: “Bare Quantifiers, Quantified NPs and the Notion of Operators at SStructure”, Rivista di Grammatica Generativa, 11, 33-63, Unipress, Padova, 1985. Giusti, Giuliana: “The Categorial Status of Determiners”, in “The New Comparative Syntax”, 95123, Longman, London and New York, 1997. Haegeman, Liliane: “Introduction to Government and Binding Theory”, Blackwell Publishers, UK, 1994. Kayne, Richard S.: “The Antisymmetry of Syntax”, Linguistic Inquiry Monographs, 25, The MIT Press, Cambridge, Massachusetts, 1994. Kayne, Richard S.: “French Syntax - The Transformational Cycle”, 1-65, The MIT Press, Cambridge, 1975. Laenzlinger, Christopher: “Comparative Studies in Word Order Variations”, Thèse de doctorat d‟État, Université de Genève, 1996. Laenzlinger, Christopher: “Adverb Syntax and Phrase Structure”, in “Configurations” ed. AnnaMaria di Sciullo, 99-127. Somerville: Cascadilla Press, 1996. Lonzi, Lidia: “Which Adverbs in SPEC, VP?” Rivista di Grammatica Generativa, 15, 141-160, Unipress, Padova, 1990. Nølke, Henning: “Recherches sur les adverbes: bref aperçu historique des travaux de classification”, Langue Française, 88, 117-123, Larousse, Paris, 1990. Obenauer, Hans-Georg: “Aspects de la syntaxe A-barre”, Thèse de doctorat d‟ Etat, Université de Paris VIII, 1994. Pollock, Jean-Yves: “Verb Movement, UG and the Structure of IP”, Linguistic Inquiry, 20, 365424, MIT Press Cambridge Mass, 1989. Robert, Paul: “Le Nouveau Petit Robert”, Dictionnaires Le Robert - Paris, 1994. Sportiche, Dominique: “Adjuncts and Adjunction”, UCLA, 1994. Togeby, Knud: “Grammaire française, Vol. II: Les formes personnelles du verbe”; “Vol. IV: Les mots invariables”, Etudes Romanes de l‟Université de Copenhague, Akademisk Forlag, Copenhague, 1984. Travis, Lisa: „The Syntax of Adverbs‟, McGill University Working Papers in Linguistics, Montréal, 1988. Vecchiato, Sara: “A Note on Longtemps-Longuement”, Annali di Ca’ Foscari, Venezia, 1999. Vinet, Marie-Thérèse: “Adverbial Quantifiers, Negation and Stress Rules Effects”, University of Venice Working Papers in Linguistics, 5, 115-138, 1995. Zanuttini, Raffaella: “Negation and Clausal Structure”, Oxford Studies in Comparative Syntax, Oxford University Press, New York / Oxford, 1997. 142 ANCORA AND ADDITIVE WORDS Milena VEGNADUZZO 0. Introduction The classification and analysis of the parts of speech (or may be better, parts of sentence) has produced a long tradition of studies that have adopted different criteria (morphological, semantic, distributional) to define them. It is well known that in this kind of research the class of adverbs received less attention than others in traditional grammar. It is the least homogeneous category, the one to which all those words that do not find place in other categories are assigned. Adverbs are traditionally classified, rather approximately, according to their meaning: e.g., we have time adverbs, mood adverbs, place adverbs. In this paper I will deal in particular with the Italian adverb ancora, which shows interesting properties because of the multiplicity of contexts in which it can be used. I will discuss its main readings in order to see whether it is possible to reduce this multiplicity of values to a general, basic, abstract meaning. Then I will extend the generalization obtained for ancora to other adverbs in order to identify a group of adverbs sharing the same properties. 1. Readings of “ancora” The adverb ancora is used in many situations to express different meanings. It can have different values such as continuative, iterative, quantitative, additive, comparative or adversative. This could be considered a case of lexical ambiguity. However, I will argue that ancora has a single lexical entry allowing for different construals, depending upon the context where the adverb is inserted. We can say that the meaning that ancora acquires is determined by the interaction of its basic meaning with the argument structure of the verb, and by the combination of the semantic features of the verb and its object. In particular, the following parameters are relevant: a) the aspectual class of the VP (activity, accomplishment, state, achievement, following Vendler [1967] and Verkuyl [1989]), with special reference to the tel- 143 Milena Vegnaduzzo 1 ic/atelic distinction; b) tense; c) grammatical aspect; d) the presence or absence of the direct complement (definite or indefinite object, count or mass noun); e) the type of event (unique or non-unique). On the basis of these parameters I will characterize the contexts where each reading of ancora is allowed. Since in English each value we are concerned with is associated with a different lexical item, I will use these words as labels to easily recognize each construal of ancora. 1.1. The continuative reading 2 When used with continuative meaning , (i.e., in the sense of English “still”), ancora implies an interval at which the situation is in progress, i.e. it presupposes that the situation expressed by the verb was already in progress at the reference time. Thus ancora has scope on the state of affairs expressed by the verb. This reading is licensed under the following conditions: 3 A. if the predicate is an adjective it is alterable (ex. 1-2) B. if the predicate is a verb without complements it is durative (ex. 3-12) C. the verb is atelic (ex. 3-5) D. the verb has imperfective aspect (ex. 3-8) 1 2 3 I consider telic verbs those oriented to reach a definite natural terminal point that once reached involves a change in the situation described; I consider atelic verbs those that lack such intrinsic terminal point, go on without interruptions and can be stopped at any time. I prefer the term continuative with respect to the term traditionally used durative, because it does not only refer to the length of time but emphasizes also that the action was in progress previously, which is exactly what ancora means in this reading. Then I will use the term durative only to make reference to the action of the verb while I will reserve the term continuative to express the value of the adverb deriving from its basic meaning associated with the action of the verb. With the expression «alterable predicates» I mean durative predicates whose denotation lasts for a determined period of time after which it is subject to a change of state. When associated to this predicates ancora states the persistence of its denotation against the supposition it already came to the change of state. They are alterable predicates adjective as: young, living, sick, raw, unripe, short, eatible but not they antonyms old, dead, healthy cooked, ripe, tall, out of date. So for example «to be young» is a condition that lasts for a period of time and then undergoes a change of state because people necessarily, become old. The direction of change goes from «to be young» to «to be old» but the reverse does not hold. So the former predicate is alterable, while the latter is not. 144 Ancora and additive words 4 Examples : (1) Maria è ancora giovane. „Mary is still young.‟ (2) *Maria è ancora vecchia. „Mary is still old.‟ (3) Maria nuota ancora. „Mary is still swimming.‟ (4) Maria stava ancora nuotando. „Mary was still swimming.‟ (5) Maria nuotava ancora quando chiusero la piscina. „Mary was still swimming when the swimming pool was closed.‟ (6) #Maria ha/aveva nuotato ancora. <Mary has/had swum again.> (7) #Maria nuotò ancora. „Mary swam again.‟ 5 (8) #Maria nuoterà/avrà nuotato ancora . „Mary will swim/will have swum again.‟ (9) #Maria parte ancora. „Mary leaves again.‟ (10) *Maria muore ancora. *„Mary still dies.‟ (11) #Luca legge ancora un romanzo. „Luca reads another novel.‟ 4 5 The symbol # indicates that the sentence is not acceptable in the reading of ancora we are dealing with (in the present case the continuative one), but it is with other readings; the symbol * indicates ungrammaticality on any reading. I will give in the English translation the exact interpretation of the Italian sentence. Notice that the continuative reading is available if the Future has progressive value: «Quando arriverai Maria nuoterà ancora (= starà ancora nuotando)», „When you will arrive Mary will be still swimming.‟ 145 Milena Vegnaduzzo (12) Luca legge ancora romanzi. „Luca still reads novels.‟ In the last case the still reading is allowed because the bare plural is not a discret entity and therefore the event is atelic (cfr. Tovena). The association of ancora with stative verbs is impossible when they express generic predications such as geographic locations or when they refer to classes or species: 6 (13) #Chiasso è ancora in Italia . „Chiasso is still in Italy.‟ (14) *Le tigri sono ancora dei felini. *„Tigers are still feline.‟ With respect to stative verbs that refer to specific situations, the association with ancora is allowed only if it is possible an alteration of the described state of affairs, even if such alteration means the cancellation of the existence conditions of the state of affairs itself: (15) Mio figlio crede ancora all‟Uomo Nero. „My son still believes in the Blak Man.‟ (16) Stefano fa ancora l‟avvocato. „Steven is still a lawyer.‟ (17) Nonostante l‟età mio nonno è ancora in gamba. „Despite his age my grandpa is still fit.‟ 1.2. The iterative and quantitative readings Besides the continuation of an action, ancora can be used to express the repetition of an event (cfr. Tovena). In this case the fundamental element is the presence of a well-defined termination point: in fact the repetition shows the action in its single occurrences and therefore as completed. From this it follows that only telic predicates can express repetition. As suggested by Tovena: a) if the repetition shows up as iteration of the event, ancora corresponds to the English “again” and it has, then, iterative meaning; b) if it shows up as alteration of a participant of the event, ancora corresponds to English “one more”, “another” and it has quantitative meaning. Therefore, I claim, in the first case, ancora has scope over the event, in the second case over an argument. Iterative and quantitative meaning of ancora represent then 6 This sentence can only have a spatial interpretation in a context as: „Chiasso is still in Italy but Lugano is already in Switzerland.‟ which refers to a scalar relation of the cities. (See König 1977 p. 184). 146 Ancora and additive words two sides of a unique concept, i.e. the concept of repetition, but are licensed under different contextual conditions. The iterative reading (the “again” construal) requires: E. telic verbs (ex. 18-19) F. non-unique event (ex. 19-20) G. perfective aspect (ex. 18-26) H. specific object (ex. 19, 21-23) Examples4: 7 (18) #Maria nuota ancora . „Mary is still swimming.‟ (19) Maria legge / ha letto ancora la lettera. „Mary reads / has read the letter again.‟ non-unique event/definite object (20) ?*Maria scrive ancora la lettera7. „Mary writes the letter again.‟ unique event (21) #Maria legge / ha letto ancora una lettera. non-unique event/indefinte object „Mary reads / has read another/one more letter.‟ (22) Maria legge / ha letto ancora una lettera di sua zia. non-unique event / specific object „Mary reads / has read her aunt‟s letter again.‟ „Mary reads / has read another letter of her aunt.‟ (23) Maria lesse ancora i romanzi. „Mary read the novels again.‟ (24) Maria verrà ancora a trovarti. „Mary will came to see you again.‟ (25) Maria non ti deluderà ancora. „Mary will not disappoint you again.‟ 7 The iterative reading is possible if the Present Tense has progressive value: «Maria sta nuotando ancora (= di nuovo)». „Mary is swimming again.‟ «Maria sta scrivendo ancora (= di nuovo) la lettera». „Mary is writing the letter again.‟ 147 Milena Vegnaduzzo (26) Maria mangia ancora la zuppa (e non il riso). „Mary eats soup again (and not rice).‟ „Mary still eats soup (and not rice).‟ The sentence in (22) can have both the iterative interpretation (“again”) or the quantitative one (“another”). This ambiguity is due to the different scope of ancora: in the former case ancora takes scope over the event, in the latter it takes scope over the NP object. Same explanation for the sentence in (26): ancora can take scope either over the state of affairs (“still” reading) or over the event (“again” reading). In this case, as suggested by Tovena, the iterative reading follows from the fact that the definite NP is coerced into a „type‟ reading which transforms a unique event into a non-unique event. The quantitative reading “one more”, “another” has more applicability restrictions because it is triggered only by the presence of a direct object that therefore constitutes the element in the scope of ancora. The licensing conditions are the following: I. J. K. L. presence of a NP object presence of an event (unique or non-unique), (ex. 27-28) telic verb possibility of alteration of the NP, i.e. the substitution of the participant with another one of the same type M. indefinite NP object (ex. 29). Examples4: (27) Maria scrive / ha scritto ancora una lettera. „Mary writes / has written one more/another letter.‟ (28) Maria legge / ha scritto ancora una lettera. „Mary reads / has read one more / another letter.‟ (29) #Maria ha letto ancora la lettera. „Mary has read the letter again.‟ (30) Maria mangia ancora cioccolata. „Mary eats more chocolate.‟ (31) Maria mangia ancora un piatto di lenticchie. „Mary eats one more / another plate of lentils.‟ (32) Maria ha nuotato ancora per due ore. „Mary has swum for two more hours.‟ The last two examples suggest interesting observations. In (31) ancora can only have quantitative reading and not iterative (cfr. example (22)). I think this is 148 Ancora and additive words probably due to the quantificational properties of the NP “un piatto” which behaves as a unit of measure. In (32) ancora can have only a quantitative and not a continuative meaning as one might think. It is true that this sentence too implies a continuation of the action expressed by the verb, but such continuation is different from the one seen in (1) because it goes, so to speak, in the opposite direction, i.e. starting from a reference point (not necessarily coincident with the utterance time) towards the future. The semantic difference between the two sentences is mirrored in the syntactic structure: in (32) ancora modifies the Adverbial Phrase (AdvP) and not the verb, as the comparison with the English translation shows. 1.3. The comparative reading Another meaning attributed to ancora by traditional grammar is that of comparative adverb, because it modifies adjectives or adverbs in the comparative degree. With this reading ancora corresponds to English “even”: (33) Maria è ancora più giovane di Piera. „Mary is even younger than Piera.‟ (34) Maria è ancora più vecchia di Antonio. „Mary is even older than Anthony.‟ (35) Maria mangia ancora più lentamente di Antonio. „Mary eats even more slowly than Anthony.‟ In this case the element under the scope of ancora is the comparative adjective or adverb. Nevertheless, the comparative sense does not seem to represent a really distinct construal of ancora, because it can be easily reduced to another value, i.e. the quantitative meaning. Actually, in the examples (33)-(35) ancora can be interpreted as indicating a further degree, a greater extent of some property expressed by the adjective or the adverb. We could say that the term “comparative” indicates the function performed by ancora in the sentence more than its meaning, which can be rather considered quantitative. I will come back to this point later. In the “comparative” construal ancora can modify alterable or non-alterable predicates. On the other hand, when it modifies superlatives, ancora has only the continuative meaning and therefore it is sensitive to the opposition alterable/non-alterable: (36) Maria è ancora bellissima. „Mary is still very beautiful.‟ (37) *Maria è ancora vecchissima. *‟Mary is still very old.‟ 149 Milena Vegnaduzzo 1.4. The additive reading Finally ancora can be used also as an additive conjunction (“also”, see It. “anche”) and in Old Italian also as an adversative conjunction (“nevertheless”, “however”, see It. “tuttavia”). In the first case the origin of such use should be probably led back to a common etymological ancestor of these two forms (hinc ad horam or a reconstructed root *anc), which specialized later in the two different forms ancora and anche. Even if it is not clear which form derives from which, it is undeniable that for some time in Old Italian these two forms were used in a very flexible way, appearing in the same contexts. They later specialized and undertook the modern meaning. Nevertheless, in modern Italian traces persist of this overlap. If, in general, they determine different readings of the same sentence and are not interchangeable (see the example in (38)), in some contexts they are (39-42), while in others they show parallelisms (43-44): (38) a. Oggi esistono anche i computer portatili. „Today we have also laptop computers.‟ b. Oggi esistono ancora i computer portatili. „Today we still have laptop computers.‟ (39) a. Ieri e ancora oggi. „Yesterday and still today.‟ b. Ieri e anche oggi. „Yesterday and also today.‟ (40) a. Ancora di più. „Still more.‟ b. Anche di più. „Also more.‟ (41) a. Viene anche Antonio. „Also Anthony is coming.‟ b. C‟erano Luca, Paolo, Giovanni e ancora Carlo, Roberto, Alfredo. „There were Luca, Paul; John and still Karl, Robert, Alfred.‟ (42) a. Gli ho anche detto. „I also told him.‟ b. Gli ho ancor detto. (obsolete) „I still told him.‟ (43) a. Anche se fosse stanco. (= irreal) „Even if it was tired.‟ 150 Ancora and additive words b. Ancorché fosse stanco. (= real). „Although he was tired.‟ (44) a. Neanche. „Neg-anche.’ b. Neancora. (= non ancora, dialectal) „Neg-ancora.’ According to Tovena‟s analysis [1996] the overlap of meanings of ancora and anche takes place in the presence of coordinated structures: in this case a new participant is explicitly introduced into the action, unlike what happens in the iterative reading “again” (where there is only one participant), or in the quantitative reading “one more” (where the participant in the predication is substituted for by another of the same type). 1.5. The adversative reading Finally, ancora in Old Italian was used with an adversative meaning: (45) Se voi mi concedete ch‟io vada, io v‟andrò e se voi non me lo concedete ancora andrò. (Boccaccio, I, 270) „If you let me go, I will go and if you do not, still I will go.‟ It must be noticed that in Old Italian the adverb tuttavia, indicating the persistence of a situation or condition in the present or in the past, could be interchanged with ancora: (46) Essendo il freddo grande e nevicando tuttavia forte. (Boccaccio) „Being really cold and snowing still a lot.‟ (47) Dall‟una e dall‟altra di quelle terre correvano e corrono tuttavia strade e stradette più o meno ripide o piane. (Manzoni) „From one and another of these grounds run in the past and still run streets and little streets more or less steep or flat.‟ An interesting parallelism can be observed in the history of English: according to the Oxford English Dictionary the adversative meaning displayed by still is probably derived from the temporal meaning: (48) Nothing can make such a room healthy. Ventilation would improve it, but still it would be unhealthy. (Flor. Nightingale, Nursing, 1861, 22) In this sense, if we suppose the same thing for ancora, the Italian adverb would have under its scope the state of affairs expressed by the verb in the adversative reading as well. 151 Milena Vegnaduzzo To summarize, we have seen that each reading exhibited by ancora is determined by different licensing conditions. We also noted that in each construal ancora takes scope over a different entity. We argue, then, that the cases of multiple reading of a sentence are due to scope differences. 2. Uniform classification So far we have seen that ancora has different readings according to the context in which it is used. I will now try to show that such polysemy is only apparent, i.e., it represents the manifestation at the contextual level of a unique more general and abstract meaning that interplays with the features of the elements of the syntactic context. Many are the proposals of a unitary classification of the meanings of ancora, or of its counterparts in different languages. The majority of these theories makes use of the concepts of assertion and presupposition: when ancora is applied to a sentence F, the presupposition holds that F is true for the entire timespan at which the situation is in progress. This induces the persistence of a situation started in the past until the moment expressed by the tense of the sentence where it is contained. Such presupposition does not say anything about the state of affairs after utterance time, which therefore remains undetermined. Starting from this concept, König [1977] defines the temporal and non-temporal (spatial, comparative, adversative) meanings of the German noch as manifestations of a basic scalar meaning because both interpretations imply the selection of some points in time or entities of different type fixed in an ordered relation. But such scalar meaning cannot account for the additive meaning of noch. Consider (49): (49) Ich kenne nóch einen Mann, der flieend Russisch spricht. „I know another person who speaks Russian fluently.‟ Here the additive reading of noch should be reduced to the non-temporal use. Anyway, it cannot be accounted for within the framework of this analysis because it does not introduce a set of ordered elements. In Barker [1991] stillF means that the state of affairs in F has persisted, against some conditions. Still takes as argument the constituent to whose denotation the concept of persistence can somehow be applied. This happens directly in the temporal case, by analogy in the other ones. Then, in general, what unifies the different construals of still is an analogical extension stemming from the prototypical case, i.e., the temporal one. Nef [1981] proposes to reduce temporal, iterative and quantitative meanings of French encore to a “schéma formel d‟implicature” which is interpreted in a different way for each case. I do not enter into the details of his analysis. It is enough to say that he too keeps distinct temporal cases as “Paul est encore malade” („Paul is 152 Ancora and additive words still ill‟) from quantitative cases as “Marie a mangé encore deux gâteaux” („Mary ate two more gâteaux‟). Also Tovena [1996] thinks that the different readings of ancora depend on its sensitivity to the context, in particular on the type of event which it is applied to. For Tovena ancora is a binary operator schematically represented as ANCORA (A, B). The basic meaning of ancora is identified, then, in two components: a mapping of the argument A instantiated by a proposition (e.g. “to be young”) with the argument B which contains an “eventuality identifier” (the information provided by the verb) and an effect of continuation. Such continuation is realized in different ways according to the entity which it is applied to: if the event lacks a terminal point such continuation is an extension in time, if the event has a definite terminal point the continuation is a repetition (iterative or quantitative construal). But I think that also this explanation implies an ordering of entities as in König‟s proposal because this “effect of continuance” seems to suggest the idea of “ before” and “after”, that is a continuance in a linear dimension. Therefore it can hardly account for an example like (49). In all these theories, then, the quantitative meaning of ancora (or whatever counterpart in other languages) is the most difficult to insert in a unitary classification. In the present work I will put forward an explanation that can account for the temporal value and the non-temporal one (and all the others) at the same time. To do this I think it is worth starting from an approach opposite to that used by the authors quoted above. On the basis of the fact that the additive meaning should be connected to the others, it might be worth trying the opposite path, namely to assume the additive meaning as basic and assign a component of this type to all other meanings. Such idea was already implicitly suggested by Doherthy [1973] (who says that “a mere additive function […] lies at the bottom of all the various nochs”), and was afterwards assumed by König e Stark [1987] in terms of a basic additive meaning. Following the analysis of van der Auwera [1993] for German noch, I will show that each construal of Italian ancora can be expressed as addition of an entity to another entity of the same type or of a different type, if this is possible according to the features of the arguments taking part in the predication. But to consider the additive reading the basic component of the meaning of ancora does not mean that we set aside the presuppositional component that previous works reveal indispensable. I think, therefore, a unitary theory of the meanings of ancora should integrate in some way both these elements, presupposition and additive component. The additive component is clearly visible in the quantitative reading of ancora. Consider the following sentence: (50) t = Prendo ancora un biscotto. „I will take another / one more biscuit.‟ 153 Milena Vegnaduzzo In this case ancora indicates that some entity x (a biscuit) introduced at the utterance time t is added to an entity x‟ previously introduced at the moment t.‟ This sentence presupposes then a sentence like (51) or at least an act of taking a biscuit: (51) t‟ = Prendo un biscotto. „I will take a biscuit.‟ The quantitative reading conveys, therefore, the presupposition of existence of another entity of the same type. At this point the component of meaning I called addition means that the entity x is added to the one presupposed and both constitutes the set of elements that are the subject of the predication. It should be noted that at this level of the analysis it is not important to know the type of participant targeted by ancora, i.e. whether it is a count or a mass noun: this does not imply differences in the interpretation of what we are testing here, the presence of the additive feature. At this point the hypothesis I formulated for the quantitative reading can account for all other construals of the adverb. We only need to substitute for the discourse referent x the element under the scope of ancora. I will show how this works for each reading I have discussed above. Recall the definition of the continuative reading where ancora implies an interval at which the situation is in progress (52) Maria abita ancora qua. „Mary still lives here.‟ In this reading ancora says that the state holding at time t, specified by the ad8 verb, already held in the immediately preceding interval , and that such state persists until the time t, i.e., the state holding in the past is still in progress at the reference time t. It is important to notice that such persistence holds for each moment of time of the interval including the final one but not the first one. It is evident that within an interval the initial point, i.e., the starting point of the state, cannot be considered continuation of itself. The starting point represents the change of state from a “negative” situation, in which the state does not hold, to a positive one in which the state comes to existence. As an example, consider the following: the sentence (53) presupposes (54): (53) Maria vive ancora qua. „Maria still lives here.‟ (54) Maria viveva qua precedentemente. „Mary has been living here previously.‟ 8 I do not consider the internal length of such interval that might be a second or many years. 154 Ancora and additive words In (53) ancora conveys an interval projecting backwards, from utterance time to an undetermined moment in the past, the starting point of the interval, i.e., the point at which Maria went in the house for the first time. Therefore such moment indicates the change from a phase in which Maria did not live in the house, and it is expressed lexically by another adverb, the adverb già (“already”): (55) Maria già vive qua. „Mary already lives here.‟ Applying to this reading my hypothesis we have that the state holding at time t already held at a time t‟ preceding t. Because of the fact that the state persists unchanged, going through the interval up to the moment t, this condition holds at any moment of the interval but the initial one. This follows directly from the fact the continuative reading of ancora is allowed only with durative verbs. Then it is not necessary to postulate a condition of contiguity of the moments t‟ and t as suggested by van der Auwera since the state is homogeneous. On the basis of the additive analysis the entity x in the continuative reading stands for a state, so to speak, added to a previous state of the same type: addition means that we have something more of a same state, a further moment in which the state of living persists. I will make this explicit with a graphic representation: ---------------------------------> initial point final point già abitava x‟(t‟) ancora x(t) In the picture the continuous line indicates the interval at which the situation (except the initial point lexicalized by già) persists and it is made up of infinitesimal states of living; x(t), the state identified by ancora, coincides with the final point, while x‟(t‟) can represent any of the infinitesimal states inside the interval. So the living of Mary in that place is the sum of all the infinitesimal states of the interval including the final one but not the first one. The iterative reading of ancora presupposes the existence of an event x‟ holding at t‟ previous to the event x holding at t. In this case, too, it is not necessary to postulate a condition of non-contiguity of the elements x and x‟, as van der Auwera does, because such condition is directly inferred by the features of the element under the scope of ancora: (56) Vieni a trovarmi ancora, Maria. „Come to see me again, Mary.‟ In fact in this case the variable x does not stand for a state but for an event. As I noticed above, the event is telic and reaches a terminal point, while the state is atelic and can have infinite duration. Events are expressed by verbs of accomplishment or 155 Milena Vegnaduzzo achievement that show the situation as completed. The visits of Mary are the set of all single events repeated. This concept is shown in the following scheme: event 1 event 2 ------------------------------------------------------------------------------------> initial point final point x‟(t‟) initial point final point x(t) In the graphic the black dots represent two events of the same type. Such events have their specific internal duration represented by the triangles indicating the beginning and the end of the event. In this way we can clearly see that we are dealing with two different events that take place in the temporal axis (the dotted line from left to right). As for the so-called comparative reading the explanation of the meaning of ancora in terms of addition confirms the hypothesis that the comparative reading could be considered a case of quantitative construal. (57) Pietro è alto, ma Paolo è ancora più alto. „Peter is tall, but Paul is even taller.‟ In this case x stands for the comparative function “essere più alto” („to be taller‟) and clearly presupposes another less tall element. Therefore we can say that the quantitative reading of ancora depends upon the element under its scope: a) when it is a NP (Noun Phrase) ancora has, strictly speaking, quantitative construal, b) when it is a degree adjective or adverb ancora has a “comparative” construal. Consider now the adversative reading. I said above that the Oxford English Dictionary suggests for English the derivation from the temporal meaning. In my analysis I will adopt this point of view and therefore, also in this case, the entity substituted for x will be a state that persists against some contrary conditions. (58) Se voi mi concedete ch‟io vada, io v‟andrò e se voi non me lo concedete ancora andrò. (Boccaccio, I, 270) „If you let me go, I will go and if you do not let me still I will go.‟ When ancora is used as conjunction (“anche”) the theory makes the right prediction again because, even if in this case a new participant takes part in the predication and x and x‟ are elements of different type, the element x (a nursery) pre- 156 Ancora and additive words suppose the existence of another element. All these elements constitute the set of charities made by Mary. (59) Luisa è una grande benefattrice: nel 1972 ha costruito un ospedale e ancora una casa di riposo nel 1975. „Louise is a great benefactress: she built an hospital in 1972 and also a nursery in 1975.‟ To summarize, with this analysis I propose that the basic meaning of ancora conveys a presupposition and is essentially additive. The hypothesis proposed puts together these two components and is sufficiently general to account for all readings of ancora simply by substituting for the variable x the entity focalized by ancora, i.e. the entity falling under its scope. This hypothesis does not imply an ordering of the entities but refer to them as a set of elements which constitute the argument of the predication. Therefore it can account for the example in (49) repeated below: (49) Ich kenne nóch einen Mann, der flieend russisch spricht. „I know another man who speaks fluent Russian.‟ So, in this case the presupposition is that I know (at least) a man who speaks fluent russian; to this man (corresponding to x‟) the man (corresponding to x) referred to in (49) is added. These men constitute, therefore, the set of men who speak fluent Russian that I know. In this analysis we also saw that the application of ancora to any entity takes place in the temporal dimension which can be directly or indirectly implicated. The intuition that the use of ancora was related to the temporal dimension induced grammars to classify ancora as a time adverb in the same group of deictic adverbs such as ieri “yesterday”, oggi “today”, domani “tomorrow”, adesso “now”. If ancora has temporal value this must be understood not in the deictic sense but as relation of two not necessarily contiguous subsequent moments. By taking as fundamental the additive meaning this new classification of the readings of ancora can easily and economically account for all interpretations. The hypothesis bases its functionality on the element under the scope of ancora: once established that it is different for each reading we do not need more specifications. Each reading follows directly from the element under the scope and the aktionsart of the sentence as resumed in the table below: 157 Milena Vegnaduzzo ANCORA processes /states AdjP events continuative adversative iterative NP quantitative comparative additive The hypothesis proposed for ancora might work also for all adverbs corresponding to ancora in other European languages. As a matter of fact the counterparts of ancora in Spanish (todavía/aún), English (still), French (encore) and German (noch), all have at least three different readings: continuative, adversative, and comparative. However, other construals, such as the iterative one, can be recovered in the previous historical stages of these languages. This is the case for Spanish and English, while French and German associate to encore and noch a large cluster of readings. Therefore I would like to suggest that the meaning of ancora can be decomposed in a modular way: each language associates a higher or lower number of readings to a unique basic adverb with the meaning of ancora, while the other ones are attributed to independent lexical items. I will leave this point open to further research. 3. Application of the theory Consider now whether the hypothesis proposed for ancora can be extended to other adverbs, i.e., whether there exist other adverbs whose meaning is given by the same components as for ancora: presupposition and addition. Going back to the readings ancora can have in Italian, we noticed that in other languages some of these are expressed by independent lexical items: so in English the readings of ancora that we have considered here are expressed by “still”, “again”, “one more”, “another”, “even”, “nevertheless”, “also”. I claim that my hypothesis can account for the interpretation of these adverbs. Since the meaning of each of these adverbs corresponds to one of the possible meanings of ancora, it should follow that their meaning could be understood in terms of presupposition and addition. Consider, then, the following group of English adverbs: again, one more, another, even and the conjunctions also and nevertheless. (60) It is raining again. (61) She read the letter again. (62) She came back again. (63) She read one more/another letter. 158 Ancora and additive words (64) She ate more chocolate. (65) She drank one more/another beer. (66) John is even more intelligent than Bob. (67) She won the first and also the second prize. In these sentences the hypothesis is easy to apply. In the equivalent Italian sentences these adverbs would be all translated by ancora. Consider now an example in which my proposal seems to apply less straightforwardly and, more interestingly, the adverb could not be translated by ancora: (68) Even a child could do so. „Perfino un bambino potrebbe farlo.‟ In (68) even would be translated as “perfino”. The element under the scope of even is “a child” corresponding therefore to the entity x(t) of the hypothesis: the presupposition is that everybody is able to do that thing. However, it should be noted that Italian, too, has independent lexical items for each of the meanings exhibited by ancora, i.e., di nuovo (“again”), un altro (“one more”), perfino (“even”), anche (“also”), tuttavia (“nevertheless”), più (“more”), sempre (“always”). Consider briefly the two last adverbs, which we did not mention before. Syntactically, the Italian adverb più, can be an indefinite adjective (without article and plural) when it modifies a noun (69.a), an adverb when it modifies adjectives, adverbs or verbs (69.b), or a substantivized pronoun (neuter and singular) as in (69.c): (69) a. Dammi più pesche che fragole. „Give me more peaches than strawberries.‟ b. Maria è più intelligente di Luca. „Mary is more intelligent than Luca.‟ c. Il più è fatto. „The most of it is done.‟ Semantically, più is a quantifier and can form comparative clauses. It is interesting to notice that a) if used adverbially (ex. 70-71) più has always comparative meaning, b), if used as an adjective, it has comparative meaning when it modifies mass nouns (ex.72), but when it modifies count nouns it has comparative meaning only if there is a second element of comparison in the textual or extratextual context (ex. 73). Otherwise, più works as non-comparative quantifier expressing an indefinite quantity equivalent to adjective such as parecchi, molti, svariati (“quite a lot”, “many”, “various”) (ex. 74): (70) Marco è più alto di Luca. „Mark is taller than Luca.‟ 159 Milena Vegnaduzzo (71) Marco mangia più lentamente di Luca. „Mark eats more slowly than Luca.‟ (72) Giovanni ha più esperienza. „John has more experience.‟ (73) Marco ha più libri di Luca. „Mark has more books than Luca.‟ (74) Ho mangiato più volte in questo ristorante. „I have eaten many times in this restaurant.‟ So più exhibits a double function: it is a comparative or an additive quantifier. In both cases the behaviour of più is predicted by my hypothesis, since in the first case it behaves exactly as “comparative” ancora, in the latter as quantitative ancora. The adverb sempre as well, when expressing a continuation limited to the present time, can replace ancora and therefore it can be accounted for in my perspective: (75) Pronto? – Sono sempre io. „Hallo? – It is me again.‟ (76) Sei sempre in collera con me? „Are you still angry at me?‟ 4. Particular cases Until now we have seen that the general meaning of ancora is typical of other adverbs and conjunctions. Such hypothesis, proposed to give an economical and global account of the readings of a single adverb, acquires general value, explaining the behaviour of a determined group of adverbs and conjunctions. Ancora then represents a kind of “hyperonym” of such class of words. But the most interesting thing is that the hypothesis works not only for the adverbs that are “hyponimous” of ancora but also for independent adverbs such as even/perfino (68), più and for two other conjunctions, as I will try to show now. It seems plausible to think that also the conjunctions e (“and”) and con (“with”) should be interpreted in terms of presupposition and addition. Conjunction e has basically the logical meaning of AND combining two syntactic categories which in the sentence have the same function, including sentences: (77) Tu e io. „You and me.‟ (78) Bello e buono. „Nice and good.‟ 160 Ancora and additive words (79) Chi va piano, va sano e lontano. „Who goes slowly, goes healthy and far.‟ (80) Canta e balla tutte le sere. „She sings and dances every night.‟ In these sentences we consider only the second element: it is evident that it presupposes the existence of another element to which it is associated. Evidences of this hypothesis come from Old Italian where e meant anche (“also”), and from mathematics, where e is used in the sense of the additional operator più: “2 + 5 = 7” is read “2 e 5, 7”. We have already shown that anche and più satisfy the hypothesis, therefore, by transitivity, the conjunction e does as well. A similar argument holds for the conjunction con when used to express companionship and union: (81) Vengo con te. „I will come with you.‟ (82) Cioccolata con panna. „Chocolate with cream.‟ In this case too the second element of the sentence presupposes the existence of another element which it combines with to express the sense of the sentence. I claimed that adverbs and conjunctions can be uniformly accounted for under my hypothesis, i.e., they have the same basic meaning. But conjunctions differ from adverbs: if it is true that both conjunctions and adverbs lack inflection and agreement, only the former and not the latter are unstressed and occur in fixed positions. Therefore their behaviour differs from that of adverbs: while adverbs are essentially modifiers and represent a functional category, conjunctions set up a relation of subordination or coordination between functional categories. Consider, then, how this difference of function affects the theory put forward in this paper. We saw that the adverbs my hypothesis applies to associate an assertion with the sentence that they presuppose: they state the persistence of a determined action against the opposite expectation of the hearer. In the example considered in (52) (repeated below as (83)) ancora indicates the persistence of the living of Maria in the same place against the supposition that she moved away: (83) Maria abita ancora qui. Maria abitava qua precedentemente. x(t) x‟(t‟) On the other hand the conjunctions dispatch their action inside the sentence relating two phrasal constituents or two clauses. 161 Milena Vegnaduzzo (84) Cioccolato con panna. x‟ x (85) Maria canta e balla. x‟ x Comparing the symbolic representation of the sentences in (84-85) and the one in (83) we note that the former lacks the symbol t. In these cases the temporal relation is not relevant because the presupposition is realized in the dimension of simultaneity. Conclusion In this paper we have seen that the polysemy of the Italian adverb ancora is only apparent. It has a single lexical entry and all the different readings it can assume depend upon the context where it is inserted: each reading derives by compositionality of ancora basic meaning and the semantic properties of the argument structure of the verb. I got to define the basic meaning of ancora by considering what element it has scope on: I showed, then, that in each reading the element falling under its scope is different. I claim that the basic meaning of ancora is made up of two components: presupposition and addition. The entity falling under the scope of ancora is added to the one presupposed. Then I have shown that there is a group of adverbs, di nuovo, un altro, sempre, perfino, più and conjunctions e, con, tuttavia, anche that share the same basic, general meaning found for ancora. This is important with respect to the issue of parts of speech classification. It would be an example of the fact that adverbs and conjunctions are not categories completely isolated from each other and impermeable, but share some features. In this work I studied only a little number of adverbs and conjunctions but we have seen that they can be considered a homogeneous group on the basis of the semantic features they share. This is only a preliminary attempt to open the way to an integrated study of these two word classes to try to define a new classification on the basis of new criteria. References Abraham W. (1980) “The Syncronic and Diachronic Semantics of German Temporal noch and schon with Aspects of English still, yet, already”, in Studies in Language, 4, pp. 3-24. Barker S. (1991) “Even, Still and Counterfactuals”, in Linguistics and Philosophy, 14, pp. 1-38. Battaglia S. (1986) Grande dizionario della lingua italiana, vol I, UTET, Torino. Bertinetto P.M. (1987) Tempo, Aspetto e Azione nel verbo italiano. Accademia della Crusca, Firenze. 162 Ancora and additive words Blücher K. (1974) Studio sulle forme ho cantato, cantai, cantavo, stavo cantando. Struttura funzione e uso nel sistema verbale dell’italiano moderno, Bergen/Oslo/Tromsø. Borgato G.L. (1976) “Aspetto verbale e Aktionsart in italiano e tedesco”, in Lingua e Contesto, 2, pp. 65-197. Bosque I. (1980) Sobre la negacion, Madrid Catedra. Carlson L. (1981) “Aspect and Quantification”, in Syntax and Semantics, 14, pp. 31-64, Tedeschi Zaenen eds.. Comrie B. (1976) Aspect, Cambridge University Press. Coromines (1980) Diccionari etimològic i complementari de la llengua catalana, Barcelona, Curial ediciones catalanas, Caixa de Pensions, “La Caixa”. Cuervo R.C. (1994) Diccionario de construcción y régimen de la lengua castellana, Santafé de Bogota. Davidson D. (1967) “La forma logica degli enunciati d‟azione”, in Azioni ed eventi, trad. it. di R. Brigati (1992) Il Mulino, Bologna. Doherty M. (1973) “Noch and schon and their Presupposition”, in F. Kiefer and N. Ruwet eds. Generative Grammar in Europe, Reidel Dordrecht, pp. 154-177. Emonds J.E. (1987) “Parts of speech in Generative Grammar”, in Linguistics Analysis, 17, pp. 3-42. Garrido J. (1991) “Gestión semántica de la información pragmática en los adverbios de cambio todavía y ya”, in Foro Hispanico, 2, pp. 11- 27. (1992) “Expectations in Spanish and German Adverbs of Change”, in Folia Linguistica, 26, pp. 357-402. Grice H.P. (1975) “Logic and Conversation”, in P. Cole and J. L. Morgan (eds.) Syntax and Semantics, 3 Speech Acts, Academic Press, New York, pp. 41-58. Horn L.R. (1970) “Ain‟t it Hard (Anymore)”, in Campbell M.A. et al. (eds), Papers from the Sixth Regional Meeting of the Chicago Linguistic Society, Chicago, Illinois, pp. 318-327. Jackendoff R. (1972) Semantic Interpretation in Generative Grammar, MIT Press, Cambridge (Mass.). Jespersen O. (1924) The philosophy of grammar, London, Allen and Uhwin. König E. (1977) “Temporal and non-temporal uses of noch and schon in German”, in Linguistics and Philosophy, 1, pp. 177-198. König E., Stark D. (1987) “Function words in a bilingual German- English Dictionary. A new Approch”, in Lexicografica, 3, pp. 158-177. König E., Traugott E.C. (1982) “Divergence and apparent convergence in the development of yet and still”, in Proceedings from the tenth annual meeting of the Berkeley Linguistics Society, Berkeley Linguistics Society, Berkeley, 8, pp. 170-179. Lonzi L. (1989) “Il sintagma avverbiale”, in Grande grammatica di consultazione, a cura di L. Renzi, Il Mulino, Bologna vol. II. Losada Durán J.R. (1992) “Still yet / aún todavía: el aspecto y la negación”, in Revista de lingüística aplicada, 8, pp. 115-125. LUI (1968) Lessico Universale Italiano, Roma 1968. 163 Milena Vegnaduzzo McConnell-Ginet S. (1982) “Adverbs and Logical Form”, Language, 58, pp. 144-184. Morrisey M.D. (1973) “The English Perfective and still/anymore”, in Journal of Linguistics, 9, pp. 65-69. Mourelatos A.P.D. (1981) “Event, Processes, and State”, in Sintax and Semantics, Tedeschi e Zaenen eds. vol 14, pp. 191-211. Muller C. (1975) “Remarques Syntacto-Sémantiques sur Certains Adverbes de Temps”, in Le Français Moderne 43, pp. 12-38. Nef F. (1981) “Encore”, in Langage, 64 pp. 93-107. Parsons T. (1989) “The Progressive in English: Events, States and Process”, in Linguistics and Philosophy, 12, pp. 213-241. (1990) Events in the Semantics of English: A Study in Subatomic Semantics, Cambridge, Mass., MIT Press. Pecoraro W., Pisacane C. (1984) L’avverbio, Bologna, Zanichelli. Reichenbach H. (1947) Elements of Symbolic Logic, London. Renzi L., Salvi G. (eds.) (1991-95) Grande grammatica italiana di consultazione, Il Mulino, Bologna. Rohlfs G. (1969) Grammatica storica della lingua italiana e dei suoi dialetti, III, Einaudi, Torino. Serianni L. (1988) Grammatica italiana, UTET, Torino. Smith C. (1991) The Parameter of Aspect, Kluwer Academic Publishers Dordrecht. Tovena M.L. (1996) “The context sensitivity of the italian adverb ancora”, ms. Traugott E.C., Waterhouse J. (1969) “Already and yet. A suppletive set of aspect markers”, in Journal of Linguistics, 5, pp. 287-304. van de Auwera J. (1993) “Already and still beyond duality”, in Linguistics and Philosophy, 16, pp. 613-653. Vanelli L. (1991) La deissi in italiano, Unipress, Padova. Vendler Z. (1967) Linguistics in Philosophy, Cornell University, Ithaca, New York. Verkuyl H.J. (1989) “Aspectual classes and aspectual composition”, in Linguistics and Philosophy, 12, pp. 39-94. Vlach F. (1993) “Temporal adverbials tense and perfect”, in Linguistics and Philosophy, pp. 231-283. 164 RIVISTA DI GRAMMATICA GENERATIVA Volume 26, anno 2001 L’interpretazione morfosemantica del modo congiuntivo in italiano e in tedesco Paolo CHINELLATO The articulation of inflection in Jamaican Creole Stephanie DURRLEMAN “L-tous”, restructuring quantifier climbing Marco NICOLIS and Manola SALUSTRI On the relative position of beaucoup, guère, peu, rien and trop in French Sara VECCHIATO Ancora and additive words Milena VEGNADUZZO Caterina Santinello Bilinguismo e acquisizione infantile di L2: alcune osservazioni sull’acquisizione simultanea di italiano e tedesco da parte di bambini in età prescolare Cod. RGG026 ISSN 1122-4428