Italian parasynthetic verbs : argument structure

Italian Parasynthetic Verbs Argument Structure Silvia Darteni Thèse présentée et soutenue le pour obtenir le grade de dirigée par preparée au sein du 1 décembre 2017 Docteur de l’Université Paris 8, spécialité Sciences du langage Mme Léa Nash Laboratoire Structures Formelles du Langage Jury : Mme Léa Nash M. Phoevos Panagiotidis Mme Lucia Tovena Mme Bridget Copley Professeur, Université Paris 8 Professeur, University of Cyprus Professeur, Université Paris 7 Chercheuse, CNRS Directeur Rapporteur Rapporteur Examinatrice ii Contents Résumé de thèse Introdu tion 1 Methodologi al notes xiii 1 7 1.1 Introdu tion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2 Data in generative linguisti s 1.3 The issue 1.3.1 7 . . . . . . . . . . . . . . . . . . 8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 In the present study . . . . . . . . . . . . . . . . . . . 16 1.4 The importan e of being reprodu ible . . . . . . . . . . . . . . 18 1.5 Dierent designs . . . . . . . . . . . . . . . . . . . . . . . . . 23 1.6 1.5.1 A eptability Judgment Test (AJT) . . . . . . . . . . . 25 1.5.2 Magnitude Estimation Task (ME) . . . . . . . . . . . . 26 1.5.3 Truth Value Judgment Task (TVJT) . . . . . . . . . . 26 1.5.4 Auto Segmented Reading . . . . . . . . . . . . . . . . . 27 Con lusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Argument Stru ture: State of art 2.1 28 31 Introdu tion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 2.2 Government and binding . . . . . . . . . . . . . . . . . . . . . 32 2.3 Hale and Keyser (1993 and .) . . . . . . . . . . . . . . . . . . 34 2.4 Ram hand (2008) . . . . . . . . . . . . . . . . . . . . . . . . . 38 2.5 Borer (2005) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 2.6 Categorizers and roots . . . . . . . . . . . . . . . . . . . . . . 46 2.7 The present approa h . . . . . . . . . . . . . . . . . . . . . . . 47 2.8 Con lusions 54 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 Parasyntheti verbs 55 3.1 Introdu tion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55 3.2 Parasyntheti 55 3.3 Deadje tival parasyntheti verbs . . . . . . . . . . . . . . . . . . . . . . . . verbs . . . . . . . . . . . . . . . . . iii 59 I 3.4 Denominal parasyntheti 3.5 Con lusions verbs . . . . . . . . . . . . . . . . . 60 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63 Non ambiguous verbs 65 Introdu tion 67 4 Parasyntheti denominal verbs 69 4.1 69 4.2 4.3 4.4 4.5 4.6 5 Introdu tion . . . . . . . . . . . . . . . . . . . . . . . . . . . . Levinson's approa h to pseudo-resultatives (PR) . . . . . . . . 70 4.2.1 Impli it Creation Verbs: features 70 4.2.2 Impli it 4.2.3 Impli it entity is a root . . . . . . . . . . . . . . . . . . 72 4.2.4 Pseudo-resultative . . . . . . . . . . . . . 75 4.2.5 Strong resultatives in Italian . . . . . . . . . . . . . . . 76 Italian denominal parasyntheti s . . . . . . . . . . . . . . . . . 78 4.3.1 . . . 79 . . . . . . . . . . . . . . . . . . 79 . . . . . . . . . . . . reation verbs dier from expli it Italian BNs onstru tion orrespond to impli it Pseudo-resultatives in Italian reation verbs 71 reation verbs 4.4.1 Methodology . . . . . . . . . . . . . . . . . . . . . . . 81 4.4.2 Parti ipants . . . . . . . . . . . . . . . . . . . . . . . . 82 4.4.3 Results for ondition 1 . . . . . . . . . . . . . . . . . . 83 4.4.4 Results for ondition 2 . . . . . . . . . . . . . . . . . . 84 4.4.5 Some improvements to the methodology . . . . . . . . 86 4.4.6 Dis ussion . . . . . . . . . . . . . . . . . . . . . . . . . 87 4.4.7 To sum up . . . . . . . . . . . . . . . . . . . . . . . . . 92 Adverbs are preferred . . . . . . . . . . . . . . . . . . . . . . . 93 4.5.1 93 Methodology . . . . . . . . . . . . . . . . . . . . . . . 4.5.2 Parti ipants . . . . . . . . . . . . . . . . . . . . . . . . 97 4.5.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . 97 4.5.4 Dis ussion . . . . . . . . . . . . . . . . . . . . . . . . . 99 Con lusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102 Pseudo-resultatives in Fren h 103 5.1 Introdu tion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103 5.2 Fren h pseudo-resultatives 5.3 . . . . . . . . . . . . . . . . . . . . 103 5.2.1 Semanti 5.2.2 Dis ussion . . . . . . . . . . . . . . . . . . . . . . . . . 109 Con lusions de ision task . . . . . . . . . . . . . . . . . . 105 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110 iv II Ambiguous verbs 111 Introdu tion 113 6 115 Stativity diagnosti s in Italian 6.1 Introdu tion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115 6.2 Agrammati alities . . . . . . . . . . . . . . . . . . . . . . . . . 117 6.2.1 6.3 7 Imperative and progressive periphrasis Semanti . . . . . . . . . 117 interpretation tests . . . . . . . . . . . . . . . . . . . 122 6.3.1 Interpretation under modal 6.3.2 Future/Present 6.3.3 Contribution in narrative dis ourse onstraint . . . . . . . . . . . . . . . 122 . . . . . . . . . . . . . . . . 129 . . . . . . . . . . . 131 6.4 Experiments involving involuntarily responses 6.5 Con lusions . . . . . . . . . 132 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135 Deadje tival parasyntheti verbs 137 7.1 Introdu tion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137 7.2 Stativity-Eventivity puzzle . . . . . . . . . . . . . . . . . . . . 138 7.3 Deadje tival Parasyntheti s 7.4 . . . . . . . . . . . . . . . . . . . 139 7.3.1 Morphologi al 7.3.2 Dierent types of external arguments roles omponents of DPVs . . . . . . . . . . . 140 Are inanimate subje ts a essible in DPVs? 7.5 Classi ation of DPVs 7.6 Dierent eventualities in DPVs 7.6.1 7.7 7.9 . . . . . . . . . . 148 . . . . . . . . . . . . . . . . . . . . . . 150 dovere under già. Interpretation of . . . . . . . . . . . . . . . . . 153 . . . . . . . . . . . . . . . . . 153 7.6.2 Interpretation 7.6.3 Temporal narrative 7.6.4 Adjun ts . . . . . . . . . . . . . . . . . . . . . . . . . . 159 7.6.5 To sum up . . . . . . . . . . . . . . . . . . . . . . . . . 160 Are all DPVs 7.7.1 7.8 . . . . . . . 147 ausative? DPVs of form . . . . . . . . . . . . . . . . . 156 ontribution . . . . . . . . . . . . . 158 . . . . . . . . . . . . . . . . . . . . . 162 . . . . . . . . . . . . . . . . . . . . . . . 162 7.7.2 DPVs of surfa e . . . . . . . . . . . . . . . . . . . . . . 164 7.7.3 To sum up . . . . . . . . . . . . . . . . . . . . . . . . . 165 Causal relation . . . . . . . . . . . . . . . . . . . . . . . . . . 165 7.8.1 For e-dynami 7.8.2 Causation of stative verbs Synta ti derivations approa h . . . . . . . . . . . . . . . . . 167 . . . . . . . . . . . . . . . . 174 . . . . . . . . . . . . . . . . . . . . . . . 192 7.9.1 Causative eventives . . . . . . . . . . . . . . . . . . . . 192 7.9.2 Causative statives . . . . . . . . . . . . . . . . . . . . . 193 7.10 Causative statives and statives . . . . . . . . . . . . . . . . . . 196 7.11 Predi ate of personal taste . . . . . . . . . . . . . . . . . . . . 198 v 7.11.1 Disagreement test . . . . . . . . . . . . . . . . . . . . . 199 7.11.2 Judge Parameter (eventive reading of DPVs) . . . . . . 200 7.11.3 Judge Parameter (stative reading of DPVs) . . . . . . . 201 7.12 Con lusions 8 Stativity an be automati ally dete ted 8.1 8.2 9 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203 205 Introdu tion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205 Pra ti al appli ations . . . . . . . . . . . . . . . . . . . . . . . 206 8.2.1 Clues and notated 8.2.2 Semanti orpus rules . . . . . . . . . . . . . 207 interpretation task . . . . . . . . . . . . . . . 214 8.3 Stativity/eventivity gradient produ tion 8.4 Con lusions . . . . . . . . . . . . 215 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215 Con lusion 219 Appendix 226 A 227 A.1 List of denominal parasyntheti A.2 Semanti verbs . . . . . . . . . . . . . . 227 interpretation task ITA . . . . . . . . . . . . . . . . 230 A.3 Magnitude estimation task . . . . . . . . . . . . . . . . . . . . 234 A.4 Semanti B.1 List of deadje tival parasyntheti interpretation FR . . . . . . . . . . . . . . . . . . . 236 B 239 Bibliography verbs . . . . . . . . . . . . . 239 253 vi List of Tables 1 Resumé des tests de stativité (DPV). . . . . . . . . . . . . . . xxxix 3.1 Morphologi al . . . . . . . . . . . . . . 3.2 Distribution of denominal parasyntheti s (prexes) 4.1 Parti ipants gender and origin (Semanti 4.2 Parti ipants edu ation (Semanti . . . . . 83 4.3 Parti ipants age groups (Semanti de ision task ITA). . . . . . 83 4.4 Informants' age and edu ation (Magnitude Estimation). . . . . 98 5.1 Semanti onstituents of DPVs. 60 . . . . . . 61 de ision task ITA). . 82 de ision task ITA). interpretation task FR (senten es) . . . . . . . . . . 106 5.2 Parti ipants age (Semanti 5.3 Parti ipants edu ation level (Semanti interpretation task FR). . . . . . . 107 5.4 Semanti 6.1 Dierent readings and 6.2 Condition distribution (Interpretation under modal ITA). . . . 125 interpretation task FR). 107 interpretation task FR (senten es) onstraints of modal . . . . . . . . . . 107 dovere. . . . . . . 123 6.3 Parti ipants (interpretation under modal ITA) . . . . . . . . . 125 6.4 Experimental items (Interpretation under modal ITA). onstituents of DPVs. . . . . 128 7.1 Morphologi al 7.2 Distribution of prexes among adje tive-base . . . . . . . . . . . . . . 140 7.3 Per entages of prex distribution (DPVs). 7.4 Parti ipant so iologi al features (Lexi al lling). . . . . . . . . 149 7.5 Re ap of stativity tests results (DPV). 7.6 Eventualities of DPVs. lasses (DPVs). . 142 . . . . . . . . . . . 144 . . . . . . . . . . . . . 161 . . . . . . . . . . . . . . . . . . . . . . 162 7.7 Distin tive traits of eventive and stati ausation (provisional). 187 7.8 Distin tive traits of eventive and stati ausation (denitive). . 190 8.1 Rules involved in the sear h of progressive. . . . . . . . . . . . 209 8.2 Senten es types pi ked out by rules of table 8.1. 8.3 Rules involved in the sear h of anti ausatives. 8.4 Senten es pi ked up by rules of Table 8.3. vii . . . . . . . . 210 . . . . . . . . . 212 . . . . . . . . . . . 213 8.5 Senten es used in the experiment. . . . . . . . . . . . . . . . . 216 viii List of Figures 1 Estimation d'ampleur, resultats. . . . . . . . . . . . . . . . . . xxiv 1.1 Double obje t onstru tion, length and 4.1 Distribution of BNs among synta ti 4.2 S reen-shot of a task (Semanti 4.3 Results, ondition 1 (Semanti omplexity . . . . . . . patterns. . . . . . . . . . de ision task ITA). de ision task ITA) 21 78 . . . . . . 82 . . . . . . . 84 4.4 Results, ondition 1 (Semanti de ision task ITA) . . . . . . . 85 4.5 Results, ondition 2 (Semanti de ision task ITA) . . . . . . . 86 4.6 Example of item (Magnitude Estimation) . . . . . . . . . . . . 94 4.7 Example of item (Magnitude Estimation) . . . . . . . . . . . . 95 4.8 Referen e senten e, s reen-shot (Magnitude Estimation). . . . 96 4.9 Example of item (Magnitude Estimation) . . . . . . . . . . . . 96 4.10 Example of item (Magnitude Estimation) . . . . . . . . . . . . 97 4.11 Results, items (Magnitude Estimation) 98 . . . . . . . . . . . . . 4.12 Results, subje t (Magnitude Estimation) ondition 1 (Semanti . . . . . . . . . . . . 5.1 Results, 6.1 Gennari and Poeppel (2003: g. 1): verb reading times. . . . . 134 7.1 Distribution of adje tives s ale (deadje tival parasyntheti s). 7.2 Answer means (Lexi al lling). 8.1 Regression Table for Tregex Rules. ix interpretation task FR) 99 . . . . 108 . 142 . . . . . . . . . . . . . . . . . 150 . . . . . . . . . . . . . . . 217 x Remer iements Je tiens à remer ier tout d'abord Léa pour le soutien, les en ouragements qu'elle ne m'a souvent donnés pendant es 5 longues années, ainsi que pour la disponibilité à prendre des rendez-vous à des horaires improbables. Mer i à Bridget pour son aide, ses onseilles et son humanité; j'espère être seule- ment un des premiers et nombreux thésards qui auront le plaisir de travailler ave elle. Isabelle, mer i pour la onan e et la possibilité de travailler sur l'a quisition dans le domaine génératif. Mer i à Elena pour les é hanges sur la distributed morphology. Mer i à Sarra et Coralie dont la ompéten e et la disponibilité ont été indispensables pour les parties expérimentales de thèse. Mer i à Saveria et Asaf pour les ette onseils en matière d'expérien es et de données empiriques. Un grand mer i aux lles du labo pour avoir adou i mon par ours ave leur amitié ; en parti ulier, Adèle et Heglyn, sans elles je n'aurais jamais appris le français, Samantha, mon oin de Vénetie à Paris. Je remer ie le Centre Italian e de Paris et à Roberto pour m'avoir permis de faire le travail que j'aime et de pouvoir a omplir mon do torat, et je remer ie aussi mes étudiants (Annie et Mi hel, Catherine, Olivier, Cristine, Jeannine, Joseline, Vanessa, Odile, Daniel, ...) pour avoir porté le soleil dans mes soirées de travail après des journées entières de re her he. Mar o, mer i d'avoir été toujours plus sûr que moi du résultat de ette aventure. A mon petit Geremia, grâ e à toi, ça a été la n de do torat la plus sereine du monde. Cette thèse a été partiellement nan ée par La Fondation des Treilles, Centre d'études sur le bassin méditerranéen, année 2016, domaine Littérature. La Fondation des Treilles, réée par Anne Gruner S hlumberger, a notamment pour vo ation d'ouvrir et de nourrir le dialogue entre les s ien es et les arts an de faire progresser la Elle a ueille également des Treilles (Var) réation et la re her he ontemporaines. her heurs et des é rivains dans le domaine des www.les-treilles. om . xi xii List of abbreviations BN, denominal parasyntheti verb of type (make) X be ome(s) an N DPV, deadje tival parasyntheti verb EI, event identi ation ICV, impli it reation verb LDD, long distan e dependen ies ME, magnitude estimation task PR, pseudo-resultative SC, small onstru tion lause sg, singular pl, plural m, mas uline f, feminine perf., perfe t det., determinant ger., gerund xiii xiv Résumé de thèse La présente thèse porte sur la stru ture argumentale des verbes parasynthétiques italiens et français. Elle est divisée en deux parties en relation ave la atégorie grammati ale des bases et les types de phénomènes générés. Avant d'entrer dans le vif du sujet, la thèse s'ouvre par des réexions générales à propos de : (i) l'état de l'art et les diérents prennent en adres formels qui ompte la stru ture argumentale et aspe tuelle des prédi ats verbaux ; (ii) la méthode de ré olte des données typique de la grammaire générative et quelques possibles améliorations ; (iii) la morphologie des verbes parasynthétiques. Elles sont dé rites dans les hapitres 1, 2 et 3. La première partie porte sur des verbes parasynthétiques à base nominale qui parti ipent à une onstru tion pseudo-résultative (Levinson, 2007), et ne présentent pas d'ambiguïtés aspe tuelles, a atastare `amon eler'. Elle ontient les omme impilare `empiler', hapitres 4 et 5. La deuxième partie porte sur des verbes parasynthétiques à base adje tivale qui présentent des ambiguïtés de le ture entre une interprétation stative et une événementielle. Elle ontienne les hapitres 6, 7 et 8. Ce dernier ap- plique au domaine du traitement automatique du langage naturel une partie des résultats pré édemment obtenus. En parti ulier, il se propose de dénir les règles utiles à l'identi ation de verbes statifs utilisables sans l'intervention humaine. Notes méthodologiques Le hapitre 1 analyse la méthodologie utilisée habituellement par la gram- maire générative en matière de ré olte des données, et dé rit d'autres méthodologies issues des s ien es ognitives, omme la psy holinguistique, qui se révèlent exploitables également dans le domaine théorique. Le hapitre insiste sur la possibilité de ognitives et la grammaire générative pour ainsi que pour les nalités. Pour ollaboration entre les s ien es e qui on erne les méthodologies, ela, des proto oles expérimentaux plus stri ts doivent être respe tés. xv En grammaire générative, la méthodologie habituelle de olle te de don- nées est l'introspe tion (Cowart, 1997). Celle- i, ainsi que les jugement de grammati alité, est la première forme d'a logies ès aux données. Ces méthodo- ependant présentent des in onvénients. Parmi es in onvénients, on trouve : la di ulté de répli ation de l'expérien e ; l'impossibilité de onduire des analyses statistiques sur les données ; la di ulté de dis riminer les épiphénomènes ; l'utilisation de onnaissan e linguistiques expli ites ; l'exposi- tion prolongée aux même données. Les jugements de grammati alité présentent des ara téristiques parti- ulières par rapport aux méthodologies des autres s ien es ognitives : (i) nombre d'informateurs trop exigu ; (ii) informateurs non naïfs ; (iii) nombre d'options de réponse trop exigu ; (iv) utilisation de pool expérimentaux trop petits ; (v) analyse des données non systématique (S hütze & Sprouse, in press). Les expérien es omposées de jugements de grammati alité ne respe tent pas, normalement, la méthode s ientique, en parti ulier pour e qui on erne omment arment Gibson & Fodorenko (2013) :  the results obtained using this method are not ne essarily generalisable be ause of (a) the small number of experimental parti ipants (typi ally one) ; (b) the small number of experimental stimuli (typi ally one) ; ( ) ognitive biases on the part of the resear her and parti ipants ; and (d) the ee t of the pre eding ontext.. leurs généralisations, En outre, il est di ile de ontrler des autres paramètres pouvant in- uen er le jugement des lo uteurs, notamment pour e qui on erne le ontexte d'interprétation, la fréquen e des mots utilisés, la plausibilité sémantique et l'identi ation de l'objet d'étude par les parti ipants. Les parti ipants aux expérien es ont, en eet, la tendan e à juger la grammati alité des si le onstru tions selon un ontexte d'interprétation n'est pas rendu expli ite dans les instru tions, haque parti ipant est mené à s'en à ontexte qu'ils produisent. Or, onstruire un qui sera diérent par rapport eux des autres. Les jugements exprimés ne seront pas ommensurables. La fréquen e dans la langue du matériel lexi al dont les items expérimentales sont omposés peut rendre plus ou moins fa ile l'interprétation des onstru tions syntaxiques. A parité de moins fréquents rendent moins a onstru tion syntaxique, des mots eptable la � onstru tion . Les jugements de grammati alité sont souvent menés de manière informelle. Cela dérive du fait que les informateurs font souvent partie de l'en- Cela est déterminé par le réseau ognitif que haque mot onstruit : mineure est la distan e entre deux mots et majeure est la probabilité qu'un lo uteur puisse établir un lien qui mène à une possible interprétation, même en présen e d'une agrammati alité. � xvi tourage du her heur. Cela peut auser une observer expe tan y (Gibson et al. 2013 : 100), un biais par lequel les informateurs auront tendan e à juger les items expérimentaux en fon tion des d'étude du onnaissan es qu'ils ont de l'objet her heur et non en fon tion de leurs. Des informateurs ons ients de l'obje tif de l'étude à laquelle ils parti ipent sont portés à tomber vi time de bias de onrmation (Gibson et al. 2013 : 99) et de royan e (Evans, Barston, Pollard 1983). Tous les biais possibles dé rits sont plus fréquents dans le as de l'utili- sation de la méthode dite d'auto-investigation, dans laquelle l'informateur et le her heur sont une seule et même personne (Levelt 1972). Toutefois, la présente thèse re onnaît la valeur de l'auto-investigation dans deux as spé iques. Le premier est le as d'une investigation s ien- tique sur un phénomène linguistique qui n'a jamais été tage. En eet, dans e ontexte, le auto-investigation pour ir ons rit davan- her heur doit for ément pro éder à une omprendre les points d'intérêt possibles, la façon la plus adaptée pour le dé rire, ainsi que les méthodes expérimentales les plus onformes à l'investigation de l'objet d'étude. Le deuxième as onsiste dans la re her he dans les faits basiques d'une langue (ordre des mots, a ord sujet-verbe, ...). Ce hapitre dé rit l'appro he et les solutions théoriques appliqués au ours de la thèse. La syntaxique lexi ale des verbes est formée de trois proje tions fon tionnelles : rP, v P, Voi e P. rP est une proje tion relationnelle non-événementielle (A edo-Matellan 2006) qui met en relation l'objet dire t et la base verbale qui a nature de ra ine. La présen e de ette proje tion détermine la sémantique ausative de la dérivation (Hoekstra 1988 ; S häfer 2008). La proje tion mantique vP a omme tête une tête fon tionnelle dont la valeur sé- aktionsart hange selon l' deux sémantiques sont proposées du verbe. Dans le vbe ome hapitre 7, par exemple, pour les verbes événementiels du abbellire `embellir' ; vrelation pour les verbes statifs ausatifs. Voi e P est responsable pour l'introdu tion de l'argument externe type (Krat- zer 1996). Le hapitre 1 poursuit en dé rivant les impré isions de ertaines données linguistiques rapportées dans des études de linguistiques formelles et qui ont été mises en lumière dernièrement. Le statut des données en linguistique formelle, en parti ulier en syntaxe formelle, fait l'objet d'un rée tion propre dans les dernières années par un nombre roissant de her heurs (Edelman & Christiansen (2003), Ferreira (2005), Wasow & Arnold (2005), Featherston (2007), Gibson & Fedorenko (2010a, 2010b), xvii inter alia ). Le hapitre rapport des as où les données, ré oltées grâ e à l'utilisation de méthodes informelles se sont révélées être problématiques : et heavy NP shift ; onstru tions anglaises à double objet fa teurs inuençant la position de la préposition dans les phrases anglaises (Wasow & Arnold, 2005) ; interprétation des onstru tions relatives du sujet et de l'objet ; extra tions multiples de pronoms wh (Gibson & Fedorenko, 2013). Il a été souligné que le re ours à des méthodes de olle te de données plus stri te est parti ulièrement important pour les langues qui possèdent un grand nombre de variables diatopiques. Le hapitre se poursuit en dé rivant les fa teurs qui peuvent inuen er les résultats des jugements de grammati alité (Keller, 1998), ainsi que différentes méthodologies expérimentales qui peuvent s'avérer utiles pour les synta ti iens formels. Des jugements de grammati alité plus stru turés peuvent améliorer la qualité des donnés en grammaire générative. Pour ela, il a été démontré que quatre fa teurs peuvent inuen er négativement les données re ueillies et ils doivent être ontrlés. Ils sont : l'é helle d'évaluation, les instru tions, diérentes problématiques liées aux sujets expérimentaux, diérentes problématiques liées aux tâ hes expérimentales. Le hapitre 1 dé rit ertaines des méthodologie qui pourraient être utiles en grammaire générative. En parti ulier, il prend en judgment test and Sora e, ompte : a eptability Magnitude Estimation Task (MET) (Bard, Robertson 1996), Truth Value Judgment Test (TVJT) (Gordon & Chafetz, (AJT), 1986), le ture auto-segmentée. Pour ha une de es méthodologies expéri- mentales, les se tions proposent une petite des ription du design et les aspe ts les plus utiles pour des synta ti iens. Il ne dé rit pas les méthodologies utilisées au ours de la thèse. Elles sont exposées dans les données sont utilisées pour la En on lusion, le hapitres où leurs onstru tion de la théorie. hapitre 1 prend en onsidération ertaines probléma- tiques que l'utilisation de petites expérimentations, qui souvent ne respe tent pas la méthode s ientique, peuvent entraîner. Le hapitre ne vise pas à mettre en dis ussion l'utilité du re ours aux jugements de grammati alité non-stru turés et de l'auto-analyse. Il re onnaît une pla e à es méthodo- logies dans les phases préliminaires d'étude d'un phénomène linguistique, ou dans le as de phénomènes qui regardent des faits basiques de la langue (ordre des mots, a de ord sujet-verbe, ...). Il tient à mettre l'a ent sur la prise ons ien e de la part de la plupart de synta ti iens génératifs à propos de l'importan e du respe t de proto oles expérimentaux plus stri ts. Le ment a deux d'é hanges ave hange- onséquen es envisageables. La première regarde la possibilité les autres dis iplines ognitives qui ont omme objet d'étude les langues et le langage. La deuxième regarde la fondation de la même en- xviii treprise générative sur des données qui puissent être reprodu tibilité des expérien es ave ontrlées grâ e à la lesquelles elles ont été ré oltées. Cadres formels : état de l'art Le hapitre 2 fournit une des ription des adres formels qui s'o upent de la dénition de la stru ture argumentale des verbes. En parti ulier, il résume les théories qui rendent ompte de la sémantique des verbes morphologiquement dérivés. La relation entre stru ture argumentale, nombre d'argument et leurs rles sémantiques est intimement liée au type de aktionsart des verbes. La stru - ture argumentale d'un verbe est impliquée par la sémantique événementielle du même verbe en diérentes manières. (1) Daria (2) Daria mord son frère. (3) Daria aime sa femme. agent  but agent  expérient détenteur  fin ourt trois kilomètres. Dans les exemples du (1) au (3), le type de mine le rle sémantique des arguments. Pour aktionsart du verbe déter- ela, une étude approfondie des nombre et type d'arguments d'un verbe est né essaire dans le adre d'une re her he sur les ambiguïtés aspe tuelles. Le hapitre 2 propose une aperçue des diérents adres formels portant sur la stru ture argumentale dont quatre en parti ulier sont développés. En parti ulier, il en prend en ompte quatre : théorie du gouvernement et liage ; Hale & Keyser (1993 ; 2002) ; Ram hand (2008) ; Borer (2005). Après avoir dé rit la théorie du gouvernement et du liage, dans laquelle le lexique et la syntaxe parlent deux langages diérents qui ont besoin de règles de onversion pour pouvoir dialoguer, et dans laquelle la question de la dé- termination de l'événementialité du verbe n'est pas prise en hapitre se poursuit ave onsidération, le la théorie proposée par Hale & Keyser (1993 et ss.), où la syntaxe est responsable soit de détermination de l'événementialité du verbe soit du rle des arguments dans dans Ram hand (2008), où la ette événementialité. Le même arrive rst phase, organe syntaxique, est responsable aktionsart. de la dénition du nombre et du type d'arguments, ainsi que de l' Toutes es théories re onnaissent, à des degrés diérents, le lexique porteur d'informations utiles à la Le adre formel de Borer (2005), au tan e syntaxique au lexique. Dans omme réation syntaxique. ontraire, n'attribue guère d'impor- e dernier ils sont emmagasinées seulement xix des ra ines, sans au une information sur les stru tures dans lesquelles elles peuvent être insérées. Une autre forte diéren e qui distingue les dé rits et adres formels pré édemment elui de Borer (2005) onsiste dans les prin ipes retenus détermi- aktionsart. Pour Hale & Keyser (1993, ss.) and Ram hand (2008), nants de l' le prin ipe déterminant est le type et le nombre de sous-événements présents. Pour Borer, les plus important est la présen e ou absen e de téli ité. Le hapitre analyse omment ha un des quatre adres représentent les verbes statifs des diérents groupes. Au un d'entre eux ne propose une solution satisfaisante, et question, ertains ne prennent pas du tout en onsidération la omme la théorie du gouvernement et du liage. Verbes parasynthétiques Le hapitre 3 dé rit le pro essus de dérivation parasynthétique. La parasyn- thèse est un pro essus morphologique des langues romanes qui dérive d'une réinterpretation d'un autre pro essus de dérivation de la latinité tardive : les verbes préxés ont été interprétés omme synonymes de leurs orrespondants non-préxés (Ia obini, 2004). La parasynthèse ne regarde pas seulement le domaine verbal, mais ainsi les domaines nominal et adje tival. Cette thèse s'o upera seulement du premier. La première dénition de la parasynthèse a été proposée par Darmesteter (1894). Cette dernière la dé rit omme un pro essus lexi al impliquant un préxe, une base et un suxe qui se ombinent simultanément et dont le produit de dérivation intermédiaire n'est pas attesté dans le lexique de la langue. Cette hypothèse de formation (Darmesteter, 1894 ; Ia obini, 2004) est une des trois formulées dans la littérature. S alise (1990) parasynthèse omme le produit d'une suxation à laquelle su xation. Corbin (1987) la laquelle su onsidère onsidère la ède une pré- omme un produit d'une préxation à ède une suxation. Cha une de es hypothèses sur les étapes de dérivation de la parasynthèse présent des problèmes. La première ne respe te pas l'hypothèse de la ramiation binaire. La deuxième ne semble pas respe ter le troisième assigne aux préxes la fa ulté de ale, e qui ne se vérie dans au un autre S alise (1990) semble être Le hanger de mirror prin iple. La atégorie grammati- as de la langue. L'hypothèse de elle qui pose moins de problèmes. hapitre se poursuit en dé rivant les deux groupes de verbes para- synthétiques qui onstituent l'objet de groupes d'appartenan e. xx ette étude, en les plaçant dans leurs Les verbes dérivés des adje tifs ont une sémantique ausative qui peut être dé rite par la paraphrase faire l'objet plus A. Le degré du hangement qui a lieu sur l'objet et qui est exprimé par la base verbale est laissé inexprimé (Ia obini, 2004). Ces verbes se divisent en deux ongurations syntaxiques : ils peuvent alterner entre une stru ture transitive et une in hoative pronominale, ou entre une stru ture transitive et une in hoative non pronominale. 221 verbes italiens ont été identiés omme abbellire `embellir', Il font l'objet d'étude de annerire omme appartenant à `noir ir', appesantire ette atégorie, � `alourdir' (DPVs) . ette thèse. Les verbes dérivés des substantifs peuvent être divisés en trois groupes selon la sémantique de la base sur laquelle ils sont formés : ausatifs, lo atifs et instrumentaux. Le premier groupe à son tour se divise entre trois sous- � groupes selon la paraphrase que les verbes génèrent : faire devenir S , faire devenir omme N,  auser/prendre N. Les verbes dérivés des substantifs peuvent parti iper à quatre ongurations syntaxiques transitifs ou intran- sitifs, alterner entre une stru ture transitive et une intransitive ou entre une stru ture transitive et une intransitive pronominale. 57 verbes de la ausatifs du type `faire devenir S' ont été identiés. Seule leur transitive a été prise en onsidération dans à pouvoir parti iper à la omme impilare, `empiler', ette étude, atégorie onguration ar elle est la seule onstru tion pseudo-résultative (Levinson, 2007), a atastare, `empiler de façon désordonnée'. Première partie : verbes non-ambiguës La première partie de ette thèse analyse base nominale. Elle en analyse le ertains verbes parasynthétiques à omportement dans la onstru tion pseudo- résultative. La onstru tion pseudo-résultative (Levinson, 2007) est onstituée d'un adje tif qui modie l'entité dénotée par la base du verbe. Dans l'exemple suivant, on peut voir que l'adje tif high, `haut', modie la base verbale pile, `pile'. (4) La → John made a high pile of books. Jean empila les livres hauts. → Jean t une haute pile de livres. John piled books high. onstru tion pseudo-résultative est grammati ale en anglais, et ne pose au un problème théorique tatives ar l'anglais peut former des (Washio, 1997) de type adje tival. L'étude de les langues romanes est plus intéressante onstru tion dans ar elles sont de type � La liste omplète est reportée dans l'appendi e. � Où S orrespond à la base nominale. xxi ette ela strong resulverb frame (Talmy, 1991) et par strong onséquent ne peuvent pas parti iper aux onstru tions résultatives. Dans e adre, une étude a été menée pour re ueillir des données sur l'italien et le français. En parti ulier, des expérien es d'interprétation ont été onduites sur des lo uteurs natifs d'italien et de français pour vérier la grammati alité de la onstru tion pseudo-résultative (PR) dans es deux langues. Le hapitre 4 reporte les résultats d'une expérien e de dé ision sémantique qui a été onduite sur 106 lo uteurs natifs de l'italien. Il montre lo uteurs natifs de l'italien a eptent la omme les onstru tion PR dans 85% des analysés quand l'objet dire t est expli ite (5), et dans 99% des as as quand l'objet dire t est pronominal (6). (5) Giovanni ha im-pilai-to G. a im-pile- i libri altii . perf. det.m.pl. livre.m.pl. haut-m.pl. G. a empilé les livres hauts. (6) Quando Giovanni ha messo Quand G. li 3.m.pl.a a a posto i perf. à pla mis- ha impilati a im-pile- e libri, det.m.pl. livre-pl., alti. perf. haut-m.pl. Quand G. a rangé les livre, il les a empilés hauts. L'a ord morphologique expli ite des adje tifs italiens est parti ulière- ment utile dans la onrmation de l'hypothèse avan ée par Levinson (2007) sur la stru ture de la a onstru tion PR. En eet, le fait que l'adje tif soit ordé morphologiquement ave l'objet dire t, alors qu'il modie l'entité impli ite (la base) du verbe, est pris omme un omportement révélateur de la nature de la base verbale. Elle est une ra ine et pas un substantif atégo- risé. Deux autres points permettent de mettre en éviden e la nature rielle de la base verbale. Le premier est atégo- onstitué du fait que la (a)teli ité (Pustejovsky 1991 ; Ja kendo 1991) des verbes parti ipants à la onstru tion PR n'est pas ae tée par le type de base présente. Par exemple, la phrase n'indique pas lairement le nombre de piles que le sujet a révélateur du fait que la base verbale, en n'étant pas réées. Cela est atégorisée, n'est pas spé iée pour le trait de nombre. (7) Jean empila les livres. Le deuxième point est onstitué d'un test lexi ale. Il est basé sur l'hy- pothèse qu'un verbe dérivé d'une ra ine peut être modié par des adjoints référant au même hamp sémantique sans générer des phrases ina ontrairement à un verbe formé sur un substantif xxii atégorisé. eptables, (8) Sandro ha allineato le tessere del domino lungo una ir onferenza. Sandro a aligné les domino le long d'une ir onféren e. (9) ... Si sono spinti giù per la rampa e hanno ammassato in un mu le � operte hio he fanno da letto ai nuovi ospiti . Ils ont des endu la rampe et ils ont mis dans un tas les ouvertures qui font de lit aux nouveaux htes. (10) Oggi appaiono separati uno dall'altro non solo per le su essive ero- sioni operate sulla dorsale dal Torrente Cormor, ma an he per l'azione di due sistemi oniugati di faglie verti ali he in tempi re enti hanno spezzato in segmenti la dorsale spostandone leggermente le singole � porzioni. Aujourd'hui ils semblent séparés l'un de l'autre par les érosions qui ont opéré sur la dorsale du torrent Comor, mais aussi par l'a tion de deux systèmes de failles verti ales qui ont assé en segments la dorsale en déplaçant haque portion dans les temps ré ents. Grâ e à supposée être (11) es points, la stru ture argumentale de la phrase en (11) est elle reportée en (12). Carla sbri iola i bis otti ni. Carla fait des nes miettes de bis uit. V oice auserP (12) DP Carla vbe omeP V oice auser rP vbe ome DP bis otti [F℄ r' r=INTO a-, in-, s- P √ √ bri iola � http://ri 18/10/2016. er a.repubbli a.it/repubbli a/ar hivio/repubbli a/2010/08/30/nei-box-sotterranei-hote � http://www.geos 18/10/2016. AP n- [uF℄ ienze.units.it/geositi/vedigeo1.php?ID_GEO=221, xxiii ��� � ��� �������� ������� ������� � ��� �� �� �� �� � � � ��� �� �� � � � ��� ��� �� � � � ���� ���� � ��� � �� � � �� �� � � �� �� � �� � � � �� � ��� � � �� � � �� � �� ������ �� � �� � �� �� � � � �� � �� �� � ��� � � � �� � � �� � ��� � � �� � ��� ���� � �� � ��� �� � � ���� � � ��� � ���� �� � �� � � ��� �� � � �� � ��� �� �� � ���� � � �� � ��� �� �� �� � �� � �� � � �� � � � �� � ��� �� � ��� �� ��� �� �� � Figure 1 : Estimation d'ampleur, resultats. Les lo uteurs natifs de l'italien parti ipant à la première expérien e sur phrases omme (13) et (14), ont exprimé de manière informelle leur préféren e pour des phrases (13) omme (15), où la modi ation est faite par un adverbe. Quando gio ano, i bambini in olonnano i lego storti. Quand les enfants jouent, ils empilent les lego tordus. (14) Quando gio ano on i lego, i bambini li in olonnano storti. Quand ils jouent aux legos, les enfants les empilent tordus. (15) Quando gio ano on i lego, i bambini li in olonnano onfusamente. Quand ils jouent aux legos, les enfants les empilent onfusément. Une expérien e d'estimation d'ampleur (Bard, Robertson, Sora e 1996) a été menée pour déterminer si les sensations reportées de manière informelle par les lo uteurs étaient s ientiquement tés à la se tion 4.5 et dans le graphe onrmées. Les résultats, repor- (à la page xxiv), montrent que les adverbes synonymes des adje tifs en fon tion pseudo-resultative sont ee tivement préférés. La se tion propose une motivation à adverbes par rapport aux adje tifs dans la adverbes peuvent avoir deux s ope ette fa ilité d'interprétation des onstru tion PR. Notamment, les quand ils modient un verbe résultatif. xxiv lower s ope se produit quand l'adverbe modie wide s ope quand il modie la partie verbale. Un L'objet de la partie résultative, un ette étude est parti ulièrement intéressant dans le pano- rama des prédi ations se ondes dans les langues romanes (Talmy 1991, 2000 ; A edo-Matellan 2012 ; Folli 2001 ; inter alia ). En eet, l'italien montre des a- ra téristiques singulières par rapport aux autres langues de la même famille. Si les onstru tions resultatives prépositionelles sont pleinement produ tives, omme on s'y attend, les onstru tions résultatives adje tivales le sont par- tiellement (Folli, 2001 ; Napoli, 1992), alors que on s'attendrait à qu'elles ne le soient pas. Pour ela le hapitre 5 analyse l'a essibilité à la onstru tion pseudo- resultative de 44 lo uteurs natifs du français grâ e à une expérien e d'inter- � prétation sémantique (16) équivalente à elle onduite pour l'italien. Quand Jean essaye de ranger ses aaires, il les amon elle hautes sur le bureau. Les résultats sont intéressants ar ils montrent que l'a essibilité à ette onstru tion est en français aussi plus faible qu'en italien. En parti ulier, sur 8 verbes testés, seulement 3 ont reporté des valeurs hypothèse de grammati alité de la Il est assez remarquable que phonologique dire t ave (17) a. empiler b. tresser → → onstru tion : ompatibles ave empiler, tran her, tresser. es verbes soient les seuls qui ont un rapport la base verbale. pile tresse . entasser � tas d. amasser � amas Le fait qu'ils aient un rapport phonologique transparent ave est le fa teur déterminant pour la possibilité d'y fet, si la base est a une leurs bases onstruire la PR. En ef- essible phonologiquement aux lo uteurs, elle l'est aussi syntaxiquement. Cela permet à l'adje tif de la modier (19). (18) ... empile les livres hautes. � Dont les items expérimentaux sont reportés dans l'appendi e. Toute l'expérien e a été validée du point de vue de la orre tion linguistique par un lo uteur natif. xxv (19) vP v rP DP r' les livres em- Au √ r=INTO P √ AP pile hauts ontraire, les verbes qui ne possèdent pas un rapport phonologique transparent ave leurs bases ne sont pas perçus par les lo uteurs natifs omme syntaxiquement dérivés. Cela empê he à l'adje tif de modier la base verbale, ette dernière n'étant pas présente dans la dérivation (21). (20) ... amasser les livres hauts. (21) Voi eP ... vP v DP amasser les livres hauts Pour résumer, la première partie de la thèse prend en onsidération des verbes parasynthétiques italiens et en étudie les intera tions ave la onstru - tion pseudo-résultative (PR). Grâ e aux résultats de deux expérien es sur des lo uteurs natifs de l'italien, l'a eptabilité de la onduites onstru tion PR a pu être pré isée. Elle est préférée quand l'objet dire t est pronominalisé (99% de taux d'a eptabilité) plutt que lorsqu'il est l'objet dire t expli it (85% de taux d'a eptabilité). Les adverbes synonymes des adje tifs (s'ils sont présents dans le lexique de la langue) reçoivent plus fa ilement une interprétation PR. Cela est expliqué par leur plus grande orrespondan e entre syntaxe et sémantique. xxvi Pour vérier si l'italien o upe une position parti ulière dans le pano- rama des langues romanes pour e qui on erne la onstru tion PR, omme pour les se ondes prédi ations adje tivales (Di Napoli 1992 ; Folli 2005), le hapitre 5 reporte les résultats d'une expérien e d'interprétation sémantique onduite sur le français. Ils montrent qu'en français la pas généralement a onstru tion PR n'est eptable, sauf pour les verbes dérivés qui ont une rela- tion phonologique expli ite ave leur base. Dans e as l'a eptabilité de la onstru tion monte signi ativement. Deuxième partie : verbes ambiguës La deuxième partie de tiques ette thèse porte sur les verbes italiens parasynthé- ausatifs du type faire N plus A, qui entretiennent une double le ture aspe tuelle : stative et événementielle. (22) Daria abbellis e la stanza. Daria embellit la hambre. (23) Le foto abbellis ono la stanza. Le photo embellissent la hambre. Cette partie tou he diérentes problématiques liées aux questions de la stativité, de la ausalité et de leur rapport. En parti ulier, il est mis en éviden e que stativité et ausalité ne sont pas opposées, mais qu'elles peuvent être présentes dans un même verbe ; le fait qu'elles soient souvent séparées est dû à des questions d'environnements formel appelé ognitives qui déterminent une fa ilité de ausatifs dans le for e-dynami onstru tion as de verbes événementiels. Le adre (Copley & Harley, 2015) a été adopté. Les verbes parasynthétiques à base adje tivale pris en ompte sont divi- ins(ingiallire, sés en trois groupes, selon la sémantique de leur base : psy hologiques ( tupidire, `abrutir'), de forme ( ingrandire, `agrandir'), de surfa e `jaunir'). Seuls les deux derniers sont étudiés, dans le ulier, les verbes de forme sont supposés impliquer un qui don entretient une hapitre 7. En partihangement physique, ausalité énergétique et qui en onséquen e dérive une le ture événementielle (24). Les verbes de surfa e sont supposés générer deux le tures qui sont mises en éviden e par l'(in)anima ité de l'argument externe. Dans le as d'arguments externes inanimés, les verbes de surfa e ont une le ture stative ; dans le le ture événementielle � as d'arguments externes animés, ils ont une (25). � Au moins qu'ils ne sont pas interprétés omme inanimés, omme en : (1) Pierre illumine la piè e par sa présen e. xxvii (24) a. Daria appesantis e la bar a. Daria alourdit le navire. b. La assa di emento appesantis e la bar a. La aisse de béton alourdit le navire. (25) a. Daria ingiallis e la asa. Daria jaunit la maison. b. L'erba se a ingiallis e la asa. L'herbe sè he jaunit la maison. La distin tion entre diérentes le tures aspe tuelles impose une autre problématique théorique, notamment Le hapitre 6 prend en elle liée aux diagnosti s de la stativité. onsidération les diagnosti s les plus fréquemment utilisés dans la littérature, pour en analyser la abilité. Les diagnosti s qui se révèlent être pertinents sont utilisés dans le des riptive, propose une mise à jour du pouvoir y insérer les verbes statifs Le hapitre 7, qui après une partie adre formel for e-dynamique pour ausatifs. hapitre 8 reporte les résultats d'une ollaboration dans un projet in- ternational entre CNRS-SFL (Fran e) et Emory University (Georgia, US) à propos de la détermination automatique de l'aspe t verbal. En parti ulier, il dé rit les étapes qui ont été suivies pour la réation d'un gradient de stati- vité/événementialité des verbes anglais extrapolés d'un orpus de Twitter. Diagnosti s pour la stativité Le hapitre 6 analyse les diérents diagnosti s de la stativité présents dans la littérature. En parti ulier, il les divise en deux groupes selon le phénomène qu'ils mettent en éviden e. Le premier groupe est qui utilisent des onstitué de diagnosti s ritères syntaxiques. Le deuxième de eux qui utilisent des ritères sémantiques. Parmi les diagnosti s qui utilisent des ritères syntaxiques, don la di ho- tomie grammati al/agrammati al, on trouve l'impossibilité pour les verbes statifs de parti iper de manière li ite à l'impératif et à la périphrase progressive (Bertinetto, 1991 : 30). Le hapitre met en éviden e le fait que e type de diagnostiques n'est pas able. Pour e qui on erne l'agrammati alité des statifs dans la périphrase pro- gressive, on peut voir que e diagnosti semble fon tionner des verbes statifs prototypiques (26). Toutefois, mann (2004 : 347), omme sorire, ela n'est pas le `sourir', en (27), ou omme prévu ave omme armé par Gross- as pour la totalité des verbes statifs, amare, xxviii `aimer', en (28). (26) *Sta possedendo inque ase. Il est en train de posséder inq maisons. (27) Sta sorendo. Il est en train de sourir. (28) Maria sta amando questo aè. Marie est en train d'aimer e afé. Il faut remarquer que l'exemple (28) dé rit une situation pon tuelle. En eet, le progressif italien for e une le ture parti ulière, limitée dans le temps : [...℄ the Italian dia hroni data show that at the beginning the progressive refers to purely durative situations and only later has it spe ialized as an aspe tual form, not expressing purely durativi y, but imperfe tivity , (Squar tini 1998 : 102). En d'autres termes, des états permanents ou des a tivités � qui durent toute la vie d'un individu (29) onstru tion aspe tuelle. A sont agrammati ales dans l'ex lusion des statifs du progressif, arme la majeure a dans ette ette e propos, Squartini (1998), tout en armant � eptabilité des SLP onstru tion. L'usage du progressif est en expansion dans l'italien retta, 1993 : 220), ontemporain (Ber- e qui peut en expliquer la grammati alité de ertains statifs, notamment les SLPs. (29) Maria sta lavorando a s uola. Marie est en train de travailler à l'é ole. L'autre diagnosti souvent utilisé dans la littérature pour dis riminer entre verbes statifs et événementiels, sur la base de ritères syntaxiques, est l'impératif. Selon Squartini (1990) et Levin (2007), l'agrammati alité qui se produit est due à une manque d'agentivité, e qui automatiquement ex lu les verbes statifs. Toutefois, on peut voir dans les exemples suivants que, même en étant dépourvus d'agent, ils sont parfaitement a eptables sous l'impéra- tif. (30) Ri ordati di santi are le feste. Pense à observer le jour du repos. (31) Non desiderare la donna d'altri. Tu ne ommettras pas d'adultère. Cet example est agrammati al là où il est interprété omme si l'a tivité de Marie se déroule de manière durable pour toute la vie de Marie, interprétation possible pour le progressif espagnol. � Stage level predi ates. � xxix Le hapitre se on lut en formulant l'hypothèse selon laquelle l'agram- aktionsart mati alité de l'impératif ne reside pas dans l' du verbe, mais dans l'impossibilité de la personne à laquelle l'impératif s'adresse d'inuen er l'événement. En on lusion, ni l'impératif ni la périphrase progressive ne sont des diag- aktionsart nosti s ables pour la dis rimination de l' Du statif en italien. oté des diagnosti s reposant des ambiguïtés sémantiques omme dis- riminant entre verbes statifs et événementiels, on trouve : l'interprétation sous verbe modal, l'orientation temporelle et la d'une ontribution à l'avan ement haîne narrative. Giorgi & Pianesi (1997) prennent en onsidération les diérentes interpré- tations que les verbes modaux anglais génèrent, en parti ulier déontique ou épistemique. L'interprétation déontique must on erne un ordre sur une a tion qui doit être réalisée. L'interprétation épistemique hypothèse à propos de l'état des `devoir' : on erne une hoses, une spé ulation sur une situation présente. La diéren e de le ture que les verbes modaux peuvent engendrer est liée aktionsart à l' du verbe lexi al. Les verbes statifs peuvent générer les deux le tures (32), les verbes événementiels seulement la le ture déontique (33). (32) (33) Le Daria doit aimer Pierre... a. pour b. pour être une bonne femme. Daria doit ommettre une erreur si bête. ourir le Marathon de Paris... a. # pour abîmer ses b. pour min ir. haussures de ette façon. hapitre reporte les résultats d'un test d'interprétation sémantique qui a été onduit sur 188 lo uteurs natifs de l'italien, et qui avait obje tif l'étude de la validité de �� omme ette distin tion de le ture en italien. Les résultats montrent que le verbe modal italien aktionsart à diérentes interprétations selon l' dovere `devoir' onduit du verbe lexi al. Les verbes, qui ont été in lus dans l'expérien e pour être probablement statifs, ont été �� Une réplique d'un autre test onduit sur lo uteurs natifs de l'anglais et dont les résul- tats sont ontenus dans le hapitre 8. xxx jugés, sous verbe modal, �� tiques . Au omme générateurs de le tures epistémiques et déon- ontraire, les verbes in lus en étant probablement événementiels ont généré seulement une le ture déontique (Tableau 6.4). Des exemples d'items expérimentaux sont fournis dans les phrases suivantes. (34) Carla deve onos ere il ontenuto del testamento di Maria. Carla doit onnaître le ontenu du testament de Marie. (35) Il libro sulla storia d'Italia deve interessare Maria. Le livre sur l'histoire italienne doit intéresser Marie. (36) Sandro deve s iogliere del burro. Sandro doit faire fondre le beurre. La possible ambiguïté de le ture générée par les verbes modaux est liée à l'orientation temporelle des phrases qui les ontiennent. Une phrase onte- nant un verbe statif sous modal requiert que la situation soit réalisée dans le présent ; au ontraire, une phrase ontenant un verbe événementiel sous modal requiert que la situation soit réalisée dans le futur (Condoravdi 2002 : Note that the temporal interpretation of the omplement in [a stative senten e℄ is present-like, while in [an eventive senten e℄ is future-like. [The stative one℄ means that given what we know now it follows that you love Lin now, while [the eventive one℄ means that to be in line with requirements you need to kiss Lin sometime in the future  (Katz 2006). 69) :  (37) Daria deve amare Maria oggi/*domani. Daria doit aimer Marie aujourd'hui/*demain. (38) Daria deve orrere la maratona di Parigi oggi/domani. Daria doit ourir la marathon de Paris aujourd'hui/demain. Un autre diagnosti qui utilise des ritères sémantiques qui sont de distinguer entre statifs et événementiels est sibilités d'avan ement de la apable onstituée des diérentes pos- haîne narrative par les deux aktionsarten (Dry, 1983 ; Katz, 2003). L'exemple (39) rée une haîne narrative qui s'ouvre par Daria qui arrive à la maison, et se poursuit par l'événement de sa lle qui s'assoit et par l'événement du hien qui s'endort. L'exemple (40) qui s'ouvre ave l'arrivée de Daria pendant que sa lle était assise et le �� était endormi rée une . �� Même si ette dernière a été séle tionnée moins fréquemment. �� Les imparfaits sont des statifs dérivés. xxxi haîne narrative hien (39) Daria est arrivée. Sa lle s'est assise sur le anapé et le hien s'est anapé et le hien était endormi sur le tapis. (40) Daria est arrivée. Sa lle était assise sur le endormi sur le tapis. La ontribution à la narration diérentes dans les deux exemples pré é- dents est en outre démontrée par le fait que, pour le premier, un dans l'ordre d'apparition des verbes que, pour le deuxième, un ne onduit à au un (41) (42) En hangement dans l'ordre d'apparition des verbes hien s'est endormi sur le tapis et sa lle s'est anapé. Daria est arrivée. Le assise sur le onduit à une narration diérente, tandis hangement dans la narration. Daria est arrivée. Le assise sur le hangement hien était endormi sur le tapis et sa lle était anapé. on lusion, le hapitre 6 porte sur les tests syntaxiques et sémantiques qui sont normalement utilisés omme diagnosti s de stativité. Il démontre que les diagnosti s qui utilisent des agrammati alités ne sont pas ables. Au ontraire, eux qui utilisent des ambiguïtés sémantiques sont apables de tran her une ligne nette entre verbes statifs et événementiels. Le hapitre 6 rapporte les résultats d'une expérien e de Gennari & Poep- pel (2003), qui met en éviden e une diéren e des temps de le ture des verbes statifs par rapports à eux des verbes événementiels dans le adre d'une ex- périen e de le ture auto-segmentée. Les verbes statifs sont lus plus vite que les verbes événementiels, à parité de onditions (Tableau 6.1). Verbes à base adje tivale La détermination des éléments responsables de la stativité est un problème omplexe. L'étude des verbes qui présentent une alternan e systématique entre deux aktionsarten, et qui onstituent don des paires minimales, peut l'é lair ir. Les diéren es aspe tuelles qui peuvent avoir lieu dans les verbes parasynthétiques à base adje tivale sont analysées. En parti ulier, le analyse la sémantique hapitre 7 ausative qui génèrent deux le tures en relation à la sémantique de la base verbale. L'appro he théorique utilisée est la 2015) qui a été revue et dont nouveaux introduits. C'est le for e-dynamique (Colpey & Harley, ertains éléments ont été modiés, ou des as de abdu tion, une omposante ausative, introduite dans la dérivation par le lo uteur, qui est responsable de la le ture xxxii stative- ausative et dont la orrespondan e ognitive est onrmée par la présen e d'un paramètre de juge (Laherson, 2005 ; Stephenson, 2007). Ambiguïtés de le ture aspe tuelle aktionsarten Diérents types d'ambiguïtés entre existent. Des verbes sta- tifs peuvent être utilisés dans des stru tures qui en for ent l'interprétation événementielle (43). Des verbes événementiels peuvent être utilisés dans des stru ture qui en for e l'interprétation stative (44). Des verbes peuvent être lus soit omme événementiels soit omme statifs à parité de stru ture syn- taxique et d'éléments lexi aux (45). Ce dernier point de vue théorique as est le plus intéressant du ar il permet de déterminer les éléments, internes à la stru ture lexi ale, qui sont à l'origine de la stativité. (43) Ce (44) Daria (45) afé est en train de plaire beau oup à Daria. ourt des Marathons. a. Les arbres entourent le hâteau. b. Les soldat entourent le hâteau. Il faut souligner que, ontrairement à rature, le verbes statifs ne e qui a été onstaté dans la litté- onstituent ni un groupe homogène ((Rappaport Hovav & Levin 1998 ; Harley 1995 ; Ram hand 1998), ni un primitif aspe tuel (Pylkkänen 2000 ; Rothmayr 2006). Par onséquent, des stru tures diérentes pour la ma ro- atégorie de statifs peuvent être supposées. Composantes morphologiques La deuxième partie de la thèse porte sur les verbes parasynthétiques à base adje tivale dont la paraphrase est faire l'objet A, faire l'objet plus A, où A orrespond à la base adje tivale du verbe. L'étude a ir ons rit 221 verbes italiens dont la paraphrase est faire l'ob- jet A, faire l'objet plus A et qui rentrent dans parmi eux a e are `aveugler', �� `rendre moins brut' addol ire `dou ir', e type (appelés DPVs), ingiallire `jaunir', sgrezzare . La se tion 7.3.1.1 démontre que la nature de la base verbale est élément non  ation. Pour elle de atégorisé (ra ine) grâ e à deux indi es : l'é helle et la modie qui on erne la première, si la base était �� s'attendrait à voir une inuen e de l'é helle �� La liste omplète est dans l'appendi e. �� L'é helle est dénie atégorisée, on adje tivale sur la sémantique omme :  a pair < S, � δ > onsisting of a set of obje ts and an asymmetri ordering relation along some dimension δ  (Kennedy & M Nally 2002 : 8). xxxiii verbale. Les DPVs, d'un oté ne sont pas formés sur une lasse spé ique d'adje tifs (Tableau 7.1 à la page 142), de l'autre, ils ne montrent pas de omportements diérents liés à la présen e d'adverbes de degré mente (46) ` omplètement', Il ghia io ha parzialmente ompleta- `partiellement'. ompletamente infreddolito i bambini. La gla e a omplètement refroidi les enfants. (47) La pioggia ha infradi iato pazialmente i panni stesi. La pluie a mouillé partiellement le linge mis à sé her. (48) Il sole ha ompletamente arrostito Giovanni. Le soleil a omplètement rti Jean. (49) La vin ita al Lotto ha parzialmente arri hito Maria. Le gain au Loto a partiellement enri hi Marie. En outre, les DPVs ne peuvent pas être formés sur des bases modiées. Le fait que les morphèmes modi ateurs ne puissent pas apparaître dans la base verbale est signe du fait que la base n'est pas (50) atégorisée. bello - bellissimo - *abbellissimare beau -extrêmement beau - faire extrêmement beau (51) grande - grandissimo - *ingrandissimire grand - extrêmement grand- faire extrêmement grand Les résultats des tests montrent lairement que la détermination du type d'é helle de la base verbale n'est pas possible. Pour onsidérée ela, la base verbale est omme une ra ine, dont la sémantique n'a pas été limitée par le atégorisateur. La stru ture proposée est don (52) la suivante. XP √ NP L'autre P N √ il bambino bello omposante morphologique des DPVs est le préxe, dont la dis- tribution parmi les trois lasses ( a-, im-, s-) est reportée dans le Tableau 7.3 (à la page 144). En a ord ave S alise (1990), les préxes sont sables de l'introdu tion de la sémantique onsidérés as, les DPVs peuvent alterner, sans eets sur le sens, ave pondants formés à l'aide du suxe ausatif xxxiv omme respon- ausative. En eet, dans -izzare ou les verbes -i are. ertains orres- (53) lombardo - lombardizzare Lombard - lombardiser (54) virtuale - virtualizzare virtuel - virtualiser (55) illombardire (attendu) faire lombard (56) invirtualire (attendu) faire virtuel Pour ela, ils sont positionnés dans la tête fon tionnelle r, qui est respon- sable de l'introdu tion de la sémantique ausative. Rles du sujet D'après Kratzer (1996), le sujet n'est plus onsidéré omme introduit par le verbe lexi al, mais plutt par une tête fon tionnelle appelée doit être sémantiquement a entraîne deux aktionsart l' ordée ave aktionsart l' Voi e. Voi e du verbe lexi al. Cela onséquen es importantes : le sujet n'a au une inuen e sur du verbe ar la relation est de type as endante (de aktionsart le rle du sujet est un reet de l' v à Voi e) ; du verbe. La se tion 7.4 reporte la méthodologie et les résultats d'une expérien e qui enquête sur de possibles diéren es d'a sujets inanimés des DPVs italiens. En essibilité entre sujets animés et on lusion, au une diéren e d'a essi- bilité liée à l'(in)anima ité des sujets n'a été mise en éviden e. On peut don on lure que les DPVs sont plausibles tant ave des sujets animés qu'ave des sujets inanimés. Diérentes atégories des DPVs et leurs aktionsarten La se tion 7.5 analyse les diérentes sémantique de la ra ine atégories de DPVs en relation ave la onstituant la base verbale. Selon la qualité dé rite par la base, on peut distinguer trois groupes de DPVs : de forme, de super ie et psy hologique. Le premier groupe, omme un allargare onstitué par les DPVs qui ont une base de forme `élargir', appiattire `aplatir', rimpi iolire `réduire', dé rit hangement physique qui a lieu sur l'objet dire t. En d'autres termes, l'objet subit un hangement dans l'une de ses ela entraîne en une diéren e ristique prise en ara téristiques intimes, et lairement identiable, un δ sur une ara té- ompte par la base. Par exemple, si un mur est large de dix entimètres et qu'il a été élargi de a été produite sur une des inq entimètres, une diéren e physique ara téristiques fondamentales du mur, sa largeur. xxxv Le deuxième groupe, omme imbian are onstitué par les DPVs qui ont une base de surfa e `blan hir', insozzare `salir', annerire `noir ir', dé rit un hangement externe à l'objet même qui ne modie pas vraiment une de ses ara téristiques internes. Si un mur est peint en bleu, on peut pas identier lairement un δ sur la ouleur du mur, il n'y a pas eu une modi ation intrinsèque du mur. Le troisième groupe est ra ine psy hologique, intristire, onstitué par les verbes qui ont omme instupidire, `abrutir', `attrister'. Ce groupe n'est pas pris en omme base une rimbe illire, `abrutir' et ompte dans l'étude. La se tion 7.6 utilise les diagnosti s de stativité dé rits par la se tion 6 an de mettre en éviden e les diérentes le tures générées par les DPVs de forme et les DPVs de surfa e. En parti ulier, les diagnosti s suivantes sont utilisés : interprétation sous verbe modal (57) à (60) ; interprétation de l'adverbe la ontribution à la già dovere `devoir' - exemples `déjà', - exemples (61) à (62) ; haîne narrative - exemples (39) à (40) ; et les adjoints instrumentaux possibles - exemples (65) à (66). Ces diagnostiques soulignent omment les deux groupes n'ont pas les même possibilités de générer une le ture stative. Seuls les verbes de surfa e peuvent la générer. Un résumé est présenté dans le tableau 1. (57) a. Giovanni deve allargare il muro entro domani an hé il lavoro sia nito. G doit élargir le mur avant demain an que le travail soit terminé. b. L'umidità deve allargare il muro entro domani an hé il lavoro sia nito. L'humidité doit élargir le mur avant demain an que le travail soit terminé. (58) a. Giovanni deve appiattire il us ino entro inque minuti per an- dare a letto. G doit aplatir le oussin dans inq minutes pour aller se ou her. b. I ollant devono appiattire il sedere di Giovanna in un'ora anhé possa andare alla festa. Les ollants doivent aplatir le derrière de G. dans une heure an qu'elle puisse aller à la fête. (59) a. Il pittore deve imbian are la tela entro domani per nire il lavoro. Le peintre doit blan hir la toile avant demain pour terminer le travail. xxxvi b. *La pittura deve imbian are la tela entro domani per nire il lavoro. *La peinture doit blan hir la toile avant demain pour terminer le travail. (60) a. Il delinquente deve insozzare la porta entro due minuti an hé il lavoro sia nito. Le délinquant doit salir la porte dans deux minutes an que le boulot soit terminé. b. *Il fango deve insozzare la porta entro sabato an hé il lavoro sia nito. *La boue doit salir la porte avant samedi an que le boulot soit terminé. (61) a. *Giovanni allarga già il bu o del salotto. G élargit déjà le trou dans le salon. b. *L'umidità allarga già il bu o della u ina. *L'humidité élargit déjà le trou dans le salon. (62) a. Il pittore imbian a già la tela del Caravaggio. Le peintre blan hit déjà la toile du Caravaggio. b. La pittura imbian a già la tela del Caravaggio. La peinture blan hit déjà la toile du Caravaggio. (63) a. Daria è arrivata, ha ingrandito il bu o e si è seduta sul divano. Daria est arrivée, elle a agrandi le trou et s'est assise sur le anapé. b. La mua si è formata, ha ingrandito il bu o ed è morta. La moisissure s'est formée, elle a agrandi le trou et est morte. (64) a. Daria è arrivata, ha imbian ato la tela del Caravaggio e si è seduta sul divano. Daria est arrivée, elle a blan hi la toile du Caravaggio et s'est assise sur le anapé. b. La verni e è stata stesa, ha imbian ato il muro e ha s hiarito la stanza. La peinture a été étalée, elle a blan hi le mur et é lair i la piè e. (65) a. ? ?La mua ha allargato il muro on le sue spore. ? ?La moisissure a élargi le mur ave ses spores. b. La mua ha allargato il muro a ausa delle (sue) spore. La moisissure a élargi le mur à ause de ses spores. xxxvii (66) a. ? ?La nebbia ha allungato la rotta on la sua densità. ? ?Le brouillard a allongé la route ave sa densité. b. La nebbia ha allungato la rotta a ausa della (sua) densità. Le brouillard a allongé la route à ause de sa densité. Les résultats des tests de stativité montrent que les verbes de forme peuvent générer seulement une le ture événementielle. Au ontraire, les verbes de surfa e peuvent générer une le ture stative ou une le ture événementielle. Ces deux le tures sont mises en éviden e par l'(in)anima ité du sujet : un sujet animé est lié à une le ture événementielle �� , un sujet inanimé est lié à une le ture stative. �� Même s'il faut souligner que, omme on s'y attend, un sujet animé peut être lu omme inanimé, en générant une le ture stative. xxxviii Dovere Già Contrib. Temp. Adjoints Animé deontique * � on, instruments Inanimé deontique * � a ausa, instruments DPVs de surfa e Animé deontique * � on, instruments Inanimé épistemique � on, not instruments Table 1 : Resumé des tests de stativité (DPV). xixxx DPVs de forme La se tion 7.7 analyse au moyen de paraphrases si les DPVs des deux groupes présentent une sémantique ausative. Si, d'un té, les DPVs de forme ne posent au un problème en étant toujours événementiels, de l'autre té, les DPVs de surfa e, en pouvant être interprétés nementiels, posent un dé. En eet, la n'est pas souvent prise en par eux qui omme statifs et évé- oexisten e de stativité et ompte par les ausalité adres formels génératifs, surtout onsidèrent la stru ture argumentale omme étant un produit de sous-événements. Les paraphrases reportées dans la se tion 7.7 montrent que soit les DPVs de formes (67), soit (67) a. eux de surfa e (68) ont une sémantique Giovanni ha allargato il bu o. il fatto → G. ha fatto qual ausative. osa per ausare he il bu o sia più largo di prima. G a élargi le trou. → G. a fait quelque hose pour auser le fait que le trou soit plus large. b. → L'umidità ha allargato il muro. per ausare il fatto L'umidità ha fatto qual osa he il muro sia largo. L'humidité a élargi le mur. → L'humidité a fait quelque hose pour auser le fait que le mur soit plus large. (68) a. Il pittore ha imbian ato la tela. per ausare il fatto → Il pittore ha fatto qual osa he la tela sia (più) bian a. Le peintre a blan hi la toile. → Le peintre a fait quelque hose pour auser le fait que la toile soit (plus) blan he. b. La pittura ha imbian ato la tela. tela ha ausato il fatto → L'esistenza della verni e sulla he la tela sia bian a. La peinture a blan hi la toile. → L'existen e de la peinture sur la toile a ausé le fait que la toile soit blan he. Causalité et for e-dynami s Les relations de ausalité sont exprimées diéremment dans les langues, no- tamment par : des moyens morphologiques ; des moyens syntaxiques ; sans moyens spé iques. L'étude des deux premiers mantique Pour ausative entraîne un as nous montre que la sé- hangement dans la stru ture argumentale. ela, l'étude de l'expression de la ausalité est intrinsèquement liée à l'étude de la stru ture argumentale. Les se tions pré édentes ont souligné que des phrases omme (69) et (70) génèrent deux le tures aspe tuelles diérentes. Grâ e à l'emploi des adjoints, nous avons pu voir que les haînes ausatives sont elles aussi diérentes (71) et (72). xl (69) Daria embellit la (70) Les photos embellissent la (71) L'enfant égaye la fête ave (72) La musique égaye la fête ave Pour rendre ompte de for e-dynami mel de hambre (ave es des tableaux). hambre (ave les petites leurs ouleurs). hansons. son rythme/*ave omportements, le stéréo. ette thèse utilise le adre for- (FD). Cette appro he formelle est née dans les s ien es ognitives, mais peut être mise à prot en linguistique formelle (Copley & Harley 2015 ; Copley & Wol 2014 ; Copley 2015) par e qu'elle identie les parties onstitutives qui sont responsables de la rit d'une façon détaillée ette appro he, et en propose une extension d'expliquer les verbes statifs abdu tion, une for d' ausalité. La se tion 7.8.1 déapable ausatifs. Notamment, il introduit le on ept e virtuelle qui est introduite dans le système par le lo- uteur et qui est responsable de la �� de la situation (Sour e et Thème En étudiant les prin ipes réation du lien ausal entre les individus ). ognitifs qui déterminent les diérents patterns argumentaux, l'appro he FD est arrivé à identier que l'élément fondamental est la transmission de for e d'un parti ipant à l'autre. La ausalité est don une intera tion asymétrique entre entités. Ces entités peuvent avoir des tendan es de type diérent : au mouvement ou à la stase. Les tendan es des entités impliquées dans la situation se somment et donnent lieu à la tendan e ontraire à ment, à se fermer ; elle de la porte : Daria immobile, applique don porte, (73) et ausalité. Par exemple, en (73) la porte Daria, ontraire à même en restant elle appliquée par la ela entraîne dans l'état résultant de la porte ouverte. Daria ouvre la porte. Les avantages de l'appro he FD sont plus stase a une a une tendan e au mouve- a une tendan e à la stase. une for e de sens Daria omme garder (74a), ou rester lairs dans le as des verbes de (75a). Même en étant événementiel, e type de verbes n'implique pas des événements, voire la grammati alité des périphrases progressives (74b et 75b). Les ausalité (74) (75) adres formels qui analysent la omme un sous-événement ont des di ultés à en rendre a. Daria garda la porte ouverte. b. Daria est en train de garder la porte ouverte. a. Daria resta au lit toute la matinée. �� Sujet et omplément d'objet xli ompte. b. Daria est en train de rester au lit. Pour traduire en linguistique les éléments appartenant à la for e-dynamique des s ien es ognitives, il faut que En d'autres termes, les for es es éléments aient une validité linguistique. ognitives doivent être dis riminées dans la langue. Les expérien es de Wol (2003) ont montré que diérents types d'événements ausatifs dans le monde sont dé rits par diérents moyens linguis- tiques. Notamment, des haînes ausatives lexi ales, et des ausatives indire tes (77) donnent lieu à des haînes ausatives dire tes (76) à des ausatives syntaxiques. (76) Daria ouvrait la porte (*par Pierre). (77) Daria t ouvrir la porte (par Pierre). La se tion 7.8.1.1 résume les prin ipes fondamentaux de l'appro he FD né essaires à sa tradu tion dans la linguistique formelle (Copley & Harley 2015 ; Copley 2015 ; Copley & Martin 2014). En parti ulier, les deux prin ipes de base sont : la for e linguistique (f ) et la situation linguistique (s). Une for e linguistique (f ) est temporellement et spatialement située et se produit à partir des individus présents dans la situation et de leurs propriétés. omme :  a fun tion from an initial linguisti situation s to the ( eteris paribus, linguisti ) nal situation s', whi h orresponds to a on eptual net for e ϕ. The latter is a (mental representation of ) an input of energy that arises from all the individuals and their property attributions in a on eptual situation σ  (Copley & Harley 2015 : 15). C'est une fon tion de type hs, si, de situation à situation. Elle est dénie Une situation linguistique (s) est formée des objets et de leurs propriétés (Barwise & Perry 1983 : 7 ss.), elle est délimitée par le lo uteur dans son a on eptual situation σ , whi h is a spatio-temporally bounded annotated snapshot of individuals and their property attributions  (Copley & Harley 2015 : 14). Elle est de type situation, hsi. a te de langage. Elle est dénie Ave omme :  es deux moyens formels, l'appro he formelle FD est formaliser les verbes forme (don événementiels) omme appesantire par (79). (78) apable de ausatifs événementiels. Par exemple, les verbe DPVs de Giovanni anneris e la stanza. John bla kened the room. xlii `alourdir' (78) sont formalisés V oice (79) auser Pheti V oice DP auserhe,f ti ' Giovanni Voi e v Phf ti auserhf t,he,f tii v be omehst,f ti rPhsti DP r a- la stanza √ nera La ontribution sémantique des têtes fon tionnelles petit vbe ome et V oicea est : (80) �vbe (81) �V oicea ome � = λp λf. p(f in(f )) tive � La sémantique de = λπ λx λf. π(f ) & source(x, f ) Voi e est dynamique, puisqu'elle est de type hft, hhe,ftii et elle séle tionne proprement l'argument externe qui est dénommé Sour e dans la terminologie originale de Copley & Harley (2015), et i i Causer. Ces moyens formels se révèlent inadéquats pour la formalisation des verbes statifs. En eet, la présen e d'une for e énergétique génère automatiquement des verbes événementiels, tique (et ar elle garantit le passage d'une situation linguis- ognitive) à l'autre. D'autres moyens pour ins rire les verbes statifs ausatifs dans l'appro he FD sont requis (se tion 7.8.2.2). Tout premièrement, il faut souligner la diéren e entre ausalité (Copley & Harley, 2015). Si, d'un né essairement un événement pas for ément un ausatif, de l'autre hangement. Cela est stase de (74) et (75), où au un Dans l'étude présente, on té, un té, la ausalité n'implique lairement présenté par les verbes de hangement est produit. onsidère qu'il y a du individu n'est pas dans le même état à deux moments le hangement et hangement implique hangement quand un t1 et t2 . Par onséquent, hangement est intimement lié au temps. Il en dérive que les prédi ats événementiels, étant les seuls à pouvoir faire avan er le temps de référen e, sont les seuls à pouvoir donner lieu à des xliii haînes ausales. tive Nous avons vu que les prédi ats statifs également peuvent impliquer la ausalité. Nous avons avan é l'hypothèse qu'il soit dû à la présen e d'une phrase réduite (SC) dans la partie basse de la dérivation, de la même manière que pour les verbes événementiels ausatifs. La SC est don responsable seulement de la dénition d'un état de l'objet dire t. Le fait que pour les verbe événementiels, l'objet dire t entre dans un état nouveau (déterminant un hangement), et pour les verbes statifs, l'objet dire t est dans un état (sans en déterminer un Pour résumer, nous hangement), est du au type de tête fon tionnelle royons que la présen e d'une rP dans la partie basse de la dérivation est responsable de la sémantique ausative et que le hange- ment est déterminé par la présen e d'une tête verbale événementielle qui est apable de faire progresser le temps de référen e et don l'instauration de deux états pour le même individu en Le fait que les v. on epts de hangement et de t1 et vbe ome de permettre t2 . ausalité apparaissent sou- vent ensemble dans la langue est dû au fait que dans la réalité nous sommes apables de voir les liens de ausalité entre individus grâ e au hangement. Par exemple, si quelqu'un tou he un bouton et que juste après la lumière s'allume, on peut ton est onstater  Quelqu'un a allumé la lumière , même si assé et que l'allumage est onséquen e d'un pi le bou- éle trique. Certaines ontraintes doivent être respe tées pour pouvoir établir une relation de salité, notamment : une priorité temporelle, une ontiguïté spatiale et une au- ontiguïté temporelle, une ovarian e (Hume 1739/1969, 1748/1955). Des exemples où un lien de ausalité erroné est établi, ainsi que les ré- sultats des expérien es de Thorstad & Wol (2016) à propos de l'illusion de ausalité et de la per eption de ausalité sont reportés à la se tion 7.8.2.1. Nous avons démontré que la présen e d'un ment la présen e de la présen e de la ausalité, toutefois le hangement implique for éontraire n'est pas vrai : la ausalité ne détermine pas for ement un hangement. Causalité sans hangement Après avoir déni l'autonomie de la la thèse se poursuit ave ausalité sans à ausalité par rapport au la se tion 7.8.2.2 qui prend en hangement, 'est-à-dire le as de la hangement, ompte le as de la ausalité statique (opposée ausalité énergétique). La se tion 7.8.2.2 démontre la non-appartenan e de la ausalité statique au modèle FD. En outre, elle montre l'importan e de la psy hé dans les expressions linguistiques. La langue distingue entre phénomènes qui ont une réalité physique dans le monde et phénomènes qui ont une réalité psy hologique. Elle le distingue par des moyens morphologiques. Cela est très important pour pouvoir établir une autre type de xliv ausalité, elle statique. Ce type de ausalité est notamment elle ensé prendre ses origines d'une réalité psy hologique, réée par la psy hé du lo uteur. On imagine que la personne qui pronon e (82) établit un lien entre individus dans une même situation, entre du lien entre individus et, en the drape et the room. La subje tivité onséquen e, la nature génératri e de la psy- hé du lo uteur est soutenue par le fait que plusieurs personnes peuvent Cela est Cela est un ensemble insensé de dé rire la même situation par des moyens linguistiques diérents : une hambre, hoses. ou Cela est un magasin, ou La relation entre sujet et objet n'est pas elle de Figure-Ground, omme proposé par Ram hand (2008 :55) pour les ILPs. (82) The drape darkened the room. Le drap a assombri la hambre. (83) Mary darkened the room. Mary a assombri la hambre. Le fait que la la ausalité statique ait un statut bien diérent par rapport à ausalité énergétique est mis en éviden e par l'impossibilité de hangement sur l'état de la hambre. Pour onstater un the room e qu'on sait de (82), est sombre maintenant, elle était sombre dans le passé et elle sera sombre dans le futur. Au un hangement de la hambre n'est exprimé linguistiquement. En outre, au une for e énergétique est présente. Cela entraîne que, à une phrase ontrairement omme (83), il n'y a au une transition de situation. Le modèle FD représenté en (84) ne peut pas être appliqué. ��� ��� ��� ��� ��� ��� (84) Un autre modèle doit être formulé pour donner représentation aux statifs ausatifs. Notamment, e modèle ne peut pas impliquer deux situations, étant donné l'absen e de for es énergétiques qui pourraient garantir le passage d'une situation initiale à une nale. xlv La se tion 7.8.2.2.2 avan e l'hypothèse que la la présen e d'une seule situation qui ausalité statique implique ontient un individu et ses propriétés. Car les DPVs de surfa e sont des verbes à deux arguments, ils impliquent s deux situations ( a et et sb ), ha une x ontenant un individu et ses propriétés ( y ). � � �� �� �� (85) Il faut se demander quel est l'élément responsable de l'instauration du lien de ausalité entre les deux individus, étant donnée l'absen e de for es énergétiques. Nous proposons que le lien de ausalité est introduit par la personne qui pronon e la phrase. En eet, au un rapport de ausalité entre les deux individus de (82) n'arrive dans le monde. Une personne est responsable de son établissement, grâ e à sa a une relation apa ité abdu tive. Elle ausale entre les deux individus, pour onsidère qu'il y e qu'elle onnaît du monde. [o℄ne morning you enter the kit hen to and a plate and up on the table, with bread rumbs and a pat of butter on it, and surrounded by a jar of jam, a pa k of sugar, and an empty arton of milk. You on lude that one of your house-mates got up at night to make him- or herself a midnight sna k and was too tired to lear the table. This, you think, best explains the s ene you are fa ing. To be sure, it might be that someone burgled the house and took the time to have a bite while on the job, or a house-mate might have arranged the things on the table without having a midnight sna k but just to make you believe that someone had a midnight sna k. But these hypotheses strike you as providing mu h more ontrived explanation  (http://plato.stanford.edu/entries/abdu tion/ , DouL'abdu tion entre en jeu quand, par exemple :  ven : 2011). L'abdu tion est présente quand une personne établi un lien entre deux entités, en royant que la propriété de l'une d'entre elles est responsable d'une des propriétés de l'autre, sans qu'un lien physique de ausalité puisse être ren ontré. Cela a deux pour onséquen es importantes : sans la présen e d'une personne onstater et établir une relation de ausalité entre deux individus, la ausalité statique n'existe pas ; la présen e dans la situation ognitive des deux individus est obligatoire. Par exemple, une situation sans la présen e d'un anapé ne peut pas onduire à la phrase (86). xlvi (86) La Le anapé assombrit la piè e. ausalité statique ne produit pas de visibles, une personne ne peut don deux individus à moins que omme seul un individu pas tirer des liens de ausalité entre eux- i soient présents. De la même manière, ons ient est sa présen e rend possible la hangements, il n'y a pas d'eets apable de produire l'abdu tion, seule réation d'un lien de le anapé ausalité entre la piè e. La et ausalité stative est le lien entre deux propriétés de deux individus faite par la apa ité abdu tive d'un être pensant. Le lien n'est pas physique, mais il est instauré par la psy hé de l'être pensant. Quand une personne établit un lien de ausalité entre sa et sb de (85), elle établit qu'une propriété non-pré isée de x est responsable d'une propriété de y, qui est représentable par p(y). Cela est possible (dans tables e as : x et y) ont des ara téristiques parti ulières qui sont interpré- omme anté édents. Par exemple, le béton ara téristiques qui peuvent être interprétées jaune, ontrairement à ar les objets le guano. en (87) ne possède pas de omme anté édents de l'arbre Cela explique la diversité sémantique entre les deux exemples. (87) ? ?Le béton jaunit l'arbre. (88) Le guano jaunit l'arbre. En eet, le guano a une tendan e vers les arbres jaunes qui le béton ne possède pas. Pour résumer, la ausalité statique et la ausalité énergétique se dis- tinguent par deux fa teurs fondamentaux. Premièrement, l'élément générateur de la ausalité statique est l'abdu tion, énergétique dont il est la for e énergétique. Par statique, les situations sont né essairement présenté par s1 l'être dans la ontrairement à la ausalité onséquent, dans la ausalité ontemporaines, omme il est re- en (85). Deuxièmement, les tendan es des individus sont vers ausalité statique et vers l'agir pour la ausalité énergétique. L'absen e d'une for e énergétique explique la variété de prols argumentaux pris par les verbes statifs. La manque de for e énergétique (qui, étant une for e ve torielle, est a- ra térisée par un vers) permet aux humains d'utiliser diérents prols argumentaux pour exprimer une même situation ognitive statique. Ils utilisent e ou pour pouvoir mettre en éviden e diérents éléments, rar hie argumentale m'est produite dans la réalité. (89) a. Les photos sont sur le mur. xlvii ar au une hié- (90) b. Les photos embellissent le mur. a. Daria a peur des b. Les au hemars. au hemars eraient Daria. Dérivations La se tion 7.9 reporte les dérivations de l-sémantique et l-syntaxe des DPVs événementiels et statifs. Les deux atégories de verbes ne se distinguent ni dans leur ( omme on a déjà vu, elles sont ausatives à ausalité ause des préxes), ni dans la nature de leurs bases (qui sont des ra ines). Elle se distinguent par la sémantique de la tête fon tionnelle DPVs événementiels ont vbe ome . v : les DPVs statifs ont vrelation ; les v détermine également Voi e, qui est responsable V oicesour e ou V oice auser . La sémantique de une sémantique diérente de la tête fon tionnelle de l'introdu tion de l'argument externe : (91) �vbe (92) V oice (93) Giovanni anneris e la stanza. ome � = λp λf. p(f in(f )) auser = λf.causer(x, f )p(f in(f )) Jean noir it la piè e. V oice (94) auser Pheti V oice DP auserhe,f ti ' Giovanni Voi e v Phf ti auserhf t,he,f tii v be omehst,f ti rPhsti DP la stanza r a- √ nera Il faut remarquer que la tête fon tionnelle qui assure la relation vrelation est une tête prédi ative ausale entre une propriété du sujet et une propriété de l'objet. xlviii (95) �vrelation� = λp λs. p(s) (96) V oicesour (97) Il divano anneris e la stanza. e = λs.source(x, s)p(suc(s)) Le anapé noir it la piè e. V oicesour e Phsti (98) V oice′sour e he, sti DP il divano Voi esour v Prelationhsti ehst,he,stii v relationhst,sti rPst DP r la stanza La diéren e entre statifs a- √ nera ausatifs et statifs non- ausatifs est prise en ompte à la se tion 7.10. En parti ulier, nous faisons que la ausalité est générée par la présen e d'une tête fon tionnelle relationnelle rP (S häfer, 2008) dans la partie basse de la dérivation. La diéren e entre un verbe statif ausatif et un verbe statif non- ausatif est la présen e d'une phrase réduite dans la l-syntaxe du premier. Cela est onrmé par le fait que la tête fon tionnelle verbale des DPVs de surfa e est une tête prédi ative, de la même manière que par les verbes statifs non- ausatifs. Paramètre de goût personnel La se tion 7.11 analyse la présen e d'un paramètre pragmatique de goût dans les DPVs. Le paramètre de goût est introduit par un prédi at de goût personnel, il représente une opinion et pas une question de faits obje tifs (Laherson 2005). Cela est évident en (99) et (100) où l'armation peut être relativisée par la question : pour qui ?. (99) (100) Le gâteau est bon. La voiture est belle. xlix Dans les DPVs le paramètre de goût personnel est introduit par la ra ine verbale. DPVs statifs et événementiels se distinguent par les parties que e paramètre peut relativiser. Ce phénomène est visible grâ e au re ours aux tests de désa ord (Stephenson 2007), dans lequel les prédi ats de goût per- sonnel admettent une ontradi tion (101), ontrairement aux autres types de prédi ats (102). (101) (102) A : La voiture de Daria est super. B : Oui, elle l'est. C : Non, pas vraiment. A : La voiture de Daria est rouge. B : Oui, elle l'est. C : # Non, pas vraiment. Le test de désa sarten ord peut être appliqué aux DPVs de diérents aktion- et il met en éviden e que le paramètre de juge peut relativiser toutes les parties dans le ase de DPVs statifs (105 et 106). Il relativise la partie resultative dans le as de DPVs événementiels (103 et 104). (103) A. Cosa fa Giovanna ? Qu'est que fait Jeanne ? B. Abbellis e la stanza. Elle embellit la piè e. C. Oh no, non la abbellis e per niente, quei quadri sono disgustosi ! Oh non, elle ne l'embellit pas du tout, es adres sont dégoûtants. (104) A. Cosa fa Giovanna ? Qu'est que fait Jeanne ? B. Abbellis e la stanza. Elle embellit la piè e. C. ? ? ?Oh no, non fa niente !/Oh, no, lava i piatti ! Oh non, elle ne fait rien !/ Oh non, elle lave les assiettes. (105) (106) A. Qu'est que font es eurs sur la table ? B. Elles l'embellissent. C. Oh non, elles ne l'embellissent pas du tout. A. Pourquoi la table est-elle ainsi ? B. C'est à C. Oh non, ause des eurs. e n'est pas pour ça, l 'est à ause de la lumière. La possibilité du paramètre de juge de relativiser aussi la partie dans le as de DPVs statifs est déterminée par la nature même de la stative. La ausative ausalité ausalité stative est générée par l'opinion personnelle du lo uteur. Cela permet à un autre lo uteur de mettre en question Sour e et Thème. En e lien ausatif entre e sens là, la nature d'abdu tion est révélée. La stativité peut être déte tée automatiquement Le hapitre 8 dé rit les étapes qui ont servies à la réalisation, de manière automatique, d'un gradient de stativité des verbes anglais. Il une onsiste en ollaboration à un projet plus vaste (CNRS-SFL et Emory University) qui a omme but l'identi ation automatique de l'orientation temporelle de phrases de orpus. Nous avons vu dans le hapitre 6 que les verbes statifs entraînent des ontraintes temporelles diérentes des verbes événementiels. Il est lair que l'identi ation des verbes statifs est fondamentale pour un projet qui porte sur la dénition automatique de l'orientation temporelle. Il y a deux façon de pro éder. La première onsiste à lister à la main les verbes anglais qui normalement ont une sémantique stative. La deuxième onsiste à traduire les diagnosti s les plus e a es en règles qui soient préhensibles par un parser (dans e les points forts et les faiblesses de as : Tregex). Le om- hapitre 8 dé rit haque appro he et montre omment la deuxième est préférable. Le hapitre se poursuit en expliquant les diagnosti s périphrase progressive, alternan e ausative et hoisis, notamment simple present, et en reporte leurs tradu tions pour Tregex. Dans le but de générer un seul gradient de stativité, les résultats obtenus par la fouille dans un être normalisés. Il faut elles orpus de Twitter des règles Tregex doivent omprendre quelles sont les règles les plus puissantes, apables d'identier un verbe statif d'une manière e a e. Pour ela, des données humaines (se tion 8.2.2) ont été re ueillies au moyen d'un test d'interprétation sémantique soumis à 25 lo uteurs natifs de l'anglais. Les résultats obtenus ont été normalisés par une fon tion de régression logistique entre jugements humains et valeurs obtenues dans la fouille de i rend une équation qui assigne un poids à être in luse dans le projet prin ipal. li orpus. Celle- haque règle Tregex et qui peut Con lusions La thèse porte sur la stru ture argumentale de deux types de verbes parasynthétiques italiens. Dans sa première partie elle s'o de la upe de la dénition de la grammati alité onstru tion pseudo-résultative en italien et en français. Pour ela, on a re ouru à la ré olte des données de lo uteurs natifs des deux langues. Une réexion sur les méthodologies expérimentales de la grammaire générative fait partie des études préliminaires. La deuxième partie porte sur la dénition de l'existen e et de la représentation de la Le dernier ausalité statique. hapitre applique ertaines dé ouvertes de l'étude au domaine du traitement automatique du langage naturel. lii Introdu tion The single engine hypothesis (Marantz 1997, .; Harley 2005; gues for the existen e of only one single linguisti is responsible for the same synta ti inter al.) ar- generative engine whi h reation of both senten es and words by means of the rules. Thus, the internal stru ture of words, the of morphologi al building blo ks, is synta ti ombination in nature. Therefore, the study of words and their stru ture is useful to the general synta ti dis ussion. Furthermore, sin e the lexi al-synta ti stru ture of verbs (l-syntax) ontains fun tional heads responsible for the introdu tion of verbal arguments, the study of derived verbs is parti ularly informative about syntax. In fa t, derived verbs an lead to the identi ation of the role and the merge position of morphologi al omponents with respe t to the arguments of the verb (i.e. Hale & Kayser 2002). This work fo uses on morphologi al derived Italian verbs, namely parasyntheti verbs (Ia obini 2004, ess of derivation inter al.). The parasyntheti morphologi al pro- reates verbs, adje tives and nouns in almost all Roman e languages, it is in fa t a Latin pro ess maintained in its histori al evolution. The label parasyntheti verb identies morphologi al produ ts hara ter- ized by the simultaneous presen e of a prex and a sux and the la k in the lexi on of intermediate derivational steps. lasses, distinguished for the It in ludes many dierent sub- ategori al nature of their base (adje tive or noun) and for the semanti s they generate ( ausative, lo ative, ...). Chapter 3 proposes a general overview of the whole and morphologi al issues in order to lass, with histori al referen es ontextualize verb sub- lasses studied in this work. Parasyntheti verbs are interesting for a general dis ussion about argu- ment stru ture and lexi al-syntax building blo ks be ause their morphology is parti ularly transparent, ex eption made as we will see for the omplex prex-sux nature. The present work analyses two sub-groups of parasyntheti arise spe i theoreti al issues depending on the base, nominal or adje tival. verbs, whi h ategori al nature of their The dierent nature of theoreti al 1 on erns 2 involved leads to the two distin t parts of the present work. The rst part deals with the l-syntax of denominal verbs and their role in pseudo-resultative onstru tion (Levinson 2007). Thus, it involves the general dis ussion about grammati ality of se ondary predi ations in Roman e languages, parti ularly in Italian and Fren h. The se ond part aims to elu idate the nature of stativity and stative verbs. It is divided into three hapters: (i) stativity diagnosti s, (ii) l-syntax of ausative deadje tival parasyntheti verbs and their double aspe tual readings (eventive and stative), (iii) an appli ation of synta ti and semanti stativity diagnosti s for the automati extra tion of temporal orientation of senten es. Chapter 1 ontains an introdu tory se tion about methodology. A small ontribution to the debate about methodology in generativism is proposed, fo using parti ularly on experimental proto ols of data olle tion and pos- sible biases produ ed by the employment of small experiments and autoanalysis. The produ tion of reliable data allows more protable inter-dis iplinary ex hanges with other s ien es investigating languages and language fa ulty. Chapter 2 summarizes synta ti frameworks about synta ti stru ture of ausative and stative verbs. Furthermore, it spe ies the framework and the theoreti al assumptions whi h are going to be employed in this study. The rst part is entitled non-ambiguous verbs, as the verbs in question do not entertain dierent aspe tual readings. Chapter 4 on erns the general domain of se ondary predi ations (i.e. weak and strong resultatives) in Roman e languages, with parti ular attention to Italian. The belong to the impli it hapter demonstrates that denominal parasyntheti s reation verb sion about pseudo-resultative Results of a linguisti larly if the onstru tion. questionnaire show that pseudo-resultative opinions lass whi h is fundamental to the dis us- ondu ted on Italian native speakers onstru tion is grammati al in Italian, parti u- onstru tion involves a pronominal dire t obje t. Sin e informal olle ted after the questionnaire point out that speakers prefer a orresponding adverb to the pseudo-resultative adje tive, a magnitude estimation task (Bard, Robertson, Sora e 1996) has been the opinion of informants, showing the higher a ondu ted. It onymous to pseudo-resultative adje tives. We will see that adverbs two s opes, as expe ted when o onrms eptability of adverbs syn- ur with resultative verbs. an have Adverbs s ope either over the result proje tion, being adjun ts of SC, or over the eventive proje tion, being adjun ts of little v )P. The availability of pseudo-resultative onstru tion has been tested for Fren h by means of a stru tured questionnaire ( hapter 5). The results afrm the parti ularity of Italian with respe t to se ondary predi ation within 3 the Roman e panorama. The pseudo-resultative onstru tion is mu h less available in Fren h than in Italian. A possible explanation to this varian e resides in the Fren h lower phonologi al orresponden e between verbs and base nouns. This makes harder for speakers to a ess the base noun. Con- sequently, it leads to the impossibility of establishing a predi ation between the base noun and the pseudo-resultative adje tive. The se ond part develops the topi It of stativity-eventivity alternation. on erns stativity diagnosti s for Italian, the behaviour of parasyntheti deadje tival verbs and a pra ti al appli ation of synta ti automati extra tion of spe i stru tures from a The notion of verbal stativity is it as the simplest aspe tual ontroversial, some resear hers lass and that stativity ombine with other aspe tual inuen ing elements, su h as (Pylkännen 2000). Thus, stativity and its nature be ame a the synta ti onsider lass (Dowty 1979; Van Voorst 1992), others assume that statives are not a simple and uniform an diagnosti s for the orpus. ausativity entral issue of debate in the last de ade. In fa t, before starting a debate about the nature of stativity, it is worth isolating synta ti stru tures or semanti expression of stativity. must dene what onstraints that are involved in the Just as a biologist insterested in the study of roses riteria dene a rose, a synta ti ian interested in stativity must dene a set of rules whi h dene a stative verb. For this reason, hapter 6 reports stativity diagnosti s whi h have been proposed in the literature and shows that some of them are not reliable sin e they dis riminate for phenomena related to stativity. I propose other diagnosti s whi h are more trustworthy and apparently In parti ular, we will see that synta ti onstru tions ross-linguisti ally valid. apable to identify sta- tives (i.e. imperative and progressive) do not oer a good level of exa titude, while semanti tests (interpretation under modals, interpretation with tem- poral adverbials and temporal narrative ross-linguisti ally valid. Furthermore, experiments that onstraints) are more reliable and hapter 6 des ribes some behavioural an be employed as stative diagnosti s. Some of the stativity diagnosti s des ribed are employed in whi h analyses parasyntheti type of verbs hapter 7, verbs with adje tival base. We will see that this an generate a double aspe tual reading (stative or eventive) depending on the semanti s of the base. The semanti s of the base is shown to determine whether the verb an alternate between the two aspe ts. Namely, if it involves a semanti s of form, dimension or weight the resulting verbs do not alternate between a stative and an eventive reading; if it involves a semanti s of olor, brightness or beauty the resulting verbs an alternate between the two readings. This is shown to be related to the physi al hange 4 of the Theme: if a hange o urs, the stative reading is ex luded. We will see that the whole volves reason, a new approa h to a lass of parasyntheti ount for stative ausation is explored and partially updated to ausatives, namely the for e-dynami tion (Copley & Harley 2015; Copley & Martin 2015; that deadje tival verbs in- ausal readings, in both eventive and stative interpretations. For this ausation does not involve hange, even if approa h to inter al.). ausa- We will see hange automati ally involves ausation. In the ase of parasyntheti sation is introdu ed by a stati stative for e ausatives, it is assumed that au- alled abdu tion whi h is introdu ed in the system by the speaker and whi h generates a ausal link between the Causer and the Theme. aused by the Causer's The state of the Theme is existen e, without the intervention of energeti for e or of a judge parameter (Stephenson 2007) further in the introdu tion of hange. The presen e onrms the speaker's role ausal meaning. The judge parameter is a pragmati modier whi h relativizes the proposition su h as a for the speaker was added. The judge parameter does not have the same referential possibilities, when the verb is interpreted as eventive the judge parameter an relativize to the speaker's opinion only the result of the verb. When the verb is interpreted as stative, the judge parameter sult, the fa t that a for the an relativize to the speaker's opinion: the re- ausation has taken pla e, and the individual responsible ausation. I argue that the presen e of a predi ative result proje tion (rP) (A edoMatellan 2006) whi h involves the dire t obje tsSC in both aspe ts is responsible for the meaning. ausal meaning. The la k of rP would derive in a non- ausal The dieren e between stative ausative and eventive stru tures resides in the semanti s of the little ausatives involve a stati tive for e-dynami vbe v ausative fun tional proje tion. Sative vrelation, while eventive ausatives involve an even- ome . We will see that a reliable identi ation of stativity is important outside the theoreti al world. Chapter 8 reports my natural language pro essing proje t (held by Dr. ontribution to a wider Copley, CNRS, and Dr. Wol, Emory University) whose aim is the automati dete tion of temporal orientation of senten es. The target of the present study was to produ e synta ti tomati identi ation of stative verbs in a steps whi h orpus. Chapter 8 reports dierent ondu ted to the denition of synta ti an be used by a parser. The rules for the au- rules for stativity that hapter further des ribes how we were able to dene a gradient of stativity for English verbs. The gradient was obtained with the interpolation of the results of the parsing of a twitter the results of a semanti interpretation task orpus and ondu ted on English speakers. 5 Stativity plays a big role, as it is shown throughout this dissertation, in the temporal senten e orientation, sin e it imposes dierent temporal onstraints. For this reason, the identi ation of stative verbs is parti ularly important in a proje t aiming to automati ally dene temporal orientation. 6 Chapter 1 Methodologi al notes 1.1 Introdu tion This dissertation fo uses on the lexi al syntax of verbs built on a nominal and on an adje tival base. We will see during sthe investigation that these verbs play an important role in pseudo-resultatives ( hapter 4). an be interpreted as stative or eventive, and Moreover, they an be dened as aspe tually variable ( hapter 7). The stru tures investigated here do not onstitute basi and obvious pa- rameters of Italian, like word-order between determinants and nouns, preposition and nouns, verbal morphology and the verb. that part of language that Rather, they belong to annot be simply investigated with informal meth- ods, sin e their interpretations depend on external fa tors, su h as intonation, world knowledge, and lexi al material. We will analyze some of these disturbing external fa tors and we will see how they resear h results when not properly ontexts, an inuen e ontrolled. Even though the present work is not meant to be experimental, during its onstru tion a ree tion was made about standard methods of data olle - tion in the generative framework. Parti ularly, I dis uss usual methods of the generative enterprise grammati ality is not resear h on erns on erning synta ti and semanti phenomena whose learly evident to all native speakers. The present onstru tions whose (a)grammati ality is often very di- ult to determine be ause of: (i) low frequen y in the everyday language, it is the ase of pseudo-resultative mutual inuen e that linguisti onstru tion in Fren h and Italian; (ii) the elements have at the interfa e between syn- tax and semanti s, su h as stative/eventive alternation. We will see in this hapter that the semanti a eptability of senten es is subje t to bias, su h as frequent exposition to the stru ture or the respe t of the experimenter's 7 ������� �� �������������� ����� 8 expe tan y. The onstru tion of proto ols is sometimes di ult be ause experimen- tal questions are di ult to translate in everyday language making them omprehensible to naive parti ipants. Possible short omings of experimen- tal proto ols employed in the present work are presented within dedi ated se tions (se tion 4.4.5). Spe i experimental designs employed for the dis- sertation are presented in dedi ated se tions (6.3.1.1, 4.4.1, 4.5.1, 5.2.1, 7.4, 8.2.2) where their results are fundamental for the theoreti In this reasoning. hapter, I will fa e the problem of reliability of experimental pro- to ols in generative syntax, furthermore I will delineate some reasons why the appli ation of experimental proto ols employed by other ognitive dis- iplines, su h as psy holinguisti s, would be preferable for higher reliability, possibility of an interdis iplinary relationship and s ienti The adequa y. hapter reports possible issues implied in the use of non stru tured designs, su h as the la k of repeated measures and the impossibility of isolating variables. Furthermore, it analyses fa tors inuen ing linguisti olle tion, su h as the la k of ontext of interpretation, lexi al material and its frequen y, and target stru ture awareness by parti ipants. be ome more important in two so io-linguisti stru tures data ases: These issues (i) studies of languages with a ri h panorama, where dierent varieties hara terized by spe i oexist; (ii) studies of non-frequent stru tures (se tion 1.3.1). In support of more stru tured investigating methods, pointed out dierent studies have ases in whi h non-formal methods of data olle tion lead to the formulation of in orre t theories, and are reported in se tion 1.4. Se tion 1.5 reports a number of designs useful in syntax-semanti s resear h; some of them will be employed in the following for the sake of 1.2 hapters, while others are reported ompleteness. Data in generative linguisti s Investigation in generative syntax begins with the work by Chomsky (1957), who analyzed and theorized some linguisti fa ts of the English grammar. An important aspe t of his framework is the fo us on the grammati al om- peten e of a native speaker, rather than on the analysis of her linguisti ompeten e. Performan e A tual observed use of language, produ tion and omprehension. Governed also by prin iples of ognitive stru ture, that are not properly aspe ts of language. (Chomsky 2006: 105) : ���� ���� �� ���������� ����������� 9 Competen e Ability of the idealized hearer-speaker to asso iate sounds and meanings stri tly in a ordan e with the rules of his language. : (Chomsky 2006: 103) Performan e provides data for the investigation about linguisti ten e, sin e linguisti ompeten e is the true obje t of study of generative linguisti s. In a generative approa h, rules of a spe i ompe- ompeten e an be dened as the set of natural language internalized by a language user. Compe- ten e must not be onfused with the more general term of ability (Shohamy 1996: 138). to dis over the grammar of some language user, we must begin by obtaining information that bears on his interpretation of senten es, on the semanti , grammati al and phoneti stru ture he assigns to them . This means that the generative enterprise does A ording to Chomsky (2006: not have dire t a ess to 105),  ompeten e, sin e it onsists in the whole set of the theory of universal grammar deals with the me hanisms used in natural languages to determine the form of a senten e and its semanti ontent  (Ibid : 107). rules governing the spe i language performan e,  The main tools of resear h in the generative enterprise in the last 50 years omprise grammati ality judgments olle ted informally. Synta ti ians often investigate the opinion of a speaker about the a eptability of a given senten e by means of a dire t question, without implementing a strong strategy for the ontrol of other variables. The re ur to a the fa t that the speaker's opinion is eptability judgments is justied by onsidered as a reliable manifestation of her internal grammar, whi h is the linguist's ultimate obje t of study. [a℄ senten e whi h is judged as grammati al by a native speaker is part of that speaker's mental grammar, while a senten e whi h is judged as ungrammati al is in violation of a linguisti rule of the speaker's mental grammar . A ording to Ionin (2012),  Generative linguists onsider the internal grammar of one single language- user as a stable and su ient representation of the set of rules governing that spe i language. This derives in the non respe t of experimental proto ols typi al of other ognitive s ien es. A ording to S hütze & Sprouse (2013), [t℄he majority of judgment olle tion that has been arried out by linguists over the past 50 years has been quite informal by the standards of experimental ognitive s ien e . These informal methods are represented by unstru tured  grammati ality judgments whi h present some issues that we will analyze further. The question of reliability of data olle ted informally has always been matter of debate within the wider eld of general linguisti s. are In fa t, data olle ted in non statisti ally signi ant ways, sin e resear hers often refer ������� �� �������������� ����� 10 to their own intuitions about their mother tongues (introspe tion) in order to validate their theories, or ask some olleague or relative for judgments. On the one hand, this issue has represented an un overed A hilles' heel whi h all other approa hes to syntax ould use in order to dis redit genera- tivism and its results; on the other hand, it is a protable method to data about basi olle t fa ts of language (word order, agreement, ...). With the development of new experimental dis iplines on language fa ulty and languages, new te hniques for developed. Sin e the synta ti olle ting impli it data have been analyses have be ome more and more subtle, involving very deli ate judgments, whi h are easily other variables, new approa hes to data onditioned by ontext or olle tion are desirable for generative syntax. The adjustment to more stri t experimental proto ols is desirable for three main reasons. The rst onsists in the grounding the theoreti lation on solid bases, i.e. on reliable data whi h are not spe u- ontestable unless the repli ation of the experiment gives other results. The se ond onsists in the possibility of using information from other dis iplines investigating languages under other perspe tives. The third help in establishing onsists in the fa t that stri t proto ols orrelations between data, leading to the possibility of interpolate many dierent fa tors. It is important to point out that introspe tion remains the rst and most powerful tool a linguist has to dene the exa t obje t of study. In fa t, without introspe tion no linguist would ever been able to realize the presen e of wh- movement or verbal aspe ts, for example. Thus, introspe tion and small experiments ( hara terized by a small number of experimental subje ts) is a good method to start an analysis. More stru tured experiments are good to produ e more subtle analyses. Linguists using introspe tion or small experiments must be aware of possible problems that these te hniques pose and be areful in the design. In the following se tion, I will report some of these problems. 1.3 The issue The usual te hnique for the onsists in the informal olle tion of data among generative synta ti ians olle tion of grammati ality judgments, or a ability judgments as Cowart (1997) points out. abstra t on ept, no questionnaire to talk of a ept- Sin e grammati ality is an an guarantee a eptability judgments, re ordable and a ess to it, making better essible. In this hapter, I will use quite inter hangeably the two terms in this latter meaning. It is worth noting that a grammati ality judgment is a response of a ���� ��� ����� 11 speaker to a pre ise senten e, whi h (is presumed to) ontain(s) the linguis- ti phenomenon under observation. Generative linguist  has made an impli it promise that (i) there is a relevant population of speakers for whi h the reported judgments hold, (ii) the example senten es provided are representative of a lass of senten es as des ribed by the linguist, and (iii) with speakers randomly sampled from the relevant populations and senten es randomly sampled from the relevant lass, an experimenter would nd more or less the same judgments that the linguist reports  (Marantz 2005: 10). This would be true, if the number of observations (items and subje ts) were larger. In every experimental dis ipline, resear hers presume that parti ular subje ts, randomly hosen, are representative of the whole lass. What makes the strength of experimental dis iplines is the fa t that a large number of repli ations redu es the probability of assigning high weight to a peripheral behavior whi h belongs only to a parti ular subje t or to a parti ular observation. Non-stru tured grammati ality judgment, if not well dierent issues, namely it does not respe t basi method: it does not show enough parameters whi h are to the isolation of epiphenomena and an inuen e the to answer, the inuen e of the ondu ted, presents prin iples of the s ienti olle tion, the use of expli it knowledge ontext, the time of exposition to the same pool; it does not register physi al responses, it does not produ e data that an be analyzed by means of statisti al te hniques of validation. Usually, the unstru tured grammati ality judgments are dierent from the methods used by other ognitive dis iplines investigating natural lan- guages under dierent perspe tives, namely (S hütze et Sprouse in press ): a. Small number of informants b. Non naive informants . Small number of response options d. Small item pools e. Non systemati data analysis With respe t to the question about s ienti methods, it is worth noting that it is not a general problem of the generative enterprise, whi h on the ontrary is responsible for having ins ribed linguisti s within s ien es, but a problem of data olle tion. This fa t is parti ularly survival of the generative enterprise within the eld of whi h ae ts fruitful ex hanges with other dis iplines. hallenging for the ognitive s ien es, ������� �� �������������� ����� 12 Why do methods of generative syntax, whi h are omposed of small ex- periments and introspe tion, not (usually) respe t the s ienti A method? ording to Gibson & Fodorenko (2013), it has been pointed out that the results obtained using this method are not ne essarily generalizable beause of (a) the small number of experimental parti ipants (typi ally one); (b) the small number of experimental stimuli (typi ally one); ( ) ognitive biases on the part of the resear her and parti ipants; and (d) the ee t of the pre eding ontext .  In a non-stru tured a eptability questionnaire, it is di ult to he k for other parameters that enter in the judgment. It is well known that speakers are inuen ed by dierent fa tors when judging a senten e, su h as the text of interpretation, the frequen y of lexi al material, semanti on- plausibility, identi ation of the obje t of study, respe t of the resear her's expe tation, ... Speakers are used to ontext of interpretation reate a attest whether a senten e is a eptable or not. Non-stru tured a in order to eptability questionnaires do not (usually) dene a possible ontext of interpretation, leaving to ea h informant the task of dening it. This leads to the quen e that ea h evaluation is dierent The onse- ondu ted against an unknown and probably ontext of referen e. ontext and the linguisti register against whi h informants are sup- posed to interpret the experimental pool must be made expli it at the beginning of the experiment. Possible disagreement between informants due to a dierent an be ontext of interpretation. If we add the usual small number of informants of non-stru tured questionnaires, we qui kly understand that data obtained annot be interpreted as being signi ant of a population (of senten es/stru tures and of informants). Highly frequent lexi al material an fa ilitate the interpretation of du- bious grammati al senten es and, the other way around, infrequent lexi al material an lower the a eptability rate of senten es for reasons independent from the grammati ality of synta ti Furthermore, the ality of word stru tures employed. hoi e of lexi al material is at stake in the prototypi- ombinations. Ea h word a tivates a net of ne tions with other words and on eptual on- on epts, the more the link is tight the more it is easy to get a possible interpretation for a senten e, deriving its possible re overy in ase of (mild) agrammati ality. Unstru tured a the interests. al. eptability questionnaires are usually small entourage This of the resear her, ondu ted within whi h is probably aware of her an entertain problems of observer expe tan y (Gibson 2013: 100), involuntarily falsifying results. Furthermore, informants et an ���� ��� ����� 13 easily understand the obje t of study and answer onsequently, using their notions about normative grammar, thus using their expli it knowledge:  if learners re ognize whi h stru ture is being tested in the AJT, they may draw upon ons ious, expli it knowledge, as learned in the lassroom, and the results may not inform us about the learners' underlying grammati al intuitions  (Ionin 2015). Aware parti ipants an be vi tims to onrmation (Gibson et al. 2013: 99) and belief bias (Evans, Barston, Pollard 1983). Whenever informant and resear her are hosted within the same person, re urring to auto-investigation method, the resear her must be very are- ful to more frequent bias. Among them: (i) expe tation for nding data in the onrmation of a hypothesis an lead to onsider grammati al what is not grammati al; (ii) repeated exposure to stimuli an inuen e their a ept- ability (Levelt 1972). It is worth noting that auto-investigation te hnique is useful and ne essary for the rst part of the job, the one in whi h an interesting phenomenon is isolated. Unstru tured a eptability questionnaires are not usually submitted to repeated measures. This is a problem under dierent points of view. Firstly, no possible statisti al analysis an be made on data, thus no possible validation is produ ed. Se ondly, individual os illations and assume an unrealisti Unstru tured a annot be relativized weight. eptability judgments are ondu ted with little means, usually by a dire t oral question and an unregistered answer. They do not re ur to the re ord of any physi al involuntary response of the informants, only to her overt and manifested opinion about a linguisti However, primary intuitions whenever fa t. olle ted in a formal way an on- stitute a sour e of data, but they are not the only one, as argued by Wason & [p℄rimary intuitions are a legitimate form of eviden e for linguisti hypotheses, but they should have no privileged status relative to other forms of eviden e . Arnold (2005: 1485):  The fa t that involuntary responses are not registered is not a big problem if questionnaires are planned and of the informant ondu ted in ways su h that the opinion annot be inuen ed. The question about the status of data in formal linguisti s, parti ularly in formal syntax, arises in these last years with an in reasing number of studies, namely Edelman & Christiansen (2003), Ferreira (2005), Wasow & Arnold (2005), Featherston (2007), Gibson & Fedorenko (2010a, 2010b), among others. These studies argue for the need to follow stri t experimental proto ols for data olle tion in formal linguisti s. They take as eviden e reported by eminent synta ti ians whi h were wrong a ases of data ording to results of ������� �� �������������� ����� 14 more stru tured experiments. Sprouse & Almeida (2012) intervene in the debate assuming the validity of traditional data olle tion methods. tional methods obtain Parti ularly, they show that tradi- omparable results as more stri t methods, showing a dis repan y of 2%. Their (2013) study takes as referen e Adger (2004) introdu tory manual of syntax, whi h reports examples of well studied phenomena and basi English stru tures. Whoever asked if example 107 is grammati al in English will answer quite easily that it is not. This is due to the fa t that it involves word order. As already mentioned, basi synta ti fa ts are easily olle ted, independently of the methodology employed. (107) The *John eats apple the. orpus tested by Sprouse & Almeida (2010) is onstituted of senten es like (107). The position taken by Sprouse & Almeida has been hallenged by Gibson & Fedorenko (2013), who reply by showing the importan e of using quantitative methods in linguisti resear h. Gibson & Fedorenko (2013) hen e GF (2013) illustrate that the mathemati al tools and reasoning implied by Sprouse & Almeida (2013) are not adapted and onsequently derive an optimisti result. The ratio of 5% whi h is used is not adequate, in fa t it onservative when applied to single of ontrasts. In this latter an be su iently ontrasts, but not when applied to a pool ase, it is impossible to establish whi h belong to the 5%, so it remains unknown whi h are Imagine to analyze an arti le in whi h 60 ontrasts orre t and whi h are not. ontrasts are reported (without any quantitative measurement), you know that the global reliability rate arise to 95%, this implies that there are 3 wrong ontrasts. You don't know whi h they are. The identi ation of the ombination of wrong and orre t orre t ontrasts must be found among a lot of ombinations. In other words, you have a box ombinations, exa tly 34220 ontaining 6 balls, 4 of whi h are white and 2 bla k. You have to extra t the two bla k balls rst, without looking inside the box. How many times should you repeat the extra tion in order to get both bla k balls extra ted? This is a ase of simple ombination. The number of extra tions you have to make in order to get the right (C{n,k} ) is determined as follows, where n is the total number of balls, at h k is the number of white ball we want to extra t. (108) Cn,k = In the Dn,k Pk = n! k!(n−k)! ase of the extra tion of balls, we have 1/15 probability to get the two bla k balls extra ted rst. In the probability of at hing the wrong ase of 60 linguisti ontrast is one over 34220. ontrasts, the ���� ��� ����� 15 This leads to the impossibility of onstru ting a reliable linguisti theory, sin e it must generalize over all examples reported, wrong ones in luded. non quantitative methods have no hope of re ognizing these errors. [...℄ Experimental methods were required to do this evaluation be ause experimental methods are the only way to obje tively determine whi h hypothesized ontrasts are real  (Gibson, Pianadosi and Fedorenko 2013: 233). Then  The ondu tion of quantitative experiments in linguisti s makes theories omparable be ause lear a between reality and data ( eptability rates allow to dis over dis repan ies ibidem : 238). Moreover, small experiments for for linguisti variation. olle ting linguisti data annot a We already said that not all senten es straightforward judgments as (107). Cases in whi h linguisti stake are mu h more problemati an re eive variation is at in an informal questionnaire. to Wasow & Arnold (2005), the level of a ount A ording eptability of senten es (109) to (111) is not uniform among all Ameri an English speakers. (109) Chris might an go. Wasow & Arnold (2005, ex 1a) (110) Pat's a Red Sox fan, and so aren't we.Wasow & Arnold (2005, ex 1b) (111) He don't like that. Wasow & Arnold (2005, ex 1 ) Quantied experiments are non-quantied, to a apable, ount for linguisti ontrary to the non-stru tured and variability. Using small unstru tured questionnaires in the ase of more subtle linguisti readings. an generate some issues even phenomena, su h as ambiguous eventive I identify two main reasons in favor of more stri t experimental proto ols in data olle tion in the syntax-semanti s interfa e. The rst has to do with the use of spe i the importan e of the theory-internal terminology, the se ond on erns ontext. For example, a non-stru tured questionnaire about aspe tual readings must previously dene to the informant ea h aspe tual ally lass. This automati- ommuni ates to the informant the position of the resear her about the topi , The onsequently leading to expe tan y biases. ontext appears to be fundamental in the interpretation, but in a small test, it is not stri tly out being ontrolled. Then, it an inuen e results with- onsidered among variables of the experiment. Imagine that in a small experiment, the resear her is introdu ing senten e (112) with the two ontexts below respe tively, whi h dier in just one word. (112) Il grumo ostruis e l'arteria. The lot is loting the artery. ������� �� �������������� ����� 16 (113) a. Dopo attente analisi, dopo aver ri ontrollato l'esito di queste ontro dierenti pareri, i medi i hanno dato il triste referto alla famiglia del paziente: da ieri il grumo ostruis e l'arteria. After attentive analyses, after having double he ked medi al reports, do tors gave the sad new to the patient's family: sin e yesterday the lot was obstru ting the artery. b. Dopo attente analisi, dopo aver ri ontrollato l'esito di queste ontro dierenti pareri, i medi i hanno dato il triste referto alla famiglia del paziente: da sempre il grumo ostruis e l'arteria. After attentive analyses, after having double he ked medi al reports, do tors gave the sad new to the patient's family: sin e always the lot obstru ts the artery. It appears lear that two ontexts an inuen e judgments. Consequently, ontext must be en ountered within the ontrolled variables of the experi- ment. I suppose that there are more reliable methods to syntax-semanti s interfa e. For example, olle t data at the on-line experiments are better suited for disentangle two readings, by means of un ons ious answers (reading times, pla e of o ular xation, ...), and not only of expli it judgments. 1.3.1 In the present study The present study analyses, for the most part, some fa ts of Italian. The soiolinguisti that this dimension must be 1.3.1.1 � panorama of this language is parti ularly olorful, and I assume onsidered in the pro ess of data olle tion. Italian linguisti panorama Italian is the national language of the Italian Republi and it is taught and spoken all over its territory, in the Vati an City, in the Republi Marino, in two Switzerlan antons (Marazzini 1994: ountries in whi h Italian plays an o ial role, other 458). of San Besides these ommunities in the world use Italian as their rst language, namely Italian rst generation emigrants. On the Italian territory there are many diale ts that must be re ognized as languages, autonomously derived (for the most part) from Latin. � I do not want to ommit myself whether Italian is parti ular as opposed to other languages in this respe t, or whether this state of aairs results form the availability of a big amount of data that has emerged from diale tologi al and so iolingusti resear h, not ondu ted in other linguisti areas of the world. ���� ��� ����� A 17 lassi ation of linguisti areas, based on linguisti features, has been ondu ted by linguists, even though the panorama is still evolving, sin e  a stati nature of diale t territories does not exist and has never existed  (Rohlfs 1972: 11, my translation). Two isoglosses, imaginary lines joining points of deep breaking in the diale tal ontinuum, are identied: La Spezia- Rimini and Roma-An ora. The derived three geographi parts onsist in the three diale tal super-varieties of Italian: the Northern, the Central and the Southern (Marazzini 1994: 466). Italian is the national language, whi h is taught in s hools and whi h is vehi le for mass-media and art. � However, even though the uni ation of Italy dates from 1861 , Italians still have a good relationship with their own diale ts. A � ording to 2006 ISTAT statisti s, 48% of Italians de lare to speak alternatively standard Italian and a diale t; only 45% of Italians de lare to speak ex lusively standard Italian (D'Agostino 2007: 55). The label Italian does not des ribe a uniform language throughout the national territory. A Italian languages. ording to De Mauro (1972): there are dierent regional These regional Italian languages are dierent uses of the national language whi h are made in ea h region. They derive from histori al melting of diale ts and national language. Regional varieties must be onsidered in experiments. stru tured questionnaire is used, possible fa t and a parti ular regional Italian orrelations between a linguisti ould not be established. asking for judgments in an informal way dard Italian Firstly, if a non- Se ondly, an be interpreted as a test of stan- ompeten e by informants, deriving in ensure about synta ti stru tures whi h are otherwise well judged and employed. Italian is only one of the languages presenting a rieties. Languages in the world present spe i ompli ated pi ture of va- so iolinguisti frames, derived for substratum languages, so ial fa tors su h as edu ation rate or prestige versus popular divide. A ording to Cowart (1996: 39)  we use experiments to estimate the properties of a population on the basis of tests applied to a sample drawn from that population . In order to do that, a s ientist must exa tly dene the population previously to the experiment. I suggest an attentive treatment of languages with a guisti panorama, they require tion, in relation with geographi areful data omplex so iolin- olle tion, sin e linguisti or so ial fa tors, varia- an play a big role in the (a)grammati ality judgments. � Not all territories were onquered at this period. Veneto region was annexed in 1861; Trento, Trieste and their regions at the end of the First World War. � National institute for statisti s. ������� �� �������������� ����� 18 Another fa tor that must be ods of data onsidered in the hoi e of stru tured meth- olle tion is the frequen y of stru tures studied in the everyday language. The investigation of non-frequent stru tures, lasses, prevent the resear her to built on parti ular verb onrm intuitions against a orpus analy- sis, sin e the low frequen y does not depend on the agrammati ality of those stru tures but on the intrinsi low rate of produ tivity. In the rst part of the present dissertation, I will analyze a non-frequent stru ture of Italian, namely pseudo-resultative It is onstru tion (Levinson 2007). lear that stru tured and quantitative methods are very important in order to guarantee reliability of the whole theoreti 1.4 apparatus. The importan e of being reprodu ible In this se tion, I will report ases in whi h informal data olle tion gave wrong data or in whi h the data were unable to determine signi ant inuen ing fa tors. [L℄anguage should be analysed by the methodology of the natural s ien es, and there is no room for onstraints on linguisti beyond those typi al of all s ienti inquiry work. (N. Smith, Foreword to Chomsky 2000: vii) With the advent of the minimalist program and its laiming for a unique pla e of grammati al representations (the generative engine), the methodologi al tradition should evolve in this dire tion. A ording to Wasow & Arnold (2005), linguisti s should follow the usual methodologi al expe tations of other or psy holinguisti s. Parti ularly ( • ognitive dis iplines, su h as psy hology ibid : 1483-84): The number of subje ts should be large enough to allow testing the results for statisti al signi an e. • The order of presentation of stimuli (that is, linguisti examples) should be randomized. • Subje ts should be ignorant of the hypotheses being tested, preferably with double-blind presentation of stimuli (naive speakers). • Data olle ted should be subje ted to appropriate statisti al analysis. ���� ��� ���������� �� ����� ������������ 19 The respe t of more rigid experimental proto ols fa ilitate the re ognition of generative linguisti s as a ognitive s ien e and allows bidire tional ex hanges with other dis iplines. It has been noti ed that non-stru tured data olle tion an lead to the spread of wrong data, on whi h a part of theory has been built. For example, Wasow & Arnold (2005) hen eforth WA (2005) study the positions of NPs in dierent problemati synta ti double obje t pattern or heavy NP shifts. environments, su h as Fillmore (1965: 2930) assumes that senten es su h as the one in (114) and (115) are agrammati al be ause of the interrogation of the rst obje t of a double obje t (114) Who did I buy a hat? (115) Who did you give this book? onstru tion. Langendoen et al. (1973) performed a test on 160 English native speakers, asking them to insert the dative preposition without to in double obje t questions hanging their meaning. If Fillmore were right, the expe tation is that only one answer is grammati al, onsequently only one insertion pla e is allowed, namely the one in whi h the dative preposition marks the dative obje t, and follows the verb. (116) Who did you oer (117) Who did you show to to the man? the woman? However, Langendoen et al. (1973) dis overed that many speakers pla e the dative preposition at the end of the senten e, as it were an o of preposition stranding onsequent to the questioning by the wh urren e pronoun of the dative obje t. to? (118) Who did you oer the man (119) Who did you show the woman These results to ? ontradi t Fillmore's hypothesis, sin e the questioning of the internal obje t of double obje t onstru tions should be agrammati al, and then unre overable for speakers. These rst ndings were further supported by another stru tured test about possible answers to double obje t questions of the usual type. If they were agrammati al, the answer should involve the reading in whi h the internal dative obje t is in pla e. But again, many informants answer in the presumed impossible way. (120) Who did you show the woman? ������� �� �������������� ����� 20 a. I showed the woman my daughter. b. I showed my daughter the woman. WA (2005: 1490) further analyze another popular statement about in- uen ing fa tors of the position of English parti le-verbs has been said that the internal omplexity of nominal onstru tions. It onstituent (number of intermediate nodes) is determinant for the separation of the parti le from the verb (Chomsky 1975). WA (2005) ondu ted some tests questionnaires and order to verify the vera ity of Chomsky's statement. They orpus sear h in onstru ted min- imal pairs of senten es whi h were equal in number of words, but diering in synta ti (121) Two omplexity. a. The hildren took everything we said in. (WA 2005: 1490, ex.8) b. The hildren took in everything we said. . The hildren took all our instru tions in. d. The hildren took in all our instru tions. onditions are interpolated: omplexity of the internal obje t and position of the parti le with respe t to the verb. More Complex Less Complex V ... part a d V-part b In order to test whether omplexity of NPs plays a role in syntax, WA build other quadruples on dierent stru tures involving dierent orders: dative alternation (122), and heavy NP shift (123. (122) (WA 2005: 1490, ex. 9) The ompany sends what Ameri ans don't buy to subsidiaries in other b. a. The ountries. ompany sends subsidiaries in other ountries what Ameri- ans don't buy. . The ompany sends any domesti ally unpopular produ ts to subsidiaries in other d. The ountries. ompany sends subsidiaries in other ountries any domesti- ally unpopular produ ts. (123) (WA 2005: 1491, ex. 10) Nobody reported where the a ident took pla e to the poli e. ���� ��� ���������� �� ����� ������������ b. a. . d. Nobody reported to the poli e where the a Nobody reported the lo ation of the a 21 ident took pla e. ident to the poli e. Nobody reported to the poli e the lo ation of the a WA (2005) dis overed that, in the ase of Verb-Parti le Chomsky's intuition about the inuen e of NP's ident. onstru tion, omplexity for the deter- [a℄n analysis of varian e revealed that the intera tion between omplexity and ordering was signi ant (P i 0.001) by subje ts, but not by items (P h 0.1)  (ibid : 1491). mination of parti le's position was The above ndings onstituent onrmed:  onrm Chomsky's intuition about the inuen e that omplexity has on preposition position in senten es. WA do not ex lude that length does not play a role, For this reason, they study two English about the order in Verb-Parti le ant fa tor but obje t However, ontrary to Chomsky. � orpora (written and oral) . Results onstru tions show that length is a signi- omplexity is not. On the other hand, results about double onstru tion are more lear sin e the relative length between the two obje ts is determinant. Either length or omplexity play a role in the order of obje ts, as gure 1.4 shows (WA 2005: 1493, Table 1). Figure 1.1: Relation between length and omplexity in double obje t on- stru tion ordering, (Wasow & Arnold 2005: 1493, Table 1). � Of the sele ted 1393 o urren es of dative alternation and 3268 o urren es of verbparti le onstru tion, they make a three-points s ale depending on the omplexity of the NP involved and a s ale of word lengths of NPs. ������� �� �������������� ����� 22 These results show that experiments with a more stri t proto ol must be ondu ted in order to determine the grammati ality of a onstru tion and the reasons of (a)grammati ality. Gibson & Fedorenko (2013: 102 .) point out that some well-known judgments reported in the literature happen to be in orre t, for this reason they analyze in depth three phenomena. The rst on erns subje t and obje t modifying relative usual assumption is that double nested relative lauses. The lauses are more di ult to be understood when they modify a subje t (124) than double nested relative lauses modifying the obje t (125). (124) The man that the woman that the dog bit likes eats sh. (Gibson 1991, ex. 342b) (125) I saw the man that the woman that the dog bit likes. (Gibson 1991, ex. 351b). From the untested assumption that (124) is more a theory of nested relatives has been formulated, a ompli ated than (125), ording to whi h the stru - ture of (124) has a higher number of open dependen ies, whi h determines this di ulty. In a third phase, an on-line test re ording reading times (Gibson, Desmet et al. 2005) shows that senten es like (124) are read faster than senten es like (125), ontradi ting the intuition formulated in previous analyses. The se ond and the third ases analyzed in Gibson & Fedorenko (2013) involve multiple wh-extra tion ee ts. The se ond ase regards the asymmetry in the extra tion of two wh-words in wh-questions (Chomsky 1977): (126) a. Who ate what? b. *What did who ate? The higher grammati ality of (126a) is supported by quantitative experiments (Clifton, Fanselow and Frazier 2006; Fedorenko subsequent laim by Kayne (1983) stating that the a improves when a third wh- pronoun is inserted (127) is (127) et al. 2006), but the eptability of (126b) ontradi ted. *What did who ate where? The third ase analyzed on erns data whi h led Chomsky (1986) to for- mulate the Va uous Movement Hypothesis in order to explain the (presumed) dieren e in a eptability between senten es like the following (Chomsky 1986, ex. 108; reported by Gibson & Fedorenko 2013: 108). ���� ��������� ������� (128) 23 a. What do you wonder who saw? b. *I wonder what you saw. A battery of tests was ondu ted by Gibson & Fedorenko in order to he k for the presumed higher a eptability of (128a) than (128b). Results ontradi t this assumption and reveal the opposite, of a eptability of both examples and the lower a We must underline that data onrming the low rate eptability of (128a). olle ted from naive speakers sentative of the high variability linked to general linguisti an be repre- variation (diatopi , stable natural phenomenon of senten e a eptability; we nd that for all the synta ti phenomena onsidered [that-tra e, NP-extra tion, ante edent in oordination℄ native speakers of Ameri an English exhibit stable, lear- ut patterns of a eptability dieren es a ross senten e types . diaphasi , ...). However, as Cowart (2006: 26) points out  In order to get rid of this variation, the number of informants and items [t℄he pro edures des ribed in this pool assume the existen e of error varian e in senten e judgments and apply various measures to ontrol that varian e. The most important of these measures are the use of multiple informants and multiple instan es of any senten e type whose a eptability is to be estimated  (Cowart 2006: 37). involved is fundamental,  1.5 Dierent designs The respe t of stri t experimental proto ols prote ts us from many dierent types of bias, as we showed in previous se tions. proto ol leads to dierent The type of experimental onsequen es, it inuen es the results, sin e ea h � design leads to dierent data . A ording to Keller (1998), we an identify four main fa tors inuen ing grammati al judgment tests: evaluation s ale, instru tions, various subje tdependent fa tors and various task-related fa tors. Grammati al s ale type determines the type of statisti treatment that an be applied, and the nature of judgments. Evaluating s ales an be nom- inal, ordinal and interval. Values reported in the rst two types do not suit on a regular s ale, this means that the distan e between two points not be the same. Values in a nominal s ale annot even be ordered, ould ontrary to ordinal s ale. In an interval s ale, points are ordered and the distan e between them is onstant. Interval s ales guarantees that distan e between two � A small experiment leads to a with a stri t proto ol. ertain type of data, the same way bigger experiment ������� �� �������������� ����� 24 points is onstant, this permits to treat data in a ontinuous way, allowing statisti al analyses for normal distributions. The inuen e of instru tions on results is a matter of debate. (1998: 6) assumes that, along with the naivety of informants, as grammati al or ungrammati al Keller on epts su h if not dened in the instru tions are not signi ant, leading to the in omprehension of the task. On the other hand, an experiment by Cowart (1997: � 55-61) instru tions are, informants have little shows that no matter what the apa ity to hange their range of judgments. Subje t-related fa tor label meets various phenomena about the inuen e that informants an have in the experiment results. Among them, one is parti ularly interesting and intuition we onsists in the naivety of the subje t. Beyond the an have about it (Cowart 1997: 60; S hütze 1996: 187), it has been demonstrated that there are signi ant dieren es between judgments given by linguists and the one given by naive subje ts (Dabrowska 2010). A series of experiments by Dabrowska (2010) shows that judgments about Long Distan e Dependen ies (LDD) dier signi antly between naive informants and professional linguists working in various theoreti al frameworks. magnitude of judgments made by naive speakers is less a The entuated then the one made by linguists, both in positive and in negative (Dabrowska 2010: 13). Linguists show a prototypi ality ee t in the judgment of LDD ( ibid : 20). Unfortunately, Dabrowska's (2010: 11) pro edure ontains a false belief in the instru tion phase, sin e instru tions dier between the experiment administered to naive informants and the experiment administered to linguists. The latter ontains expli it referen e and asks the informants not to rely on what they had learned in the ourse of their linguisti training. This kind of request unique and annot be satised sin e the judgment we have on a senten e is annot be split in two: the one we would have if we weren't linguists and the one we have sin e we are linguists. Another task-related fa tor apable of inuen ing judgments onsists in the presentation order of experimental items. Order of presentation and repetition of experimental items an ae t results in dierent ways: de reasing the grammati ality rate (if repetition happens in a short amount of time) as demonstrated by Nagata (1987 and .), blurring or in reasing grammati ality rate in the ase of linguists' disease � (S hütze 1996) . � Two experiments equal in the experimental items and dierent for the type of instru - tions: one intuitive and one pres riptive. No signi ant dieren es are found in the given judgments. � The extended exposure to the same stru ture makes its grammati ality more un ertain. ���� ��������� ������� 25 In the following sub-se tions, I will report some designs whi h are useful for synta ti studies. Dierent designs are theorized in the literature and their appropriateness depend on the type of phenomenon investigated, or on the type of statisti al analysis to be performed on results. In the following se tion we will see some of them, without the presumption of being thorough. 1.5.1 A eptability Judgment Test (AJT) The A eptability Judgment Test (AJT) is a stru tured version of the usual unstru tured grammati ality judgment test. AJT is an expli it test that asks parti ipants to evaluate plausibility of senten es following personal intuitions. In order to prevent informants from identifying the stru ture under investigation, it is important to: (i) mix experimental items and ller items, whi h are responsible for hiding the target stru ture; (ii) time the task timed (Ionin 2012) in order to for e informants to answer without a Pre ision and ess to their expli it knowledge. larity of instru tions are mandatory in order to guarantee that informants perform as expe ted. If instru tions are too te hni al or too omplex or too short, people may not properly understand what they are asked for. Dierent types of rating s ales exist: binary, Likert (on 5, 7 or 10 points), ontinuous (see se tion below for Magnitude Estimation). Ea h s ale is appropriate in relation to the type of linguisti binary s ale is appropriate for fa t that must be investigated: lear and strong ontrasts, Likert s ale for more nuan ed one (Ionin 2000). Imagine that we want to test the a eptability of the pre-nominal position of Italian appositive adje tives. We test two onditions: Adj + N and N + Adj. (129) a. Giovanni possiede un G. owns a red book. b. Giovanni possiede un G. owns a book red. rosso libro. libro rosso. The same informant should not judge both sin e it would onditions (129a) and (129b), ompromise the results due to repeated expositions. By means of a Latin square, we obtain two experimental pools, ea h of whi h all experimental items dierentiated for their ondition 1 of senten es 1 and 2 and 2 we do the opposite. ontains onditions: in pool 1 we insert ondition 2 of senten es 3 to 4, in pool ������� �� �������������� ����� 26 Item Condition 1 Condition 2 1 rosso libro libro rosso 2 giallo telefono telefono giallo 3 bian a statua statua bian a 4 ... Ea h group of experimental items must then be randomized. In this dissertation, I resort to a randomizer � http://www.sfl. nrs.fr/EVO/s ripts/randomisation- designed by Dr. Coralie Vin ent . 1.5.2 Magnitude Estimation Task (ME) Bard, Robertson & Sora e (1996) theorize a Magnitude Estimation task (hen eforth ME) design be ause they to help the needs of linguisti onsider usual s ales: (i) too theories ( ibid : ondensed 38) and (ii) not involving a on- stant distan e between two points. ME was rst applied to physi al phenomena in per eption studies (Stevens 1956) sin e it  ( ibid : provides better than ordinal s ales for measuring impressions  40). Bard et al (1996: 41) onsider it a valuable te hnique for linguisti s too be ause: (i) it does not restri t the number of values; (ii) an interval s ale is subsumed by judgments on a ratio-s ale. For what on erns the mathe- the straight line in log-log oordinates means that equal ratios on the physi al dimension give rise to equal ratios of judgments  (Bard et al 1996: 41). mati al onsequen es of this design, it is worth noting that  This design is employed in hapter 4, where it is des ribed in detail in se tion 4.5.1. 1.5.3 Truth Value Judgment Task (TVJT) The truth value judgment task (hen eforth TVJT) has been initially designed for linguisti resear h on L1 a quisition (Gordon & Chafetz 1986), but it is useful also to investigate TVJT hild language. an be designed in dierent ways, depending on the age of in- formants (whether hildren or adults) and on informants spe i ognitive issues. Sin e this dissertation does not investigate hild language, I do not an- alyze the parti ular design for this age, the reader al. (2009) where TVJT is used for dete t an refer to Conroy et hild sensibility to Prin iple B violations. In this se tion, I want to underline the usefulness of this design in � CNRS, Laboratoire Stru tures Formelles du Langage. ���� ��������� ������� ases where 27 ontext is determinant for the interpretation of senten es with dierent grammati al readings, parti ularly where the plausibility of readings dier in dierent ontexts. Sin e naive speakers are not frequently asked in everyday life to identify all readings of ambiguous senten es, they are not used to at h all readings of an ambiguous senten e, one reading is often more preferred than others. Consequently, the ontext of referen e helps informants to gure out whi h of the possible readings is orre t/plausible. Originally, the TVJT re urs to a short story, introdu ing the informant to the situation, after whi h she is asked to judge the experimental item grammati ality. This allows the resear her to sesses the grammati al stru ture of the topi he k if the informant pos- of the study (passive, et .). In adult language, TVJT is not employed to register the grammati ality rate of stru tures, it is very useful for the identi ation of readings that a same grammati al senten e of the investigated an generate in dierent ontexts. The grammati ality onstru tion must be previously onrmed by means of other designs. Even though this dissertation does not re ur to the TVJT, the design was onsidered in all ases where double readings were supposed to exist. 1.5.4 Auto Segmented Reading Generative linguisti s is able to ex hange with psy holinguisti s, from the single engine hypothesis (Chomsky 2000). This hypothesis makes possible to a ount for the derivational theory of omplexity, whi h was formulation of a the main point of dis ord between psy holinguisti s and generative linguisti s in the 70s' (Fodor, Bever & Garrett 1974). A ording to Marantz (2005: 439), this hypothesis [...℄ the more an be formulated as: omplex a representation the longer and more omplex the linguisti omputations ne essary to generate the representation the longer it should take for a subje t to perform any task involving the representation and the more a tivity should be observed in the subje t's brain in areas asso iated with reating or a essing the representation and with performing the task. The Minimalist Program assumes that there is only one module to reate stru tures, and that is syntax. to build representations makes it possible to a representations. apable The existen e of a single pla e ount for the omplexity of ������� �� �������������� ����� 28 Behavioral tests be ome to be more meaningful for generative linguists, sin e they onstitute eviden e for the stru ture and the nature of prin iples regulating the single linguisti engine. an sh in the psy holinguisti Consequently, generative linguisti s literature looking for behavioral eviden e. In order to do that, the awareness of the importan e of stru tured data tion te hniques, of s ienti methodologies for data olle - olle tion and statisti al validation has to grow in the generative tradition. Auto segmented reading is an on-line method that provides for both expli it judgment on (a)grammati ality and for behavioral data. It is a te hnique apable of re ording reading times, useful whenever a double reading is supposed to be generated by the same stru ture. Dieren e in reading times an be interpreted as a ree t of dierent stru tures. Stimuli are made up of senten es split in several segments. The point of ut is established depending on whi h part of the senten e the investigated stru ture relies. That is, if the time of pro essing of animate vs. inanimate subje ts is at stake, the rst ut will be pla ed after the subje t. reading times of stative vs. eventive verbs is investigated, the rst If the ut will be pla es after the subje t and the se ond after the verb, this way the verb alone onstitutes one single segment. There are several pre autions to be employed in the design phase. Sin e this method ompares reading times, all the experimental items must be omposed of the same number of letters, otherwise no possible an be made. read, a omparison In order to push informants to pay attention at what they omprehension question must be added at the end of ea h segmented senten e. This te hnique has fruitfully been employed by Poeppel & Gennari (2006), they show that ausal semanti s entertains orrelates in reading times. parti ular they show that non- ausative statives are read faster than eventives. Results of this experiment are analyzed in depth in In ausative hapter 7, where they are useful to the development of the reasoning. 1.6 This Con lusions hapter reports dierent issues pertaining to the use of small experi- ments and auto-analysis, whi h are te hniques used in the generative tradition. It means to be a reminder for linguists pointing out the biases aused by using small experiments and auto-analysis. Assuming the importan e of these two te hniques for the olle tion of data in a rst phase of resear h, it delineates the reason why more stru tured designs should be employed for ���� ����������� 29 deeper phases of the resear h. whi h Furthermore it reports some useful designs an be employed in the synta ti resear h. We have seen that data, derived by auto-analysis, where the resear her and the informant are the same person, onsequently to in orre t theoreti an lead to in orre t results and generalizations. Employing more stru tured experiments should be parti ularly important for those languages whi h have omplex so io-linguisti panorama, su h as Italian, where the inuen e of diale ts on the national language is still very important. Unstru tured questionnaires are not apable of identifying whi h variety of language is being tested. The use of stru tured and quantitative methods reliability. Statisti al veri ation an guarantee higher data an be performed, so iolinguisti features of informants are registered, allowing to identify possible inuen es they have on judgments. The expli itness of experimental proto ol allows the s ienti ommunity to verify the orre tness of hypotheses and to understand whether the design, the item pools, or the statisti treatment have inuen ed results. However, I assume the importan e of auto-analysis and very small experiments in the rst steps of a resear h. In fa t, they allow the linguist to identify and delineate interesting fa ts of language. In other words, without auto-analysis linguisti s wouldn't exist. 30 ������� �� �������������� ����� Chapter 2 Argument Stru ture: State of art 2.1 Introdu tion The present hapter reports some of the most ommon theories on argument stru ture. In parti ular, it analyses frameworks whi h ing and morphologi ally derived verbs. on ern ausal mean- Parti ular attention is given to the treatment of stativity. For ea h framework analyzed, parti ular attention is paid to the way in whi h it a are: ounts for dierent verbal lexi al aspe ts. Frameworks analyzed Government and Binding (Chomsky 1981), Hale and Keyser (1993), Ram hand (2008), Borer (2005). Se tion 2.7 reports the theoreti al solutions that will be employed in the present dissertation. The relationship between argument stru ture, number of arguments and their semanti topi roles, as well as the eventuality of the predi ate is a of formal linguisti s. argument realization patterns are related to synta ti semanti roles. entral Many studies along the time have noti ed that realization of spe i Argument stru ture involves the wider issue of predi ate eventuality. A deep investigation on argument stru ture annot ex lude in- vestigation on eventualities of predi ates. (130) John runs 10 miles. (131) John bites Peter. (132) John loves Mary. Agent-Goal Agent-Experien er Holder-Goal In the last half- entury, dierent theoreti al hypotheses have been formulated in order to larify these issues. Two main urrents are dete ted, depending on the weight they give respe tively to lexi on and syntax. 31 ������� �� �������� ���������� ����� �� ��� 32 lexi alist approa hes One the one hand, laim that a lexi al verb omes equipped with synta ti and semanti type of its arguments. On the basis of this lexi al information, it builds up the synta ti spe i ations about the number and stru ture. On the other hand, for stru turalist approa hes, argument roles do not depend on lexi al spe i ation, rather ex lusively on the synta ti 2.2 stru ture in whi h verbs are inserted. Government and binding Sin e Chomsky's (1993) guisti Le tures on Government and Binding (GB), the lin- ompeten e has been divided in four dierent sub- omponents: lex- i on, syntax (a. phoneti ategorial form (PF) omponent; b. transformational omponent, lexi al form (LF) omponent), omponent. If the relationship between outputs of syntax and PF, and output of syntax and LF have always been maintained (GB, Minimalism, ...), the relationship between lexi on and syntax has be ome more ontroversial. In GB, D-stru ture (deep stru ture) is generated by a set of base rules, whi h are omposed by two systems, lexi al and synta ti ategorial om- through insertion of lexi al material into stru tures generated by [synta ti ategorial omponent℄, in a ordan e with their feature stru ture  ponents,  (H. Heider & Nettel 1991: 6). Ea h lexi al item is spe ied in the lexi on for its abstra t morpho-phonologi al stru ture and for its synta ti ( ategorial and features ontextual). Base rules generate D-stru ture through insertion of lexi al items into stru tures that are generated by the omponent, in a α, with their features. tra es ategorial Those are mapped to S-stru ture by move- ordan e leaving o-indexed with their ante edents. The fundamental on eption of the Proje tion Prin iple is that lexi al information leads to synta ti lexi al information stru ture, syntax is built on the basis of the ontained in the lexi on (Chomsky 1993). Stored lexi al the initial synta ti representations are literally built on the basis of the themati representations stored in the lexi on , (Belletti & Rizzi 1988). units in lude all pie es of information useful to syntax:  Being two separate omponents, lexi on and syntax are guided by two dierent sets of prin iples. between them, break In order to assure a protable ommuni ation onversion rules must be formulated. For example, the verb is assumed to be stored in the lexi on with its semanti data on the one hand, and on the other hand, synta ti argument stru ture it an reate: (i) and phoneti information about the ausative-transitive, su h as in (133); (ii) intransitive-in hoative, su h as in (134). ���� ���������� ��� ������� (133) John broke the window. (134) The window broke. Every o urren e of break 33 is stored in the lexi on with relevant pie es of information about its argument and its themati items are assumed for it, two verbs Argument stru ture of break stru ture. Two lexi al are stored in the lexi on. break in (133) ontains a subje t and a dire t break in (134) ontains only a subje t. This is obje t; argument stru ture of represented in themati (135) break: V 1 2 (136) break: V 1 grids (135) and (136). Ea h of the arguments bears a themati themati role, whi h is spe ied in the stru ture of the verb. In order to explain synta ti variability the Uniformity of Theta-Assignment Hypothesis (UTAH) has been formulated (Baker 1988: 46). UTAH assures that arguments with the same themati synta ti role need to be generated in the same position. UTAH Identi al themati relationships between items are represented by identi al stru tural relationships between those items at the level of D-stru ture. D-stru tures of senten es like (133) and (134) are assumed to be equal, and by means of move-α and linking rules, the latter is derived from the former. (137) [John [broke [the window℄℄℄ (138) [e [broke [the window℄℄℄ Linking rules relate two distin t linguisti modules (lexi on and syntax), whose prin iples are dierent in nature. Several riti isms have been noti ed about this framework (Levin & Rap- paport Hovav 2005). Namely, no pre ise diagnosti s for the identi ation of semanti roles has been eli ited. The identi ation of a spe i done only by whi h role onsidering the verbal meaning. This leads role fragmentation, onsists in the division in many dierent sub-roles, determined by the attempt to nd fundamental traits of roles (Dowty 1991). Semanti internal organization, impossible sets ( The an be ibid : onsequently, possible sets roles la k annot be distinguished by 41). orresponden e one-to-one between semanti role and argument has been questioned by Ja kendo (1972, 1983) by means of senten es like: ������� �� �������� ���������� ����� �� ��� 34 (139) Phil sold the ya ht to Mira. (140) Mira bought the ya ht from Phil. In both senten es Phil is the Sour e, Mira Theme, but they stay in dierent synta ti that there is no unique synta ti is the Goal and positions. the ya ht is the It must be assumed position generating the same semanti role. Within GB framework a well-known attempt to derive dierent argument realization patterns is ondu ted by Belletti & Rizzi (1988), hen eforth BR (1988), with respe t to psy hologi al verbs: (141) Gianni teme questo. G. fear this. (142) Questo preo upa Gianni. This worries G. (143) A Gianni pia e To G. like-3 G. likes it. sg. questo. this. BR (1988: 291) explain their theoreti starting point:  [t℄he initial synta ti representation are literally built on the basis of the themati representation stored in the lexi on . It follows that, in examples like (141), (142) and (143), the Experien er argument (Gianni ) must undergo some onversion rule, whi h puts it in a proper S-position. Lexi alist frameworks assign to the lexi on some regulatory fun tion. The existen e of rules that link one module to another would spread regularities and endanger the possibility to understand regularities. Even though senten es (141) to (143) are all o i ates, they are not (141) and (143) are not ausative, Senten es ontrary to (142). This suggests that ar- guments are not in the same synta ti same urren es of stative pred- hara terized by the same event stru tures. position and they do not share the of a hara teristi s in relation to event stru ture, if we assume the existen e ausative sub-event head. Thus GB framework among verbs, and manti 2.3 annot deal with the issue of aspe tual dieren es annot a roles and synta ti ount for the non orresponden e between se- positions. Hale and Keyser (1993 and .) Hale & Keyser's (1993) work, hen eforth HK (1993), is a histori al land- � mark whi h leads to a new denition of argument stru ture, where relations � As dened by Marantz (2012). ���� ���� ��� ������ ����� ��� ���� 35 between arguments are derived from the type of event in whi h they o Even though substantial theoreti al dieren es ur. hara terize dierent stages of HK's work, this main point remains unvaried. HK's (2002) denition of argument stru ture an be summarized in three points: - it is the synta ti onguration proje ted by a lexi al item (lexi al proje tion); - it onsists in the system of stru tural relations holding between heads and their arguments; - it is determined by properties of lexi al items, and by synta ti ong- urations in whi h they must appear. The relationship between lexi on and syntax has been rethought thanks to an unambiguous system of relations within lexi al proje tions. However, the representation of the argument stru ture of a verb is a synta ti representation of the usual sort  (HK 1993: 64), where stru tural relationships  are expressed in relation to a head: spe ier or omplement. The fa t that theta-roles are in a restri ted number dire tly derives from two grammati al (synta ti ) prin iples: (i) nature of synta ti proje tions, dened by the Unambiguous Path hypothesis (Kayne 1984) and the Single omplement hypothesis (Larson 1988); (ii) restri t amount of lexi al proje tions (V, P, A, N). In HK's (1993) approa h themati guisti � roles do not exist, they do not have lin- validity . HK (1993) argue that theta-roles are asso iated to spe i stru tural positions, themati If V is ausation. with the roles an be derived ongurationally. omplement of V, we are in presen e of a semanti relation of The NP spe ier of the higher VP bears a synta ti ausation relation, and this marks it as Agent. hen e boils down the unambiguous synta ti relation The Agent role relation of an NP to the ausal relation between two verbs. (144) (145) e1 → e2 n > e1 → e2 � � HK (1993) show that the Themati hierar hy (Grimshaw 1990) and the UTAH (Baker 1988) do not have any status in the grammar and an be derived by means of stru tural ongurations. � Where  > expresses the semanti relation that a subje t entertains with a V'. ������� �� �������� ���������� ����� �� ��� 36 If a prepositional phrase is omplement of V, a semanti is established. The NP subje t of ti relation of hange hange entertains an unambiguous synta - relation with V, being its spe ier and it is ommonly interpreted as Theme. (146) (147) e →r n > e1 → r If an adje tival phrase is omplement of V, a hanging event gives rise to a state. The NP subje t of embedded V is interpreted as Theme being subje t to hange. (148) e →s Sin e not all languages realize those stru tural relations with the same morpho-synta ti related to spe i ategory, HK (2002) abandon these stru tural positions grammati al ategories, in favor ongurations whi h are ross-linguisti ally valid. In the newest version of HK's framework (2002), there are three possible types of lexi al argument stru tures, des ribed without resorting to fun tional ategories. (149) Head Head Comp Head (150) Head Spe Head Head (151) Comp Head* Spe Head* Head* (152) Stru ture in (149) is Comp Comp Head alled ase of a single head without monadi ; stru ture in (152) is the simplest omplement and spe ier; stru ture represented ���� ���� ��� ������ ����� ��� ���� dyadi in (150) is a basi 37 type, the head proje ts both a omplement stru - ture and a spe ier; stru ture in (151) does not involve the proje tion of a omplement and an be used only in Not all verbs omposition with another head. an parti ipate in the in hoative alternation, some verbs an proje t only an in hoative stru ture.  fundamental nature of the root  not proje t a spe ier, allowing or not a break ough and This fa t is determined by the (HK 2002: 3): nominal roots ausative stru ture. an or an- For example, do not proje t the same stru tural pattern, the dieren e in argument realization depends on the lexi al nature of their roots. (153) The pot broke. HK (2002: 1) (154) I broke the pot. HK (2002: 1) (155) *The engine oughed. HK (2002: 1) (156) I oughed the engine. HK (2002: 1) The synta ti nature of operations ondu ted on lexi al items is fun- damental in delimiting the range of possible stru tures. synta ti operations to whi h HK resort ned as a f usion of synta ti nu lei  onsists in One of the main onation, whi h is de- (HK 2002: 47) where the phonologi al matrix of a head is inserted into the head that governs it, giving rise to a single verbal word (HK 2002: 48). It is on omitant to merge (HK 2002: 61), it is a pro ess of opying the p-signature of the omplement into the p-signature of the head, where the latter is defe tive (empty or axed)  � (HK 2002: 63), the  synta ti stru ture is left inta t  , sin e it is a opying parti ularly  pro ess (HK 2002: 75), rather than a movement. Conation is parti ularly important in order to orre tly derive positions of heads and it is des ribed in orporation  a ording to whi h the phonologi al matrix of the head of a omplement repla es the empty matrix of the governing head  as a spe ial kind of (HK, 2002: 11). the syntax has been shown to ree t relationships between events su h as ausation and hange of state as mu h as the relationship between entities and events des ribed by themati roles  (Marantz 2013). Theta roles lose their importan e, semanti roles are The authors reate a new framework, where:  dened by the role that parti ipants perform in the event des ribed by the verb. Relationships between individuals, and between individuals and events are dened by stru tural ongurations,  [p℄arti ipants in the event will only be denable via the role they play in the event or sub-event  (Ram hand 2008: 23). A huge onsequen e ensues: sin e parti ipants are dened as part of a � Where P-signature is a set of indexes that must mat h with indexes vo abulary items. ontained in ������� �� �������� ���������� ����� �� ��� 38 sub-event, the study of eventualities and their stru tures be omes fundamental. For this reason a resear h on argument stru ture dire tly involves the resear h on eventualities. HK (2002: stativity: 219) identify three possible me hanisms (i) prepositions of prepositions of terminal ulas, i.e. entral oin iden e ( apable of in, on, with ), oin iden e responsible for hange events; (ii) all items that introdu e a predi ate rather than a their argument stru ture ( be, ost, weight ); (iii) late a DP and an adje tival head. In all these reating opposed to overt head op- omplement in δ, whi h orre- ases, elements are responsible to establish a link between: (i) the entity and an attribute, or (ii) a lo ation, or (iii) a property. It is important to stress that stativity is generally introdu ed by stasis, whi h is generated by a relationship of entral oin iden e, this means that no energy is present in the derivation. stativity [is℄ never feature of individual lexi al items, but features of a whole predi ate , (HK 2002: 214). For example, even though (157) and (158) involve the same PP in the room, they generate dierent In all these ases,  eventualities, whi h are as ribed to dierent prepositions. a preposition of entral (157) ontains oin iden e, (158) a preposition of terminal den e. This derives in the stativity of (157) and a oin i- hange of state reading of (158). (157) With father Jim in the room , we have to wat h our language. (HK 2002: 217, ex. 25) (158) Frankie walked in the room . (HK 2002: 217, ex. 26) 2.4 Ram hand (2008) Ram hand's (2008) framework an be ins ribed within the onstru tivist ap- proa hes, sin e the author assumes that only one linguisti omponent is narrow syntax and semanti omputation  pla e to transformations, the  (Ram hand 2008: 9). The lexi on does not host any kind of rule responsible for semanti tion of verb arguments, be ause argument stru ture variability is by systemati patterns and predi table forms. not as systemati as des ribed in omposed However, these patterns are onstru tionalist frameworks (Borer 2005; Marantz 1997), be ause some pie es of semanti in the lexi on in order to a realiza- information are still as ribed ount for the irregularity of argument stru ture. Lexi al items bear features whi h instru t the item about the stru ture in whi h it an parti ipate .  [S℄uperset of ategory features it a tually spell out  ���� �������� ������ ( ibid.: 97). This is 39 alled prin iple of under-asso iation. In this regard, Ram- hand's framework diers from pure onstru tionist frameworks inasmu h it � still re ognizes some kind of information in the lexi on . Ram hand assumes that there is no need to resort to linking rules, sin e the regularity of themati ture of a predi ate is in whi h it roles is due to synta ti reated by the synta ti features. The event stru - stru ture whi h it sele ts and an appear. Ram hand's (2008: 23) approa h to argument stru ture is based on the parti ipants in the event will only be denable via the role they play in the event or sub-event . From this fundamental on ept she awareness that  pro eeds to the identi ation of primitives that are relevant to event and to argument stru ture. Ram hand (2008) sees morpho-syntax as a orrelate of the semanti s of event stru ture, as Ritter & Rosen (1998), synta ti proje tions are based on event stru ture. The rst primitive individuated is mine spe i verbal morphology ausation, whose presen e an deter- ross-linguisti ally (Ram hand 2008: 23). Ram hand takes examples of Italian unergative verbs as glow and stink, whi h do not involve an Agent even though they involve an external argument. Causation is not ne essarily parallel to agentivity. (159) Giovanni puzza. John stinks. This primitive, a ording to Marantz (1984), underlies the distin tion between internal and external argument. The relevant ategory for ausation is the one of initiator, whi h is the whose properties/behavior are responsible for the eventuality oming into existen e  (Ram hand 2004: 24). It an be realized by Agent, Instru- entity  ment, abstra t Cause or Sour e. This means that, even if important in some environments,  agentivity is not synta ti ally relevant  (ibidem ). Dieren es between external semanti � roles (su h as between Causers and Agents) are determined by the intera tion with one or more sub-events. � In parti ular, Ram hand takes this prin iple to be responsible for possible o urren es of ognate obje ts with onation verbs. The item dan e is spe ied as [init, pro , N ℄. This means that when the verb omes alone, John dan ed, the [N℄ feature is realized on a overt omplement NP. Otherwise, if the verb omes with a ognate obje t, John dan ed a tango, the [N℄ feature an be underasso iated on the item and unied with the DP omplement. � With the term role Ram hand does not identify lexi al-semanti roles, be ause arguments are arguments of predi ates introdu ed by semanti interpretation (ibid.: 44), and not arguments of a lexi al item. As onsequen e a role is determined by the spe i semanti s involved in the sub-event stru ture of whi h it is subje t. ������� �� �������� ���������� ����� �� ��� 40 For example, Causers intera t with the sole with ausative sub-event, while Agents ausation and pro ess sub-events. Teli ity is the se ond primitive that onstitutes the argument/event stru - isolable in verbal meaning  and it is  asso iated with morphology and ase marking reex  in some languages (ibid.: 25). As agentivity ture, sin e it is  has been thought to be tightly linked to external arguments, teli ity has been thought to be aused by quantized internal obje ts (Krifka 1992). Contrary to this view, Ram hand assumes that it is not the presen e of a spe i on the obje t that feature auses teli ity, sin e it exists even without internal quan- tized obje ts and quantized obje ts do not inevitably yield teli ity. Rather, a dynami event has a part-whole stru ture, implying a not ne essarily attain a resultant state, as in the entity whi h undergoes this undergoer. hange, whi h does ase of gradual hange is the se ond relevant Its presen e does not for e a teli hange. The ategory, that of reading. Sin e the attainment of a resultant state is separated from the undergoing of a hange, a third relevant sultee, whi ategory an be introdu ed, that of re- h is the entity that rea hes a nal state. Some verbs, su h as break, arrive, nd, are obligatorily teli in that they systemati ally involve result properties are properties of verbal event stru ture, not of the intera tion between dire t obje t and quantization  (ibid.: 32)� . initiator, undergoer, resultee are dened as aspe tual arguments, the a hievement of a new state. This means that  be ause they are generated by dierent aspe tual proje tions, by dierent sub-events: Causing, Pro ess, Result. There are arguments whi h are not involved in the determination of verbal aspe t, su h as the  path, whi measuring s ale homomorphi with the event  (ibid.: (160) h denes 30). initP subje t of ause init pro P subje t of pro ess pro resP subje t of result res XP � Contrary to Borer (2005) and her transfer of boundedness from DP to empty fun tional eventive heads whi h leads to a teli reading. ���� �������� ������ These layers 41 ombine in a stru ture, alled rst phase, whi h is verbal in nature, but in none of its single parts it orresponds to the lexi al verb, the same way as the split of C-proje tion. Pro P is  the heart of dynami predi ates  and  it is present in every dynami predi ate  (Ram hand 2005: 40), it is the onstitutive part of predi ates, ex eption made for statives. ResP is present only when a spe i resultant state is expressed within the predi ate. boundedness  (ibidem ). Semanti s of the onstru tion is built up re ursively from the synta ti a regular and predi table way  (ibid.: variables are present in the stru ture and this framework as an- orrelate with semanti and aspe tual not represent teli ity and it does not  stru ture in  As already said, it 42). The fa t that event an be internally omplex, identies post-Davidsonian. The primitive role types are dened as stru tural relations between subje ts and heads. dynami Initiator and Resultee are states, while Pro ess being the sub-event denotes an internal hange. This latter fa t leads to a series of theoreti al spe ulations about formal relationship that intervenes between a possible rhemati omplement and aspe tual heads. orresponden e between usual aspe tual ments and a omplishments, and verbal There is no lasses, su h as a tivities, a hievelasses of this framework, sin e they are dened in relation to the number and type of sub-events they are made up with. For example, verbs of type tives. Transitive verbs of this distin t DP obje t whi h resP℄; (ii) path, [Spe init-pro in lude both transitives and intransi- lass have a DP subje t in [Spe , an o ur in two positions: (i) initP ℄, and a undergoer, [Spe , , PathP℄. If we analyze verbs of reation with two possible readings, we an better point out the dieren e between DP obje ts. (161) Giovanni pitturò un albero sulla John a perf.3sg. paint- tree tela on-the (in un'ora). anvas (in one-hour) John painted a tree on a anvas. (162) Giovanni pitturò un albero (per un'ora). John a perf.3sg. paint- tree (for one-hour) John painted a tree. In (161), the dire t obje t does not undergo a being as result of the painting pro ess. hange, sin e it omes into For this reason the DP obje t is in [Spe , pathP℄. On the other hand, in (162), the verb involves a DP obje t whi h undergoes a hange, sin e the a tion is dire tly performed on it. The obje t is in [Spe , resP℄. ������� �� �������� ���������� ����� �� ��� 42 The pro ess sub-event is heart of dynami not events, thus stative verbs do � ontain it. Statives involve neither dynami ity nor to dene whi h is the ausation . In order orre t event stru ture for statives, Ram hand (2008) briey analyses psy hologi al verbs and their argument templates (obje texperien er or subje t-experien er). The fa t that they ta ti ally internal arguments leads to the are omposed of two arguments Devoid of the pro an have real syn- on lusion that stative predi ates rheme and theme. sub-event, statives have an init argument, whi h is argued to generate the state. (163) Katherine fears nightmares. In example above, the (Ram hand 2008: 106. Katherine, be ause of fearing nightmares. The ause of her disposition, is interpreted as orrespondent stru tural representation is reported in (164). (164) (Ram hand 2008: 56, ex.34) initP Holder init Rheme If they share the same syntax, it remains unexplained why ausative statives and non- ausative statives should dier. In other words, why (163) and (164) are dierent if their stru tures are not supposed to? In Ram hand's framework, the answer resides on the dispositions of subje t, whi h pertain to world-knowledge. (165) Nightmares frighten Mary. For this reason, the present work does not adopt this framework. though it appears useful for eventive verbs, it does not seem Even apable of a - ounting for variable behavior of stative verbs. 2.5 Borer (2005) Borer's (2005) approa h is dened as exo-skeletal. This term emphasizes the independen e from the lexi on, [it℄ listemes  (Borer 2005: 7). As we dierentiating events is teli ity, is independent of the properties of spe i an see below, the feature responsible for onditioned by quantity. Sin e the stru ture is external to the lexi on, lexi al semanti s of listemes  � In Ram hand's opinion, but we will see in further ausation. doesn't or an't play hapters that this is not true for ���� ����� ������ 43 any role in the determination of teli ity. [...℄ We must reje t any a ount of teli ity whi h ru ially relies on the assignments of some parti ular role to some parti ular argument  (ibid : 122). Argument stru ture is li ensed by fun tional synta ti stru ture, and spe i ally, fun tional stru ture that is interpreted as event stru ture , (Ibid.: 30). An under spe ied listeme (possibly a root) enters in the stru ture where it � an be verbalized by the fun tional stru ture itself . Sin e the fun tional stru ture is not dependent on lexi al features �� verbal domain is generated �� , . However, a synta ti polisemy in itself, sin e the semanti onsistent polysemy in the stru ture annot generate module interprets a synta ti in a unique way, but two distin t stru tures stru ture an re eive the same interpreta- tion. Tenny (1987, 1992, .) is the rst who proposes a of argument roles: from semanti hange in the nature roles linked to the argument semanti s, to eventive roles expressing the relationship between the argument and the event expressed by the verb. Consequently, aktionsart is a synta ti synta ti ally represented and shows sensitivity to synta ti obje t whi h is stru ture. From this perspe tive, the role assigned to a dire t obje t of a transitive verb will be the same assigned to an Ex eptional Case Marking obje t of an intransitive verb: they both ontribute to the teli ity of the event. If Kratzer (1996) severs external argument, Borer severs also the internal one. In fa t, ea h verbal argument is proje ted by a fun tional head. In other words, the verb enters dierent stru tures, whi h are endowed with dierent arguments, depending on the aktionsart they reate, and not the other way around. As it has been noted several times sin e Verkuyl (1972), quantized obje ts inuen e teli ity of predi ates. Borer applies Krifka's (1992) on eption of all verbs are inherently ateli , in the sense that they do not spe ify a ulmination point, but only a path  ( Ibid.: event quantization that assumes that  74). Teli ity arises in the stru ture, by means of a parti ular fun tional is stru turally represented, while ateli ity is that whi h emerges in the absen e of teli ity  (Ibid.: 64). Generally, teli ity is generated by the max synta ti proje tion Aspq , in the spe ier of whi h is merged a DP that proje tion, it  re eives a its usative ase and is dened as subje t of - ommanded domain, hange; the head, and orresponds to a quantity predi ate. � I leave apart the distin tion between L-head and L-domain, listeme and lexeme, whi h an be thought as the dieren e between a nude root and the ategorizing head, Arad's (2002) terms. �� The listeme brings lexi al information on the arbitrary pairing between sign and ontent. Lexi al features of listemes fun tion as sort of stru ture modiers. �� Ibid.: 30. ������� �� �������� ���������� ����� �� ��� 44 The subje t-of-quantity DP (s-o-q) that expresses a quantity is < e ># ) of ranging over an empty value ( predi ate is is not Aspq Aspq , in the head of Aspq max apable . If the the event is interpreted as teli , whereas, if the predi ate it is interpreted as ateli . Aspq max is proje ted. Dierent max tion of Aspq : (i) phonologi al In a quantized transitive stru ture the onditions are responsible for the proje (assigned a Case) and semanti al (ranged by s-o-q to a transitive teli una �� ) interpreted, giving rise predi ate; (ii) semanti ally interpreted, giving rise to an usative predi ate; (iii) phonologi ally li ensed ( ase is assigned, va u- ous head), giving rise to a transitive ateli not involve a teli head, no Aspq max predi ate. The latter ase does is stru turally present, and a generi Fun tional Proje tion (FP) is generated in its pla e. A quantity listeme merges in Aspq max and ranging over its head, re eiving a spe ier position, be oming a s-o-q usative ase from Aspq . verbal argument is merged in TP where it re eives nominative The other ase before moving upwards in [Spe , EP℄ to li ense this proje tion by ranging over the head. Arguments in [Spe , EP℄ are interpreted as originator, respe ting Burzio's generalization: assigned to a a usative ase is assigned i nominative ase is distin t hain. The tree below represents the stru ture of a quantity transitive predi ate. It is worth noting that Borer does not divide the stru ture in sub-events, ontrary to Ram hand (2008), she assumes that argument stru ture is only an epiphenomena (Ibid.: (166) 220) of the event stru ture. EP SpecN OM T max < e >E SpecN OM T AspQ max Spec2 A debate arises if ateli < e2 ># VP predi ates must be split in two groups: eventives (a tivities) and statives (states). In parti ular, statives la k the verbalizing head and have a spe ial kind of event proje tion (EP). The EP proje tion �� Subje t of quantity ���� ����� ������ 45 dierentiates statives from verbs of other eventualities and does not o stru tures provided of Aspq . ur in In other words, predi ates of all aktionsarten, ex ept for statives, involve a type of EP, and the presen e or absen e of determines their ateli ity. The presen e of Aspq Aspq ex ludes statives, be ause it implies an internal event non-homogeneity. The nature and role of the stru ture responsible for stativity remains [...℄ the dieren es between eventive and stative events should not be aptured in terms of the properties of EP, but rather in terms of properties of some other stru ture, subordinate to it . The author argues for the stru ture stative verbs should have, whi h unexplained. In fa t, Borer (2005: 265) arms:  is hara terized by a stative proje tion (SP). SP is able to pre-empt the verbalizer part. In other words, the SP is apable to invalidate the verbalizer part before the stru ture is spelled out. (167) EP SpecN OM < e >E TP Spe TP T SP VP/AP ... Pre-emption of verbal ontent in statives a senten es and for adje tival stative predi ates. ounts for opular/adje tival Whether pre-emption is an operation that takes part in the derivation of all statives is not evident. Pre-emption works in English, where stative verbs example) progressive form, unless they are This annot ombine with (for oer ed in an eventive reading. an be derived from the fa t that English progressive an ombine only with verbs provided of a verbalizer head, stranding stative verbs. The ase of ausative statives is left behind and rises some issues. In Borer's framework, ausation an apply only to verbalized stru tures, onsequently it ombine with emptied stru tures su h as stative. The derivation of annot ausative statives is left unexplained. I will not adopt this framework be ause it does not give any role to the lexi on. On the one hand, I agree that dierent readings are generated by different stru tures, but on the other hand, possible patterns in whi h a lexi al item an merge are dened by its lexi al properties. I assume that impossi- ������� �� �������� ���������� ����� �� ��� 46 bility to appear in several stru tures resides in the fa t that eventuality and argument stru ture are inter onne ted, 2.6 ontrary to Borer's opinion. Categorizers and roots An issue that on erns all frameworks, independently on their assumptions about the role of lexi on and syntax, is the way in whi h words enter the syntax. Two main theories about word formation pro esses an be found in the literature: (i) theories about double nature of words; and (ii) theories about the synta ti nature of word formation pro esses. on iliate them and proposes that words roots and a tual words. su h as n, v, a) A Marantz (2000) tries to an be formed starting from both In other words, ategorizing heads ( ategorizers, an merge above a root or above another ategorizer. ategorizer is a head bearing the required distin tive features, ne essary at LF for the interpretation of root, they are interpretive perspe tives on on epts (Panagiotidis 2010). Inner (from root) and outer (from word) derivations are responsible respe tively for regular or non-transparent meaning of derived words. derived words (lower derivation) words (upper derivation) Root- an present idiosyn rasy, while word-derived annot. The ategorizer oer es its interpretive per- spe tive on the root, of whi h sele ts a partial interpretation. This pro ess prevents a further upper ategorizer to have a (168) ess to the root. x x √ root When the root merges, it denotes meanings verbal or adje tival environments. In all the root must ombine with a ompatible with nominal, ases, at some point of the derivation ategorizer responsible for and the meaning delimitation of the root in a given merges with a the ategorizer, the ategorizer does not have the rst omplex an be further omplete a x 0 n, v, a n, v, a ategorized, however ess to the root semanti s, sin e ategorizing head already narrowed it down. (169) ategori al features ontext. When the root √ root ���� ��� ������� �������� 47 This pro ess has been spe i ally formalized by Arad (2003: 747) in the lo ality onstraint on the interpretation of roots:  roots are assigned an interpretation in the environment of the rst ategory-assigning head with whi h they are merged. On e this interpretation is assigned, it is arried along throughout the derivation . Dierent diagnosti s have been put forth in the literature in order to distinguish between word derived by roots and words derived by ategorized words. When the meaning of a root has been narrowed by a not ompletely available in the ontext. ategorizer, it is Consequently, adjun ts des ribing further spe i ations of the meaning of the root are not allowed. (170) *She taped the pi ture to the wall (171) String him up with push-pins. with a rope! In (170) by Kyparsky (1982), the verb derived. The root rst merges with the tape ategorizer is supposed to be noun- n, blo king possible a ess to the whole range of root's interpretation. Agrammati ality of (170) derives from the ontrast between sele ted range of meanings by the rst ategorizer and further spe i ation of instrument (i.e. push-pins). On the other hand, the verb string, in (171), is root-derived, sin e a further spe i ation of the instrument used to a omplish the a tion is allowed. Derivational morphology is spe i of ea h synta ti ategory, it an at- ategorized element and not to a bare-root. In fa t  any further derivation takes as its input not the root itself, but an element whose semanti and phonologi al properties have been ashed out  (Arad 2003: 2). ta h only to a We will use this eviden e in parasyntheti hapter 4 and 7 in order to demonstrate that verbs are built from roots rather than from nouns or adje tives. It is worth noting that in the present approa h, roots of an element an be omplement alled [r℄ whi h is responsible (in Roman e transitive ausative verbs) for the introdu tion of a relation between the verbal base and the internal obje t (A edo-Matellan 2006). In Roman e prexed sitive verbs, semanti in little v, 2.7 The present approa h whi h is supposed to be the verbal I adopt that line of reasoning whi h isomorphi ausative tran- ontent of roots is narrowed down when they onate ategorizer. onsiders the verbal onguration as an representation of the event stru ture (Ram hand 2008; Copley & Harley 2015; inter al.). I will propose that the onguration is divided ������� �� �������� ���������� ����� �� ��� 48 into three main layers: the lower proje tion (a Small Clause) whi h an be v (Folli & Harley 2005, inter al.); Voi e (Kratzer 1996). present or not; the verbalizer littleintrodu er of the external argument the The lower part of the derivation is supposed to be a predi ation relation, a sort Small Clause �� (SC). When SC is present, the verb has meaning (Hoekstra 1988; S häfer 2008; Folli & Harley 2005). ausative In hapter 7, I propose that the presen e of a lower SC is possible either in stative or eventive verbs Prexed Roman e verbs are assumed to ontain the expression of a rela- tion between the internal obje t and the verbal base (A edo-Matellan 2006). This relation is expressed by the prex. Adopting Mateu's (2001) approa h to argument stru ture, A edo-Matellan (2006) (hen eforth AM) argues that the prex is in the head of a [r℄ proje tion, a non-eventive relation proje ting both a omplement and a spe ier. sin e it is not introdu ed by a  ( ibid.: 12), The nature of [r℄ omplement is root, prepositional element with spatial meaning ontrary to Latin whose prexes have a prepositional nature. The stru ture proposed for transitive prexed Roman e verbs by AM (2006: 13) is expressed in terms of Figure-Ground-Path (173), where: Fig- ure is the individual who moves or is stationary, Ground is the referen e of movement, Path is the relational element between Figure and Ground. (172) La infermera assen el pa ient. (173) (Catalan) FP la infermera F R R el pa ient a- √ seu I translate AM's (2006) proposal in a tripartite argument stru ture adopted in the present work. AM's [r℄ is the odies a non-eventive relation and it orresponds to Path, whi h omponent whi h relates Figure and Ground (ibidem ). In Roman e languages, it sele ts roots be ause they are not introdu ed by a synta ti omplete prepositional element. I will show in nature of the hapter 4 and 7 that the root omplement of [r℄ is supported by Italian data, in parti ular I will resort to Kiparsky's (1982) tests about the agrammati ality of further spe i ations of �� ategorized elements. I will better dene its nature further. For simpli ity, I all it SC for the moment. ���� ��� ������� �������� 49 I argue that [r℄ head is a predi ative head (Bowers 1993) sele ting a root. Moreover, prexes are its lexi al manifestation. For simpli ity, I sti k to AM's terminology for this head, therefore I will all it r and its proje tion rP. Contrary to AM's (2006) and Mateu's (2001), I will show ( hapter 7) that the presen e of a rP proje tion in the verbal derivation is responsible only for the ausative meaning of the verb and it does not involve any reading undergone by the Theme. The hange of state hange of state reading is determined by a higher fun tional proje tion. AM's [R℄ is an eventive head proje ting a omplement but only optionally an external argument (ibid.: 8). In my proposal, the fun tional eventive head is little v ; I will argue that it omes in dierent avors whi h are responsible for dierent aktionsarten. In this respe t, I assume that AM's [R℄ orresponds to my little v sin e they both determine the eventuality of the verbs, with one dieren e, namely that little v an be responsible also for stative verbs. Dierent eventualities (stativity or eventiveness) arise be ause of dierent avors of the same higher proje tion rP does not determine only its state: per se alled little v. The presen e of a lower the a hievement of a result of the Theme, but ausative stative verbs do not involve a result, while ausative eventive verbs do. A hange of state is per eived when an individual (x) is in a dierent state t in two times of his life ( 1 and t2 ). The hange of Theme state interpretation is li ensed by the presen e of an eventive fun tional head, whi h is responsible for moving forward the time of referen e. apable of moving the time of referen e, dierent states of the same individual. Stative fun tional heads are not onsequently they A hange of state of the Theme is determined by the presen e of an eventive avor of little (174) annot represent v. ... vP v rP DP r' r la stanza aThe fun tional proje tion alled little v √ P bello is responsible for making the ������� �� �������� ���������� ����� �� ��� 50 stru ture a verbal stru ture. the language, little v However, sin e dierent is not the same for all verbs, but it avors (Folly & Harley 2005; Copley & Harley 2015; v In parti ular, in eventive verbs little & Harley 2015): 178), vappear in vbe ome ase of a in ase of omes in dierent omes in dierent avors (Copley hange of state eventive verbs (177 and vemerge for The present work analyses eventive verbs of hange (denominal parasyntheti s like verbs like exist in inter al.). omplishment with in remental Theme, denominal verbs of birthing. parasyntheti aktionsarten annerire, impilare, `to pile'; and deadje tival `to bla ken'), for this reason, only vbe ome will be analyzed in details. I will propose that stative verbs ( ausative and non- ausative) are by only one avor of little v, namely vrelation reated (175 and 176), whi h is a predi ative head establishing a relation between the external and the internal argument. The fa t that a stative verb is ausative or not is determined ongurationally, it depends on the presen e or the absen e of the lower r proje tion (refer to (175) a. hapter 7). La foto resta sul muro. The pi ture stays on the wall. vrelationP b. vrelation resta (176) a. PP sul muro La foto abbellis e il muro. The pi ture embellishes the wall. vrelationP b. vrelation (177) a. Daria mangia la mela. Daria eats the apple. rP DP r √ il muro a- bella ���� ��� ������� �������� 51 vbe b. vbe ome mangia (178) a. ome DP la mela *Il mare mangia la spiaggia. (Folli & Harley 2005: 14) The sea eats the bea h. b. Il mare si è mangiato la spiaggia. The sea eats the sea up. vbe . vbe ome ' ome si SC DP Vadjectival la spiaggia mangiato The spe ier of AM's FP proje tion is the external argument, whi h is interpreted as the individual who moves, namely to the most prominent individual of the eventuality. I argue that this proje tion with the Voi e P an be translated (Kratzer 1996; Pylkkänen 2002; Harley 2012a). In order to understand the relationship between external arguments and their predi ate, it is ne essary to re all Kratzer's (1996) analysis about the severing of the external argument from the verb. In the last de ades it has been noti ed that internal arguments an very often inuen e the interpretation of the whole predi ate (Marantz 1984), ontrary to external arguments. (179) a. throw a baseball b. throw support behind a . (180) throw a boxing mat h d. throw a party e. throw a t a. lan iare la palla throw the ball b. lan iare una sda hallenge andidate ������� �� �������� ���������� ����� �� ��� 52 . lan iare una provo azione provoke d. lan iare il dado roll the di e e. lan iare la ma hina ai 100 km/h hurl the ar to 100 km/h f. lan iare un grido ry out g. lan iare un programma start a program h. lan iare uno sguardo ast a gaze Inspired by neo-Davidsonian theories whi h assume that arguments are introdu ed by predi ative heads, Kratzer (1996) develops a theory whi h treats subje ts as arguments of fun tional head, with the result that Agents and Holders are no more dire t arguments of the lexi al verb. an explain why internal obje ts while external ones do not. This move an inuen e the interpretation of the verb, Internal obje ts are proper arguments of the verb, while external ones are added via a distin t fun tional head. Kratzer's analysis is based on two theoreti al assumptions: arguments are introdu ed by heads; stru tural heads. ases ( nom; a ) are assigned by fun tional Kratzer assumes, after Hung (1998), that external arguments are introdu ed by a head alled voi e. Contrary to Hung (1998), Kratzer argues its fun tional nature for four main reasons: a. it explains defe tive distribution of voi e. If it were a lexi al head, the defe tive distribution within the paradigm of the same verb ould not be explained. b. voi e is related to a ase assignment, and we know that fun tional, and not lexi al, heads assign stru tural . serial verbs ase. an share the external argument, be ause verbal om- plex presents one ine tional morpheme and one external argument. A ordingly, external argument is introdu ed by an ine tional head =fun ( tional). d. this type of analysis is in a ordan e with previous a phrase stru ture (Pesetsky 1989; Johnson 1991). ount to English ���� ��� ������� �������� 53 Thus, external arguments are introdu ed in the derivation by a fun tional head and they identi ation (181) ombine with it by means of a semanti operation alled event (EI). Event Identi ation: he, hs, tiihe, hs, ti → he, hs, tii Lambda expression in (182) helps us to understand the role of EI for a transitive verb su h as buy. (182) λxe λes [Agent(x)(e)] [buy(T heme)(e)] → λxe λes [Agent(x)(e) & buy(T heme)(e)] If external arguments are introdu ed by a distin t fun tional head, what determines their argument role? The lexi al verb introdu es an event argument whi h denes its eventuality, the external argument is introdu ed by another fun tional head whose eventuality needs to be quently the themati ompatible with the one of the lexi al verb, onse- role assigned to the external argument is still related to the eventuality of the lexi al verb (whether a omplishment, a tivities, statives, ...). In the present work, Voi e omes in two avors ( à la Folli & Harley 2005), in relation to the eventuality of the predi ate; it must a of little v in order to get the derivation spelled out (refer to example, for ausative eventive verbs, a predi ate of for es, and it V oice auser hapter 7). For takes as its omplement omposes with the external argument whi h is interpreted as the Causer (183). For as its ord to the avor ausative stative verbs, omplement a predi ate of situation and V oicesour e takes omposes with the external argument whi h is interpreted as the sour e of the internal argument state (184). V oice (183) auser P V oice DP V oice auser ' vbe auser vbe ome P ome rP DP r prex √ ������� �� �������� ���������� ����� �� ��� 54 V oicesour (184) eP V oicesour DP V oicesour e' vrelationP e vrelation rP DP r √ prex We will adopt the for e-dynami approa h to ausation (Copley & Harley 2015; Copley & Martin 2014; Copley 2015; Copley & Wollf 2014) with the due hanges to a ount for are generated by energeti ausative stative predi ates. Eventive predi ates (linguisti ) for es whi h orrespond to energeti ( on eptual) for es. Causative stative verbs annot be generated by energeti for es (by denition). I assume that stative ausative verbs are generated by a virtual `for e' alled abdu tion whi h is introdu ed in the system by the speaker. 2.8 This Con lusions hapter reported theoreti al bases of dierent frameworks on argument stru ture and aktionsart. We saw that frameworks an be divided into two lasses depending on the weight they attribute to lexi on and syntax. We saw that the stru ture of stative verbs is often not dened. This dissertation adopts a general l-syntax framework in whi h arguments are introdu ed by fun tional heads whi h are linked to event stru ture. Verbal l-syntax determines the syntax and semanti s (aktionsart) of the prediate and its arguments. Spe i theoreti al along the dissertation. hoi es will be presented in dedi ated paragraphs Chapter 3 Parasyntheti verbs 3.1 Introdu tion The analysis of verbs with learly identiable morphologi al onstituent parts is parti ular enlightening on the nature of argument stru ture. In parti ular, the possibility of determining number and nature of morphologi al building blo ks leads to a deeper understanding of the fun tional synta ti blo ks responsible for their building ombination. For this reason, the present dissertation fo uses on the argument stru ture of deadje tival and denominal verbs. These verbs are all formed by means of a morphologi al derivational pro ess alled parasynthesis. This hapter will present parasynthesis and dis uss theoreti al issues about the nature of derivational steps in this lass of verbs. We will see a typology of parasyntheti synta ti and semanti properties of this verbs by Ia obini (2004) based on lass. We will further fo us on two sub-groups that will be the aim of the present dissertation, namely a group of denominals and a group of deadje tivals. Denominal parasyntheti verbs analyzed in the present dissertation are an be ausatives and their semanti s make something a N , where N a atastare, `to pile up'. Deadje paraphrased with  is the base noun, su h as impilare, tival parasyntheti are `to pile', ausatives and their semanti s something A, where insozzare, `to soil. 3.2 an be paraphrased by means of  A is the base adje tive, su h as abbellire, `to verbs make embellish', Parasyntheti verbs Parasynthesis is a derivational morphologi al pro ess present in all Roman e languages. It yields verbs equipped of a prex and a sux and it yields 55 ������� �� ������������� ����� 56 dierent semanti s. (185) Il ommesso in-s atol-ò The ashier in-box- il 3sg.past. det.sg.m. regalo. gift. The ashier boxed the gift. (186) La musi a in-stupid-ì The musi in-stupid- i ragazzi. 3sg.past. det.pl.m boys. The musi made the boys dumb. (187) Gli operai a- atast-arono la spazzatura. det.pl.m workmen a-heap-3pl.past. det.sg.f rubbish. Workmen made a heap of trash. Parasynthesis originates from a reinterpretation pro ess in Late Latin, a period in whi h the semanti ontent of Latin prepositional prexes blurs gradually. This leads to a reinterpretation of prexed denominal and deadje tival verbs: they are interpreted as synonymous to their non-prexed parts, leading to a loss in the semanti s of prexes whi h be ome ounter- ontentless derivational tools (Ia obini 2004). Traditionally, the denition of parasynthesis, due to Darmesteter (1894), is based on a lexi al dened as riterion. Spe i ally, a verb, a noun, or an adje tive are parasyntheti , when they are omposed of a prex and a sux and the intermediate stage of derivation is not attested in the lexi on (Ia obini 2004). We will see that this denition is problemati , sin e the la k of a derivational produ t in the lexi on does not imply the oddity of the morphologi al resulting verb/adje tive/noun. I will leave aside parasyntheti nouns and adje tives and will fo us only on verbs. Parasyntheti verbs ontain three re ognizable parts: a prex, a nominal � or adje tival base, a sux . (188) Prex + Base (N or A) + verbal sux Only three prexes s- an form parasyntheti verbs, namely ad-, in- and without privative meaning (Ia obini 2004). These prexes are produ tive only with parasyntheti of the formations. They do not ompound (Ia obini 2004). ontribute to the semanti s Contrary to Ia obini and a ording to Bertinetto (1986), who argues that these prexes have an a tional meaning �, I assume that they are morphologi al manifestation of ausation. words, they an appear only if In other ausative meaning arises from the stru ture. � We will see that the derivational or ine tional nature of the sux is at stake in the debate about parasynthesis. The reader should please onsider it in more general terms. � They ontribute to the semanti s of the a quisition of a new state. ���� ������������� ����� -ire The 57 onjun tion group is produ tive only with these three prexes, otherwise the only produ tive onjun tion in ontemporary Italian is -are. From Latin and Late Latin, the original lo ative semanti s introdu ed by these prexes swit hed progressively to a related one, namely that of hange of state (Ia obini 2004). Morpho-synta ti stages of parasyntheti sial in nature and number. verb formation are ontrover- The order in whi h prex, base and sux are ombined together is matter of debate in the literature and three positions emerge: a. simultaneous addition Darmester 1890; Ia obini 2004) [pref [X]N suf f ]N/V b. suxation followed by prexation (S alise 1990) [pref [[X]N suf f ]V ]V . prexation followed by suxation (Corbin 1987) [[pref [X]N ]V suf f ]V The rst hypothesis onsiders that the prex and the sux onstitute a dis ontinuous morpheme, be ause the otherwise expe ted intermediate morphologi al item is not attested in the lexi on (Cro o Galéas & Ia obini 1993). However, it is problemati for three reasons of dierent nature. First, a formal issue (S alise 1990: 218) onsists in the fa t that the derivation does not respe t the binary bran hing hypothesis (Arono 1976). Italian does not present other ases of ir umxes, making this kind of derivation an ad ho pro edure. Se ond, the denition of ir umx does not t the ase of parasyntheti In a series su h as Sp. embalsamar `to embalm' (← bàlsamo `balm') [...℄, it is di ult to argue that there is a dis ontinuous morpheme [en- ... -ar℄ based on a supposedly obligatory o-presen e of the sux and the prex. We have to bear in mind that there are orresponding synonymous verbs without the prex: balsamar, mas arar [...℄. (Serrano-Dolader 2015: 531). We an talk of two distin t morphemes that  [...℄ are jointly atta hed to a base  (ibidem ). Third, Italian employs two dierent in- prexes: (i) one with negative axes, in fa t  semanti value (190) whi h pre edes adje tives (S alise 1990: 220); (ii) one with intensive semanti (189) in + verb value (189) whi h pre edes verbs. ������� �� ������������� ����� 58 a. in + rompere = irrompere in + break = burst into b. in + porre = imporre in + pla e = impose (190) in + adje tive a. in + esperto = inesperto in + expert = unexpert b. in + edu ato = inedu ato in + polite = impolite As in- with aspe that it tual meaning ombines with verbs should be an eviden e omposes with verbs also in ase of parasyntheti verbs. This leads to the hypothesis that two derivational steps must be assumed for parasyntheti verbs: the rst involves the sux and a hange in ategory of the nominal or adje tival base, the se ond further adds the aspe tual prex. (191) -are, -ire 1. A/N + 2. [A/N − are/ire]V = [A/N − are/ire]V + a-, im-, s- = parasyntheti produ t The produ t of the rst stage is possible, but not ne essarily attested. At the se ond stage, the a tual parasyntheti word is generated. The fa t that intermediate produ ts are not attested is onsidered as unproblemati both by S alise (1984: 204; 1990) and Corbin (1980: 191), sin e it is possible for the intermediate produ t to be a possible but non-attested word. The se ond hypothesis has several advantages: it does not assume a spei derivational me hanism su h as bran hing hypothesis. ir umxation and it respe ts the binary However, it still does not explain morphemes order [...℄ does not explain the relationship between prexed verbs and non-prexed verbs with the same stem that are not attested, and negle ts the widely heterogeneous hara ter of su h relations  (Serrano-Dolader 2015: 530). whi h do not respe t the mirror prin iple, and furthermore it  The third hypothesis also assumes two distin t derivational steps, but it harges the prexes of the Su h a theory is problemati mati al hange of grammati al ategory (Corbin 1980). as Italian prexes normally do not ategory (S alise 1995: 477). Thus again an ad ho hange gramme hanism is assumed. The nature of verbal suxes by Serrano-Dolader (2015: 528): hanges in the three theories, as pointed out ���� ������������ ������������� ����� 59 Another possible interpretation is to argue for the ine tional as well as derivational nitive is should be tive is hara ter of the innitive ending. If the in- onsidered to be part of the verbal paradigm, its ending lassied as ine tional. If, on the ontrary, the inni- onsidered to be part of a derivational paradigm, then its ending is derivational. The hallenge of parasynthesis is well represented by these three theories whi h all present some issues. The following se tions des ribe denominal and deadje tival parasyntheti lasses with parti ular attention to the sub-groups whi h are obje t of the present study. 3.3 Deadje tival parasyntheti Italian deadje tival parasyntheti words they are verbs verbs show an ingressive meaning, in other ausative and they attest that the obje t is more A as result of the event expressed by the verb. A ording to Ia obini (2004), the resulting grade is left unexpressed, for example in abbassare (`to lower') the grade of attainment is not spe ied with respe t to the initial state. This happens independently from the base adje tive, we will see in se tion 7.3.1.1 that this is due to synta ti nature of the base whi h is not a ategorized adje tive, but rather a root. The majority of deadje tival parasyntheti verbs alternates between a transitive and a pronominal in hoative form (Ia obini, 2004). (192) a. Maria ha innervosito sua sorella. Mary annoyed her sister. b. Maria si è innervosita. Mary got annoyed. (193) a. Daria ha intossi ato sua sorella. Daria intoxi ated her sister. b. Daria si è intossi ata. Daria got intoxi ated Some verbs alternate between a transitive and a non-pronominal in hoative form. (194) a. Daria ingrassa il maiale. Daria fattens the pig. ������� �� ������������� ����� 60 bello, `beautiful' brutto, `ugly' giallo, `yellow' bian o, `white' nero, `bla k' grande, `big' stupido, `stupid' > > > > > > > > a-bell-ire, `make (more) beautiful' im-brutt-ire, `make (more) ugly' in-giall-ire, `make (more) yellow' im-bian -are, `make (more) white with an addition of white s-bian -are, `make (more) white with a loss of another a-nner-ire, `make (more) bla k' in-grand-ire, `make (more) big' in-stupid-ire, `make (more) stupid' Table 3.1: Morphologi al b. onstituents of DPVs. Daria ingrassa. Daria puts on weight. In the present study, I fo us on the transitive form of verbs whose paraphrases orrespond to make N (more) A, where A orresponds to the base adje tive and N to the ae ted dire t obje t. I identied 221 deadje tival parasyntheti verbs (hen e DPV), among a e are (`to blind'), addol ire (`to sweeten', `to alleviate'), irrigidire (`to stien'), sgrezzare (`to make rough'). The full list is reported in appendix. them: The base adje tive remains a essible in the verb semanti s, as expli itly reported in se tion 7.1. The dis ussion about morphologi al omponents of DPVs is ondu ted in hapter 7, where the nature of prexes and base are analyzed in depth. 3.4 Denominal parasyntheti verbs Italian denominal parasyntheti verbs an be divided into several sub-groups in relation to the semanti s introdu ed by the base. The relevant meaning of the base involved in the to dene and it is onstitution of denominal parasyntheti s is di ult onditioned by ultural and en y lopedi knowledge of speakers (Ia obini 2004). Three sub-groups of denominal parasyntheti s ( ausative, lo ative and instrumental) are identied by means of paraphrases. The distribution of dierent verb groups among prex type is proposed in table 3.2 (page 61). Instrumental denominal parasyntheti s express the instrument by means of whi h the resultant state is attained. This means that they are ausative in nature and the verb fo uses on the instrument employed and not on the olor' olor' ���� ��������� ������������� ����� ad Semanti s Table 3.2: (%) in (%) 61 s (%) Causative 40 36 24 Lo ative 24 76 0 Instrumental 45 16 39 Distribution among prexes of denominal parasyntheti s sub- groups (Ia obini 2004, table 2). result itself. � The parent noun is in the instrumental ase, so they an be onsidered instrument verbs in Clark & Clark's (1979) terminology. that abbottonare `to button up' in example (195) states Mary did something to ause that oat to be losed by means of buttons. It is lear that the resultative- ausative semanti s is introdu ed by the verb For example, the verb abbottonare, but only be ause the instrument employed to perform that event is the base. (195) Maria abbottonò il appotto. Mary buttoned the oat up. In other words, these verbs ontain two semanti portions: the ausative Mary did something to ause the oat to be tied, and the instrumental portion, by using buttons (Clark & Clark 1979: 778). portion, Lo ative parasyntheti parasyntheti verbs onstitute over 40% of the entire denominal lass and they divide between lo atum or lo ation verbs, in relation to the ase in whi h the parent noun is (Clark & Clark 1979: 769). Lo atum verbs des ribe  the position of one thing to respe t to another  (Clark & Clark 1979: 770), the parent noun is in obje tive responds to the individual whi h is reorganized in spa e. ase and it or- For example, in did something to ause it that the anapé had some butter on it, the butter is reorganized in spa e on the anapé. (196), Daria where (196) Daria imburrò la tartina. Daria buttered the anapé. Lo ation verbs des ribe the lo ation in whi h the obje t is repla ed and the parent noun is in lo ative (197) Daria infornò D. ase. la put-in-the-oven the tartina. anapé. Daria baked the anapé. � In Clark & Clark's (1979) terminology. ������� �� ������������� ����� 62 In (197), the anapé is relo ated into oven. Base nouns of lo ative verbs orrespond to the pla e in whi h the obje t is pla ed. lo atum verbs The present study fo uses on This ategory The base noun of orresponds to the obje t whi h is relo ated. ausative denominal parasyntheti verbs. an be further divided into three subgroups in relation to the ausative semanti s of their paraphrases (Ia obini 2004) and the relationship expressed between the subje t and the nal state denoted by the base noun. (i) (ii) (far) diventare (un) N, `(make [something℄) be ame (a) N' (far) diventare ome (un) N, `(make [something℄) be ome as/similar to (a) N' (iii) ausare/prendere/a quisire N, `(make [something℄)take, a quire N'. In (i), at the end of the event the obje t a quires the semanti features of the verbal base (198, 199). (198) Giovanni ha appallottolato la arta. John balled the paper up. (199) Giovanni ha a atastato la legna. John dumped timber. Giovanni did something to ause it that the paper be ame (like) (199), Giovanni did something to ause it that the timber is part of In (198), a ball. In a dump. In the se ond sub-group, the subje t be omes similar to the individual denoted by the base that builds the predi ate, (200). (200) Giovanni è in artape orito. G. perf.. is in-par hment- John shrivelled up. Example (200) shows that, state seems internally ontrary to the previous group, the resultant aused, in fa t the ausative event responsible for the resultant state of the obje t remains undened. These verbs are mostly intransitives. The third group of ausative denominal parasyntheti verbs is formed on abstra t noun whi h often express a psy hologi al quality. of (201) onsists in the Italian verb is The paraphrase John did something to ause it that Mary is tired, literally formed from the noun fati a, `eort'. ���� ����������� (201) 63 Giovanni ha aati ato Maria. John tired Mary. The present study fo uses on the rst sub- lass, whose paraphrase is (make) X be ome(s) an N (hen eforth BNs). In parti ular, I am interested only in this sub- lass be ause, as reported in the pseudo-resultative (202) hapter 4, it an parti ipate in onstru tion (Levinson 2007). Quando Daria mangia i bis otti, li sbri iola sottili. Whenever Daria eats bis uits, she rumbles them thin. I identied 57 (2014). The ausative denominal parasyntheti omplete list of them is reprodu ed in appendix with trans- lation and synta ti four synta ti verbs in Devoto & Oli onguration. The 57 sele ted verbs are divided into ongurations: transitive, alternating transitive/intransitive, pronominal intransitive, reexive. (203) L'orafo ha inlzato le perle. (Transitive) The goldsmith pier ed (and thread together) pearls. (204) a Medusa impietriva hiunque la guardasse. (Trans.-Intrans.) Medusa petried whoever looked at her. b Nella lotta alle di oltà, l'animo impietriva. In the ght against di ulties, spirit be ame a stone. (205) Il lo si è aggrovigliato. (Pron. Intrans.) The line twisted. (206) I ragazzi si sono a oppiati per l'eser izio. (Reexive) Students paired up for the exer ise. The present study will on a spe i 3.5 This onsider only the transitive group, sin e it fo uses se ondary predi ation that involves a dire t obje t ( hapter 4). Con lusions hapter des ribes Italian derivational pro ess alled parasynthesis and reports dierent morphologi al theories whi h try to a ount for its distin - tive traits, namely the position and nature of prexes and suxes. 64 ������� �� ������������� ����� Part I Non ambiguous verbs 65 Introdu tion The rst part of the present dissertation analyses the behavior of denominal parasyntheti verbs whose paraphrase is make X be ome an N (hen eforth BN). It examines the grammar of the pseudo-resultative onstru tion (Levinson 2007), whi h expresses an adje tival se ondary predi ation on the impli it entity denoted by the base. (207) John piled books high. → John made a high pile of books. Contrary to English, Italian shows morphologi al agreement on adje tives. This parameter is parti ularly useful in the study, onrming Levin- son's (2007) assumption for the stru ture of pseudo-resultatives. (208) Giovanni ha G. im-pila-to has im-pile- i perf. det.m.pl. libri m.pl. book. ����. high- m.pl. Giovanni piled books high. We will see in hapter 5 that Italian and Fren h, both Roman e languages, do not behave in the same way with respe t to pseudo-resultative onstru - tion. This dieren e is tied to the general issue of se ondary predi ation in Italian, and onrms the higher availability of adje tival se ondary predi ates in this language ompared to other Roman e languages (Folli 2001). This study belongs to a wider urrent resear h about se ondary predi ates and their produ tivity in Roman e languages (Talmy 1991, 2000; A edoMatellan 2012; Folli 2001; inter alia ). Italian shows some pe uliarities in this language family, namely prepositional resultatives are highly produ tive and adje tival resultatives are partially produ tive (Folli 2001; Napoli 1992). For this reason, we investigate Fren h pseudo-resultative of a semanti lower a onstru tions, by means interpretation task ( hapter 5). Results of the experiment reveal essibility of this onstru tion in Fren h than Italian. A hypothesis for this asymmetry is made: phonologi al transparen y is lower for denominal Fren h verbs, making impossible to establish a predi ative link between the base and the pseudo-resultative adje tive. 67 68 We will see that Italian speakers prefer synonymous adverbs as prediates of the impli it entity, even if they a A Magnitude Estimation task (Bard tive speakers shows the higher a ept pseudo-resultative adje tives. et al. 1996) ondu ted on Italian na- eptability of adverbs than of synonymous adje tives. This is an expe ted result, sin e BNs are resultative verbs, whi h in lude two possible layers that ation in rP, or the verbal layer an be modied by the adverb: v P. In other words, adverbs the predi- an have two readings, one of whi h modifying the impli it entity (low s ope) and the other modifying the verbal proje tion (wide s ope) (se tion 4.5). Chapter 4 Parasyntheti denominal verbs 4.1 This Introdu tion hapter analyses the behavior of 57 Italian parasyntheti verbs with nominal base and semanti s of type (make) X be ome(s) an N, where N is the base and is modied in the pseudo-resultative onstru tion. In the last de ade new insights about the vast topi i ates distinguish a new tatives. This lass is lass whi h was previously of se ondary pred- lassied within resul- alled pseudo-resultative predi ates (Levinson 2007) (PR), and is formed by an adje tive whi h predi ates over the impli it entity of the primary predi ate. (209) → John piled books high. of John. Pseudo-resultatives The pile is high as the result of the a tion an be formed only on an impli it reation verbs (ICV). These verbs denote the presen e of an impli it entity a modi ation by the adje tive. Their semanti s arms the essible for oming into being of an entity whi h is not otherwise present in the argument stru ture. will see that this synta ti feature distinguishes them from expli it We reation verbs. This hapter analyses the grammar of pseudo-resultatives in Italian, a Roman e language whi h normally does not a predi ations. For this reason, a semanti on Italian native speakers. ept true adje tival se ondary interpretation task was Results of this experiment ondu ted onrm the a ept- ability of PR, even though speakers informally report a preferen e for synonymous adverbs. A Magnitude Estimation task (Bard ondu ted in order to investigate the dieren e in a et al. 1996) has been eptability of pseudo- resutlatives and of synonymous adverbial modi ations (se tion 4.5.4). 69 ������� �� ������������� ��������� ����� 70 Morphologi al agreement on Italian adje tives in gender and number on- rms Levinson's analysis for PR adje tives. Namely, the adje tive is a prediate of individual whi h predi ates over the verbal base. As the base is nonategorized, it does not have the possibility to Consequently, the adje tive is for ed to he k adje tival φ-features. he k its features against the rst - ommanding DP, the dire t obje t. Next, I propose an analysis for PR adverbs, showing that they an be interpreted with either narrow or wide s ope. 4.2 Levinson's approa h to pseudo-resultatives (PR) This se tion presents Levinson's (2007) approa h to the pseudo-resultative (PR) onstru tion. Spe i features of PR distinguish them from true se - ondary resultative predi ations. PR are systemati ally built on impli it re- ation verbs (ICV). The fa t that Italian allows PRs is parti ularly interesting sin e it annot normally produ e adje tival resultatives (se tion 4.2.5). 4.2.1 Impli it Creation Verbs: features Pseudo-resultative this lass onstru tion is built on impli it an be identied by four spe i fall under the reation verbs. Verbs of features (Levinson 2007): (i) they lass of goal verbs (Clark & Clark 1979) (ii) they involve a shadow argument; (iii) the presen e of the dire t obje t is mandatory; (iv) the dire t obje t is an ae ted obje t. ICVs fall under the lass of goal verbs, in the terminology of Clark & Clark (1979), be ause the morphologi al base of the verb (parent noun) is in goal ase. The verbal base denotes the entity brought into existen e, with no mention to the substan e whi h it is made of. Senten e (210), by Clark & Clark (1979), onstitutes an example of goal verb and (211) represents its paraphrase. (210) Edward powdered the aspirin. (211) Edward did something to ause it to ome about that [the aspirin was powder℄. the shape, entity, form [...℄ denoted by the parent noun omes to exist by virtue of the a tion denoted by the verb  (Clark & Clark 1979: 774). English ICVs have The main semanti feature of Goal verbs is their fa titivity:  ���� ���������� �������� �� ������������������� ���� been shown to orrespond mostly to this 71 lass of verbs, where the parent noun denotes the entity brought into existen e in the event. (212) John piles the books. (213) John did something to ause the books to be into a pile. ICVs involve a shadow argument (Geuder 2000: 79), an argument whi h is not expli itly present in the argument stru ture of the verb and whi h denotes the reated obje t. In (212), the impli it entity pile is the shadow argument denoting oming into being as the result of the main predi ate. ICVs do not only involve a shadow argument, but a tually entail its reation as result of the event. ICVs require a Theme argument in the dire t obje t position; this denotes that the obje t has been relo ated in the spa e by the Agent. The la k of dire t obje t makes the senten e agrammati al. (214) John piled *(the books) (215) Mary aligned *(the boxes). The dire t obje t is ae ted. In the denition by Clark & Clark (1979: 774) denotes the entity brought into existen e, with no mention of the substan e from whi h it is made . this ae ted obje t [...℄ 4.2.2 Impli it reation verbs dier from expli it re- ation verbs impli it reation verb must not be reation verb sin e they refer to two dierent The term ICVs onstitute a onfused with the generi term derivations. lass of verbs derived from a root whi h is predi ate of individuals and responsible for naming an entity that the result of an event. In other words, ICVs entail the omes into being as reation of an entity whi h is not otherwise part of the argument stru ture of the verb (Levinson 2007: 17). (216) Teresa braided her hair → Teresa made a braid as result of braiding. For example, in (216), the impli it entity the verb braid whi h represents the event of braid is not an argument of reating a braid. The reated entity remains impli it in the argument stru ture and the verb expresses its reation, not only the way by whi h it has been reated. On the other hand, in the argument stru ture of expli it the reation verbs, reated entity is an expli it argument and is present in the argument ������� �� ������������� ��������� ����� 72 stru ture. The verb des ribes the way in whi h the being. In the following example, the bra ialetto, reated entity omes into `bra elet', is produ ed by means of a braiding pro ess. (217) Giovanni intre ia un bra ialetto. G. braids a bra elets. This intuition is further supported by and English on benefa tive verbs, do not a (218) onstru tions. ICVs, eviden e from Finnish ontrary to expli it reation ept benefa tive appli atives. *Hän leti-tti s/he ross-linguisti minu-lle minu-n aus.pst 1sg.-all 1sg-gen. braid- tukka-ni. poss-1sg. hair- (Levinson 2007, ex.195). She braided me my hair. (219) Hän leti-tti s/he braid- minu-lle pullapitko-n. aus.pst 1sg-all braided.bread-a She braided me a braided bread. ICVs and expli it reation verbs do not manifest the same behavior and are not built from the same stru ture. We will see in the following se tions that the derivation of ICVs involves an indire t relationship between the dire t obje t and the impli it entity, ontrary to expli it reation verbs in whi h the relation is dire t. 4.2.3 Impli it entity is a root This se tion analyses the synta ti nature of the impli it entity. larly, we need to establish if the entity is a ategorized root. Parti u- ategorized element or a non In fa t, the dieren e between a ategorized element and a non- ategorized root determines dierent behavior both from a synta ti and from a semanti point of view. For example, (a)teli ity of denominal verbs is determined by the (un)boundedness (Pustejovsky 1991; Ja kendo 1991) of the nominal root in dire t obje t position (Harley 2005). (220) John ate apples. (221) John ate the apple. However, (a)teli ity of ICVs is not inuen ed by the nature of the root on whi h they are built: senten e (222) does not built up. onrm how many piles John ���� ���������� �������� �� ������������������� ���� (222) 73 John piled books. Levinson (2007) argues that this is an eviden e of the indire t relationship between the base root and the ategorizer. The relationship is mediated by a relational element. I will argue further that the impli it relational elements IN and TO argued by Levinson argue that, orrespond to the r head. Furthermore, I will ontrary to English this element is morphologi ally expressed by the prex. An indire t relation between the base and the in lo ative verbs, su h as (223) ategorizer is present also box. John boxed books. When root is embedded in a relational stru ture it annot inuen e (a)teli ity of the whole verbal predi ate. (224) vP DP John v' v SC DP books PP P √ P √ box Therefore, Levinson (2007) proposes that the semanti relation between root and internal obje t is mediated by a relational element, prepositional in nature. both (225) (226) In this respe t, the r head is similar to the impli it preposition: reate non-eventive relations. a. John braided his hair. b. John made his hair in a braid. a. Jill boxed his books. b. John put his books in a box. With respe t to lo ation/lo atum verbs, ICVs do not express a simple lo ative relation between Theme and Goal. Sin e Goal is made of Theme, a sort of physi al/material relation needs to be a ounted for and it is as rib- able to the semanti s of the proper verbal part, another relational element ������� �� ������������� ��������� ����� 74 must mediate between root and internal obje t (Levinson 2007: 45). I argue that the r head is responsible alone for the orre t derivation. Levinson (2007) argues that the relational elements in the derivation are: a ongurational and a mode omponent semanti ally similar to proper preposition omponent to in ; (Krat h 2002), whose semanti s is predi ate- dependent and states that Theme and Goal share the same lo ation. For in, in (227) is responsible for the introdu hair and braid, namely hair is in a braid. to, in (228), establishes a link between John example, the expli it preposition tion of a lo ative link between The expli it and the store, onal preposition assuming that they are in the same lo ation at the end of the relevant event. In ICVs, this impli it preposition establishes a relation between an entity (the DP) and the impli it entity, stating that they arrive in the same lo ation. We an imagine that the impli it relational elements assumed to link root and impli it entity share the same relevant semanti s, stating respe tively that hair is in a braid and that braid and hair share the same lo ation being made of the same material. (227) His hair in a braid. (228) John go to the store. The semanti s assumed by Levinson (2007) for the two relational impli it elements is the following: (229) (230) IN = λf<e,t> .λye .λss .∃xe .f (x) & being-in(s)(x) & theme(s, y) TO = type-theoreti ally va uous (agreement with ause introdu ed by v) The derivation proposed for ICVs by Levinson (2007) is therefore the following. (231) vP vgoal SC DP TO her hair IN √ P braid IN and TO, in my derivation for the orrespond to the r head whi h is responsible orre t semanti s. However, I argue that the ausative meaning arise ���� ���������� �������� �� ������������������� ���� from the onguration and it is not introdu ed by little realizations of what Levinson v. 75 Prexes are lexi al alled IN TO. We have seen that the root nature of the impli it entity is derived from the fa t that (a)teli ity of ICVs is not inuen ed in relation to its (un)boundedness. For this reason, the presen e of relational elements mediating the relationship between the root and the internal obje t is proposed. The synta ti stru ture of impli it reation verbs, given in (231), involves two relational elements between dire t obje t and impli it entity. They assure the same lo ation of hair and braid, and the same material of these individuals. 4.2.4 Pseudo-resultative onstru tion Pseudo-resutlative onstru tions ontain an adje tive whi h predi ates over the verbal root denotating the impli it entity, su h as in (232). (232) A John braided his hair tight. ording to Levinson (2007: 33 .), in (232), adje tive tight is neither a pure resultative as in (233), nor an obje t depi tive as in (234), nor a predi ate of events as in (235). (233) John hammered the metal at. (234) John hammered the metal hot. (235) John hammered the metal hot 6= the event of hammering was hot. In example (233),the se ondary predi ate at rea hed by the obje t as result of the a tivity of adje tive, an obje t depi tive � hot introdu es the nal state hammering. In (234), the modies the state of the obje t during the hammering : the metal is already hot during the event. In (235), the interpretation of hot as an adverbial modi ation is not allowed, an event of hammering annot be hot. In (232), tight does not introdu e the nal state of the obje t as result of the a tivity of braiding, sin e it is not hair whi h is tight, rather the braid. In event of other words, the adje tive does not modify the state of the obje t during the � In obje t oriented depi tive onstru tions, the adje tive des ribes  an eventuality (state) pertaining to one parti ipants of the main predi ate  (Halliday 1967) at the time at whi h the main predi ate o urs.  The depi tive predi ation onstru tions have been lassied with two types, namely Subje t-Oriented Depi tives (SODs), and Obje t-Oriented Depi tives (OODs). It is a SOD if the subje t of a d-predi ate is subje t in a senten e; it is an OOD if the subje t of a d-predi ate is a dire t obje t  (Noh 2003: 22). ������� �� ������������� ��������� ����� 76 event of braiding, it is not an event modier and it adverb whi h would require -ly annot be a resultative morphology. Pseudo-resultative predi ates of the nal state of the individual denoted by the base-root of impli it Based on ross-linguisti reation verbs. eviden e, Levinson (2007) assumes that pseudo- resultative predi ates are adje tival in nature. She reports eviden e from Catalan. Catalan shows morphologi al agreement between pseudo-resultative adje tives and the obje t. � Levinson (2007) produ es only a single example (reported here in 236 ) in whi h the verb does not seem to be an ICV as no impli it entity is (236) M' he reated by the a tion of lligat els Me-dat have-1st tied tying � . ordons de les the la es sabates (ben) of the shoes estrets. (very) tight-pl (Catalan) I tied the la es of my shoes very tight. A deeper study about Catalan and pseudo-resultatives should be du ted in order to investigate whether the above example a an be on- onsidered ase of PR. This dissertation does not investigate Catalan, however it studies the onstru tion in other two Roman e languages, namely Italian and Fren h. Before analyzing PR subgroup of parasyntheti ter. onstru tion in Italian, se tions 4.3 des ribes the denominal verbs that will be studied in this Furthermore, I will show that they behave as impli it reation verbs. Se tion 4.4 presents data whi h show the grammati ality of PR in Italian. Se tion 4.4.6 proposes a synta ti hap- onstru tion analysis for these verbs and PR in Italian. 4.2.5 Strong resultatives in Italian Adje tival se ondary predi ation in Roman e languages are usually not grammati al; Roman e languages belong to the lass of verb-frame languages (Talmy 1991, 2000) and do not allow an adje tive to introdu e a resultant state with an a tivity verb. resultative For this reason, the availability of pseudo- onstru tion in Italian (as we will see further on) is interesting, sin e it is on an adje tival resultative onstru tion. � In order to express motion dire tion , Italian resorts to verbal morphology, � Original example by Mateu (2000), reported as example (107) in Levinson (2007). � The knot is not reated by the a tion of tying, it only hanges in nature. � Even though new studies point out that this is not a dual typology, sin e there are mixed languages su h as Greek (Soroli & Hi kman 2011). ���� ���������� �������� �� ������������������� ���� 77 adje tives or prepositions are not allowed for this fun tion (Talmy 1991, 2000). Italian allows resultatives in asso iation with verbs that already entail the a hievement of a resultant state, a hievement resultatives in Folli's (2001) terminology, or weak resultatives in Washio (1997)'s terminology; and does not onstru t resultatives on verbs of a tivity, a tivity resultatives in Folli's terminology, or strong resultatives in Washio's terminology. (237) Giovanni ha martellato il metallo per/*in 5 minuti. G. hammered the metal for/*in 5 minutes. (238) *Giovanni ha martellato il metallo piatto. G. hammered the metal at. (239) *Giovanni ha martellato il metallo in bri iole. G. hammered the metal in rumbles. (240) Giovanni ha rotto il vaso *per/in 5 minuti. G. broke the vase *for/in 5 minutes. (241) Giovanni ha rotto il vaso in mille pezzi. G. broke the vase in one-thousand pie es. (242) *Giovanni ha rotto il vaso aperto. G. broke the vase open. Examples (238) and (242) show that an adje tival resultative predi ates in Italian produ e agrammati al senten es, both in asso iation with a tivity verbs su h as rompere break '). martellare to hummer ') (` and a hievement verbs su h as (` On the other hand, examples (240) and (241) do not show a parallel behavior. Italian resultatives onsist in a further spe i ation of the result proje tion, whi h is already present in the verbal aspe tual stru ture. other words, in (240), Giovanni resultant state of being broke. pezzi auses il vaso In to attend a new state, the in mille il vaso, that it is not only broken, In (241), the prepositional phrase further spe ies the state rea hed by but it is broken in pie es. Sin e adje tival se ondary predi ations on a tivity verbs are agrammatial in Italian, the fa t that the pseudo-resultative implies that the two onstru tion is a eptable onstru tions have dierent derivations. In fa t, we agree with Levinson (2007) that the resultative part is not introdu ed by the adje tive, but by relational elements. In Levinson's approa h two preposition-like elements mediate the relation between impli it entity and the adje tive. In my approa h, a non-eventive relational head (r head) and it is lexi alized by the prex. ������� �� ������������� ��������� ����� 78 4.3 Italian denominal parasyntheti s The aim of this se tion is to show that parasyntheti to the lass of impli it of parasinteti vebrs, please refer to They lass hapter 3. The group of denominal parasyntheti ausative semanti s. denominal verbs belong reation verbs. For a detailed presentation of the verbs analyzed in this hapter has an be paraphrased as (make) X be ome(s) an N. For the identi ation of BN verbs, I ondu ted a sear h in Devoto & Oli (2014) with a rst renement with automati tools. I identied 57 Italian verbs distributed among four dierent synta ti ongurations: transitive, alternating transitive intransitive, pronominal intransitive, reexive. The distribution of synta ti in table 4.1 (page 78) and an example for ea h ongurations is reported ase is proposed in senten es below. The present ian. It is hapter investigates pseudo-resultative onstru tion in Ital- onstru ted on transitive verbs, therefore only the transitive sub- group of BNs is analysed in this se tion. (243) L'orafo ha inlzato le perle. (Transitive) Goldsmith pier ed (and threaded together) pearls. (244) a Medusa impietriva hiunque la guardasse. (Trans.-Intrans.) Medusa petried whoever looked at her. b Nella lotta alle di oltà, l'animo impietriva. In the ght against di ulties, the spirit hardened. (245) Il lo si è aggrovigliato. (Pron. Intrans.) The line tangled. (246) I ragazzi si sono a oppiati per l'eser izio. (Reexive) Students paired for the exer ise. Stru ture Transitives % on the total 65,45 Transitives and intransitives Pronominal intransitives 5,45 10,91 Reexives Figure 4.1: Distribution of BNs among synta ti 9,09 patterns. The following se tion reports eviden e in favor of the analysis of BNs as impli it reation verbs. ���� ������������������� �� ������� 4.3.1 Italian BNs orrespond to impli it We have seen that pseudo-resultative verbs, whi h denoted the 79 reation verbs onstru tion involves impli it reation reation ( oming into existen e) of a new entity represented by the nominal base of the verb. Two riteria are used, namely those reported in se tion 4.2.1 for English verbs, to show that Italian BNs belong to the impli it reation verb lass (ICV). 1. Italian BNs are goal verbs (Clark & Clark 1979) and they imply the reation of a shadow argument (Geuder 2000). A ordingly to Clark & Clark (1979: 774), the subje t  to ause it to ome about that the obje t is base N-ed . (247) Giovanni s-bri iol-a G. s- rumble-3. sg. il does something pane. the bread. G. rumbles the bread. (248) Daria a - atast-a i D. the books. sg. a-sta k-3. D. heaps books. libri. In (247), ` rumble', and (248), `heap', onstitute a shadow argument. 2. BNs require an ae ted dire t obje t. Example (249) shows that dire t obje t is mandatory and it expresses an ae ted argument, sin e it denotes the individual whi h is moved and reorganized in the spa e in order to a reate sta k. (249) Daria am-mu D. hi-a *(i sg. a-sta k-3. *(the vestiti) lothes) D. sta ks lothes. Italian BNs are impli it Now that the impli it reation verbs (hen eforth ICV). reation nature of ausative denominal parasyn- theti verbs has been demonstrated, these verbs they an o ur in pseudo-resultative an be used to test whether onstru tion in Italian as they do in English ( f. se tion 4.4). 4.4 Pseudo-resultatives in Italian We have seen in hapter 1 that the investigation of grammati ality is not always a simple matter. Pseudo-resultatives in Italian are hallenging in this ������� �� ������������� ��������� ����� 80 respe t. In fa t, the adje tive involved in this onstru tion an re eive two interpretations: internal obje t modier (250) or pseudo-resultative (251). (250) Giovanni sbri iola i G. bis otti sottili. rumbles the bis uits thin. John rumbles thin bis uits. ( int. Bis uits are thin before rumbling) (251) Se Giovanni mangia i If G. eats bis otti, li the bis uits, l. 3.m.pl. sbri iola sottili. rumbles thin. If John eats bis uits, he rumbles them thin. ( int. Bis uits turn into thin rumbles) An Italian speaker adje tive, in this an always get a grammati al interpretation for the ontext, and this makes di ult to investigate if senten es as the one above are well-formed or not in the PR interpretation. The desired interpretation must be made expli it in some way. For this reason, a semanti interpretation task has been designed and performed as reported in se tion 4.4.1. In the pre-test phase, I tested 4 Italian native speakers about the grammati ality of pseudo-resultatives. The judgments diered a lot and seemed to be related to Italian regional varieties. Consequently, in the experimental phase, two resear h questions were targeted, namely: a. Is pseudo-resultative onstru tion grammati al in Italian? b. Do dierent varieties of Italian present signi ant dieren es in a ept- ability of pseudo-resultatives? Results Italian with onrm the grammati ality of pseudo-resultative ausative denominal pararyntheti rate of 83,58 %), furthermore a obje t is pronominal (a verbs (a onstru tion in eptability mean eptability rate in reases when the dire t eptability mean rate of 99,5%) � onrming the ad- je tival nature of the predi ation . No signi ant dieren e in a This suggests that PR a eptability is found between Italian varieties. eptability is not related to diatopi variation in Italian. � This argues in favor of an AP analysis of the adje tive, as assumed in Levinson (2007: 72) ���� ������������������� �� ������� 81 4.4.1 Methodology The present se tion des ribes design and methodology employed for the semanti interpretation task. Experimental senten es an re eive two interpretations: one in whi h the adje tive is interpreted as modier of the dire t obje t (i senten es); one in whi h it is interpreted as pseudo-resultative (ii senten es). (252) Giovanni ha sbri iolato i bis otti ni. John rumbled bis uits thin. (i) (ii) (253) John made rumbles from thin bis uits. John made thin rumbles. Giovanni ha impilato i libri alti. John piled books high. (i) (ii) John made a pile from high books. John made a high pile. Therefore, the task must be designed in su h a way as to allow a ess to � both interpretations . This experiment is divided into three parts: (i) a so io-linguisti tionnaire, (ii) a warm-up phase with instru tions, (iii) the linguisti ques- part. Ea h parti ipant is tested on 11 experimental senten es with an ICV and 11 llers built on a non-parasyntheti Ea h senten e has two denominal ausative verb. onditions: expli it dire t obje t; pronominal di- re t obje t. Senten es and llers are presented to informants in random order. In no ase, a parti ipant is asked to judge both onditions of the same senten e. Parti ipants are asked to tations, whi h are: hoose one or both of the proposed interpre- (i) adje tive modies dire t obje t (252); (ii) adje tive modies the impli it entity (253). Interpretations are made expli it by means of paraphrases presented always in the same order: adje tive as obje t modier in rst position and � adje tive as pseudo-resultative in se ond position (gure 4.2 , page 82). This is true also for llers, for whi h only one interpretation is possible, namely the one in whi h the adje tive modies the dire t obje t. � Parti ipants are allowed to sele t both. � Figure ontents translation: When hildren play, they pile building-blo ks rooked. From building blo ks, they reate rooked olumns. From rooked building blo ks, they reate piles. ������� �� ������������� ��������� ����� 82 Figure 4.2: S reen-shot of a task (Semanti de ision task ITA). Experiment was administered via IbexFarm. present the same amount of The system managed to onditions for ea h senten e. 4.4.2 Parti ipants 106 Italian native speakers ompleted the experiment (73 female): 38 speak- ers of Northern regional Italian, 35 of Southern regional Italian, 33 of Central regional Italian, table 4.1 (page 82). North regional Italian varieties are spoken North to the isogloss Rimini-La Spezia (Pellegrini 1977). It separates northern diale ts from entral. South regional varieties are spoken South to the isogloss An ona-Roma (Pellegrini 1977). This isogloss separates entral diale ts from the southern one. Central regional-Italian varieties of Italian are in luded between the two mentioned isoglosses. Informants are divided in three groups on the basis of two position to a diale t during hildhood; and if none diale t exposition during hildhood was de lared in the so iolinguisti linguisti riteria: ex- questionnaire, pla e of birth and ba kground of parents. Male Female Total North 13 25 38 South 9 26 35 Center 11 22 Total 33 73 33 106 Table 4.1: Parti ipants gender and origin (Semanti de ision task ITA). Edu ation rate of the sample divides as follows: 10,38% of parti ipants have a high-s hool diploma, 52,83% have a university degree, 36,79% have a PhD. The three regional groups present omparable edu ation, in parti ular, speakers without a degree are less than 15% in ea h group (table 4.2, page 83). ���� ������������������� �� ������� 83 Age distribution among the three groups is less homogeneous than eduation, although the majority of speakers in ea h group is aged less than 40 (table 4.3, page 83). High-S hool Degree PhD North 13,16 60,53 26,32 South 8,57 45,71 45,71 Center 9,09 51,52 39,39 GLOBAL 10,38 52,83 36,79 Table 4.2: Parti ipants edu ation (Semanti de ision task ITA). 18-25 26-32 33-40 41-60 60+ North 5,26 47,37 13,16 31,58 2,63 South 22,86 40,00 25,71 8,57 0 Center 9,09 57,58 21,21 6,06 3,03 GLOBAL 12,26 48,11 19,81 16,04 1,89 Table 4.3: Parti ipants age groups (Semanti de ision task ITA). 4.4.3 Results for ondition 1 This subse tion reports the results obtained in the rst experimental tion, namely the one (254) ondi- ontaining an expli it dire t obje t. Se non erano esperte nella latura, le donne aggomitolavano il otone las o. If they were not ning experts, women winded loose otton. Results do not show any signi ant dieren e in answers for the three linguisti varieties, as graph 4.3 (page 84) shows, where OD stands for obje t modifying adje tive interpretation, PR for pseudo-resultative interpretation and OD PR for both interpretations. Answer rate is perfe tly similar for the three groups, and no signi ant dieren e is found. As pseudo-resultative a eptability rate obtained by the sum of PR and OD PR is more than 85% for ea h group, I assume that pseudo-resultative onstru tion is a � eptable in Italian . � To my knowledge, there are no studies about the su ient a makes a onstru tion grammati al. eptability rate that ������� �� ������������� ��������� ����� 84 ��� �� �� �� �� �� �� ����� ��� �� �� �� �� �� � ����� ������ ����� ������� Figure 4.3: Condition 1. Means of answers for speaker group (Semanti de ision task ITA). Furthermore, no dieren e is found in single experimental item results, all items re eive similar answers, as reported in graph 4.4 (page 85). I argue that PR onstru tion is a eptable in Italian with no observable dieren es between Italian regional varieties. However, the a (namely 85 %) shows that the is not fully a 15 % of eptability rate onstru tion, even though mainly grammati al, epted by Italian native speakers. ases in whi h it is not In fa t, there is a mean of hosen as possible interpretation. We will see in the next se tion that in the se ond experimental the one with pronominal obje t, the a onstru tion in reases up to 99%. This ondition, eptability rate of pseudo-resultative onrms the grammati ality of PR in Italian and shows its preferen e in one synta ti ontext. I will a ount for this behavior in se tion 4.4.6. 4.4.4 Results for ondition 2 This se tion reports results obtained by the se ond experimental whi h ontains a pronominal dire t obje t, as in (255). ondition ���� ������������������� �� ������� 85 ��� �� �� �� �� �� ��� �� �� �� �� ����� �� �� � ����� Figure 4.4: Condition 1. Means of answers for item (Semanti de ision task ITA) (255) Quando prepara When il salame di (she)-prepares the salami of bis otti, Maria li bis uits, Marie a .m.pl. io olata ho olate sbri iola ni. on with i det.m.pl m.pl. rumbles thin- When Mary prepares the ake with bis uits, she rumbles them thin. Sin e the analysis of the rst ondition has shown the absen e of signif- i ant dieren es between dierent regional varieties, data are analyzed as a unique group. Results show that pseudo-resultative interpretation of adje tive is not only strongly preferred when the obje t is pronominal, but it appears to be the only possible one. The global rate of a is in fa t of 99,68% rate eptability of PR interpretation omposed of 98.53 of PR alone and of 1.15 of OD PR (graph 4.5, page 86). When the dire t obje t is ally signi ant higher a liti ized, PR interpretation re eives a statisti- eptability rate (PRCond1 < PRCond2 onrmed by ������� �� ������������� ��������� ����� 86 ��� �� �� �� �� �� �� �� ��� �� �� ��� �� �� �� � ����� Figure 4.5: Condition 2. Means of answers for item (Semanti de ision task ITA). a t(10)=4,2691) than when the dire t obje t is an expli it DP. In se tion 4.4.6, I dis uss the results obtained in both experimental ditions. In parti ular, I argue that they hypothesis for pseudo-resultative eptability between the two on- onrm Levinson's (2007) derivation onstru tion and that the dieren e in a - onditions is determined by the presen e or the absen e of lexi al material in the dire t obje t position. 4.4.5 Some improvements to the methodology The methodology used for the semanti interpretation task was designed by myself and it was the rst time it was employed. For this reason, a margin of improvement is possible and riti al aspe ts must be pointed out for possible future appli ations of the same task. ���� ������������������� �� ������� 87 Sin e it was used at the same time for Fren h too ( hapter 5), this se tion analyses riti al points of both appli ations, for Italian and Fren h. It is worth noting that the so iolinguisti instru tions were questionnaire, examples and lear, the task was understood, no big variability in parti - ipants' answers was registered. However, some i senten es. riti isms were pointed out about the plausibility of spe- Parti ularly regarding the lexi al plausibility of involved, su h as the pair understand. by eno rotondo on epts (`round hay'), whi h is di ult to The validity of the task is not ompromised, as pointed out omparable results for ea h senten e. All experimental items need to be he ked for senten e plausibility before the experiment in order to less plausible senten es, sin e they mati al or semanti judgments. orre t an generate some interferen es in gram- For both Italian and Fren h, experimental senten es were tested for plausibility by four native speakers who did not further undergo the task. As ller senten es were built with the same morphologi al elements of experimental items (ex eption made for the verb whi h was not an ICV), this implied the impossibility of reating senten es with a double reading as ex- perimental senten es were. This generated an asymmetry in possible answers between experimental and ller senten es and onsequently it ould ause the identi ation by parti ipants of experimental items, and the re ognition of investigated stru ture. However, the use of ller senten es with two readings would have led to the use of ompletely dierent morphologi al elements. This solution would also have led to the identi ation of experimental items and of the stru ture under investigation. To summarize, the experimental design employed for investigating the a eptability of pseudo-resultative onstru tion in Italian and Fren h is per- fe tible regarding senten e lexi al plausibility. Furthermore, the onstru tion of ller senten es was di ult: (i) if llers were built similarly to target senten es, they involved one single reading; or (ii) if ller senten es were built on ompletely dierent stru tures but with two readings , they an alternative reading as target senten es. In both parti ipants to identify ller senten es and ould generate ases there was a risk for onsequently experimental target senten es. 4.4.6 Dis ussion In se tions above, we have seen that Italian ausal denominal parasyntheti verbs of the type `make X be ome N' are impli it reation verbs, in terms of Levinson (2007). In other words, they involve an impli it entity whi h is stru turally the base noun of the verb and whi h is reated by the a tion ������� �� ������������� ��������� ����� 88 des ribed by the verb itself. We have seen that Italian a epts PR onstru tion built on ICV verbs. PR onstru tions are mu h more preferred when the dire t obje t is a pronoun rather than a referential DP. With a pronoun, speakers identify the senten e as PR in 99% of ases and with a nominal the hoi e drops to 85 %. This se tion will produ e an analysis for Italian PR whi h agrees with Levinson's (2007) analysis for English PR. I propose that the Levinson's TO impli it relational element 4.4.6.1 orresponds to the r head. Pseudo-resultative derivation This se tion applies Levinson's (2007) pseudo-resultative approa h to Italian pseudo-resultative onstru tion. We will see that the pseudo-resultative adje tive modies the impli it entity. However, the impli it entity, being a root, annot he k the adje tive's φ-features. Consequently, the adje tive must he k φ-features with a ategorized upper element, namely, the rst un he ked its un he ked - ommanding DP, the dire t obje t. Chapter 2 reports two main hypotheses about word formation. A ording to the rst approa h, the pro ess has a double nature and involves lexi al and synta ti onstraints. A ording to the se ond approa h, the pro ess of word formation is synta ti in nature (Marantz 2000), are built obeying the same onstraints as senten es. onsequently, words Lexi on provides bare roots to syntax and the syntax is responsible for their means of spe i fun tional ategorization, by ategorizer heads. In order to understand whether a word is a bare-root or a element, dierent tests are disallowed on the root on e it merges with a morphology is allowed only on Italian ateogorizer; derivational ategorized roots. ausative denominal parasyntheti verbs allow modi ation by an adjun t whi h further spe ies the impli it entity. that the verbal base, whi h denotes entities su h as `heap', (256) Examples below show pila, `pile', or an be spe ied by adjun ts. Daria intre ia i suoi apelli. Daria braids her hair. (257) # Daria intre ia i suoi apelli Daria braids her hair in a braid. (258) Daria intre ategorized an be performed: further modi ations or operations ia i suoi apelli 3.sg. det. Daria in-braid- in una tre in un'a pl. her hair- ia. on iatura. in det. hairdo. ammasso, ���� ������������������� �� ������� 89 *Daria makes a ponytail out of her hair. (259) Sandro ha allineato le tessere del domino lungo una ir onferenza. Sandro lined up domino pie es in a ir umferen e. (260) Daria ammassò le immondizie in un mu Daria sta ked rubbish in a shapeless heap. (261) Piero ha impilato i dis hetti Piero piled disks in heaps. (262) [...℄ in mu si sono spinti giù per la rampa e mu hio le operte hio informe. hi. hanno ammassato in un � he fanno da letto ai nuovi ospiti. They run down the ramp and they sta ked in a heap blankets used as beds by new guests. (263) Oggi appaiono separati uno dall'altro non solo per le su essive ero- sioni operate sulla dorsale dal Torrente Cormor, ma an he per l'azione di due sistemi oniugati di faglie verti ali he in tempi re enti spezzato in segmenti la dorsale spostandone �� hanno leggermente le singole porzioni. Today they appear separated not only for repeated erosions on the Cormor river's edge, but also for the a tion of two onjugated vertial fault systems whi h, re ently, broke the ridge in segments. A possible obje tion to the fa t that these verbs are root derived omes from the fa t that some of them seem derived from nouns that show expli it derivational morphology, as in the example below. (264) Maria ha spezzettato perf. Maria has small-pie e-break- la io olata. the ho olate-bar. Maria ra ked the ho olate bar. ase of spezzettare and aaldellare, whi pezzetto, `a small pie e', and faldella, `a small layer', whi derived from pezzo, `pie e', and falda, `layer'. This seems to be the from Derivational morphology is spe i that it an atta h only to for ea h synta ti h are derived h in turn are ategory, this means ategorized element and not to bare-roots, in any further derivation takes as its input not the root itself, but an element whose semanti and phonologi al properties have been ashed out  (Arad fa t  2003: 2). However, ICVs built from derived nouns are few and the � http://ri er a.repubbli a.it/repubbli a/ar hivio/repubbli a/2010/08/30/nei-box-sotterranei-hote 18/10/2016. �� http://www.geos 18/10/2016. reation ienze.units.it/geositi/vedigeo1.php?ID_GEO=221, ������� �� ������������� ��������� ����� 90 of new parasyntheti s from morphologi ally derived bases is impossible, as shown in examples below. (265) gomitolo - aggomitolare - gomitolino - *aggomitolinare ball of wool - wind - small ball of wool - make a small ball of wool (266) pila - impilare - piletta - *impilettare pile - pile - small pile - make a small pile (267) fetta - aettare - fettina - *aettinare sli e - sli e - small sli e - make a small sli e To summarize, only few ICVs present what looks like a derived base-noun and it is not possible to reate them by means of morphologi ally omplex nominals. For this reason, I argue that pseudo-derived parasyntheti verbs are in fa t registered in the lexi on, they are not synta ti ally derived but they enter the syntax already provided of the pseudo-derivational part, they are root in nature, and in fa t they annot parti ipate in PR onstru tions. We have seen that Roman e languages are important in the pseudoresultative debate be ause of their morphologi ally overt agreement on adje tives. They show that PR adje tive does not agree with the base element but with the dire t obje t. Sin e the base element is dire tly modied by the PR adje tive, it would be able to he k for the un he ked φ-features of the adje tive if it were a nominal, yielding an expli it morphologi al agreement. However, this is not the ase as Italian examples have shown above. The adje tive agrees with the dire t obje t. Thus, the dire t obje t is the rst available nominal expression against whi h the adje tive may (268) he k its Carla sbri iola i bis otti ni. Carla rumbles bis uits thin. φ-features, as shown in (269). ���� ������������������� �� ������� 91 V oice auserP (269) DP Carla vbe omeP V oice auser rP vbe ome r' DP bis otti [F℄ r=INTO a-, in-, s- P √ √ bri iola In (269), the adje tive egorised root its bri iola n- (`thin') merges as a AP n- [uF℄ omplement of the un at- (` rumble'), without having the possibility of he king φ-features, so it nds a andidate in the rst - ommanding full DP, whi h bis otti (`bis uits'). √ The un ategorized root bri iola merges in P and moves upwards to is the dire t obje t vbe ome . The relational head r assures the right semanti relationship between the dire t obje t and the impli it entity. Its semanti s is responsible for the introdu tion of a lo ative relation and stru ture is interpreted to onal relation. In other words, the ontain the impli it entity resulting in the same lo ation of the dire t obje t and been made of the same material. The impli it entity, by means of head-head movements moves upwards and builds its semanti s. vbe ome . It is ategorized in vP by means of the head With respe t to Levinson's (2007) original analysis, I argue that the verbalizing head is responsible for introdu ing an energeti & Harley 2015). The for e (Copley ausative meaning arises from the presen e of a rP in the lower part of the derivation (Hoekstra 1988; Shäfer 2008). The Voi e head introdu es the Causer external argument and it assures it is interpreted as the individual responsible for the introdu tion of an energeti for e in the situation. 4.4.6.2 Pronominal obje t simplies PR interpretation We have seen that the root nature of the base explains why Italian pseudoresultative adje tives agree in number and gender with the dire t obje t, ������� �� ������������� ��������� ����� 92 generating the ambiguous reading between a PR interpretation and a standard DP modi ation that does not involve a result. of the semanti However, the results interpretation task have shown that the pseudo-resultative reading of the adje tive is denitely easier to obtain when the internal obje t is pronominalized. The pseudo-resultative reading in the previous rea hes 99% of ase hoi es, making it the only possible interpretation. The fa t that the adje tive, when interpreted as modier of the dire t obje t, is within a determiner phrase determines its attributive nature and furthermore prevents the pronominalization of the noun. (270) Giovanni impila i librii altii . G. piles the high books. (271) DP D NP NP AdjP libri adj alti Thus, if the dire t obje t of senten es whi h involve a PR onstru tion is pronominalized, the attributive interpretation of the adje tive disappears and the only pseudo-resultative interpretation is possible. 4.4.7 To sum up The results of a semanti interpretation task ondu ted on 106 Italian native speakers of three dierent regional varieties revealed that Italian speakers allows PR onstru tions. The a eptability rate is more than 85% in full dire t obje t and it raises up to more than 99% in ase of pronominal dire t obje t (with no signi ant dieren es in the three diatopi The rise of PR a eptability in presen e of a liti onrm the a Results in presen e of a eptability of the PR varieties). dire t obje t is a - ounted for by the attributive nature of adje tive whose subje t dergo pronominalization. ase of annot un- liti ized dire t obje t onstru tion in Italian. The root nature of the base element has been demonstrated by means of semanti tests and by means of synta ti eviden e from Roman e adje tival agreement. Levinson's (2007) analysis for pseudo-resultative onstru tion has been extended to Italian with some dieren es, namely the r head takes pla e of the impli it relational elements IN and TO and it is lexi alized by prexes. ���� ������� ��� ��������� 93 Se tion 4.5 investigates a registered dieren e in a PR adje tives and orresponding adverbs. eptability between I will argue that the split be- tween semanti s and syntax in PR adje tives is responsible for their lower a eptability, as opposed to synonymous adverbs. 4.5 Adverbs are preferred During the semanti interpretation task many speakers have informally noted that, even though they a ept the pseudo-resultative prefer, when possible, the (272) onstru tion (272) they orresponding adverbs (273). Quando Giovanni fa la torta on i bis otti, li sbri iola ni nemente . When John makes the bis uit ake, he rumbles them thin. (273) Quando Giovanni fa la torta on i bis otti, li sbri iola . When John makes the bis uit ake, he rumbles them thinly. This intuition has been task (Bard et al. 1996) onrmed by means of a magnitude estimation ondu ted on 15 native speakers. This se tion re- ports its results whi h show that adverbs are preferred to pseudo-resultative adje tives. It provides an analysis that explains this preferen e, arguing that adverbs an naturally take two s opes with result verbs, one of whi h is a low s ope modifying the resultative part (rP). 4.5.1 Methodology The theoreti assumption of ME onsists in the fa t that grammati ality is not a binary on ept, rather a gradient from non-grammati al to ompletely grammati al. ME design is parti ularly useful for the analysis of PRs adje tives and adverbs be ause both typi al a an modify the predi ate in Italian, therefore with a eptability judgment questionnaire it would be di ult to dis rim- inate over the degree of a modi ation. ME is tween them, eptability of pseudo-resultative and adverbial apable of investigating the dieren e in judgment be- onsequently it an predi t whi h one is preferred by speakers ( hapter 1). Sin e the previous semanti de ision task points out that the pronominal obje t blo ks the attributive interpretation of the adje tive and only allows the PR interpretation, experimental and ller senten es involve a pronominal dire t obje t. Parti ipants are asked to evaluate senten es and to attribute them a numeri al value. Numeri al values provided by speakers must be proportional, ������� �� ������������� ��������� ����� 94 onsequently they are distributed on a s ale whi h is not dened by the experimenter. Ea h informant establishes her own s ale by the evaluation of a referen e senten e ( alled also modulus). Values given to experimental senten es are asked to be proportional to the one of the modulus senten e. This way, values distribute on a personal s ale where the interval between values remains onstant, ontrary to normal a eptability questionnaires where the s ale is xed by the experimenter and the value given to intervals an vary form speaker to speaker. The test is administered via Ibexfarm and so io-linguisti (iii) linguisti omposes of three parts: (i) questionnaire; (ii) warm-up phase (judgment of lines length); test (judgment of senten es a The rst part is a standard so io-linguisti eptability). questionnaire asking for infor- mation about age, edu ation, spoken languages and diale ts. The se ond part onsists of a warm-up phase, where parti ipants must judge a series of lines length. First, parti ipants must evaluate the length of �� a referen e line, giving it a personal appropriate value (gure 4.6 , page 96). Se ond, they are asked to evaluate the length of other lines, proportionally to �� the length value they have assigned to the referen e line (gure 4.7 ), page 95). Figure 4.6: Warm-up phase. Referen e line, s reen-shot (Magnitude Estimation). The third part of the task onsists of the linguisti task. tions, parti ipants are asked to give an appropriate value of a �� S reen-shot translation: Line of referen e: What is its length? �� S reen-shot translation: Line 1: With respe t to referen e line, what is its length? After instru eptability to a ���� ������� ��� ��������� 95 Figure 4.7: Warm-up phase. Example of a line lenth judgement, s reen-shot (Magnitude Estimation). �� referen e senten e (gure 4.8 , page 96). Parti ipants must judge the gram- mati ality of the referen e senten e on the base of their personal opinion, and referring to a medium- ontrolled ommuni ative situation. In subsequent frames, parti ipants must evaluate experimental senten es proportionally to the value attributed to the referen e senten e. Experimental senten es �� ontain a pseudo-resultative adje tive (gure 4.9, page 96 �� an adverb (gure 4.10 Parti ipants judge one senten es for both ondition per senten e and an equal number of onditions. items, built on parasyntheti The task is omposed of 11 experimental ICV verbs, and 11 llers, randomly presented. Senten es below are example of experimental items (274) is a ase of pseudo-resultative onditions. Quando Mario ha When M. ammonti sg. molte ban onote, le pl., a -pl.f. many banknote- hia verti ali. 3.sg. mount- have-3 verti al- Senten e onstru tion. Senten e (275) is a adverbial modier. (274) ) or , page 97). pl.f. �� S reen-shot translation: When John observes the olor of apples, he them intensely sees red. Referen e senten e: Give a value to this senten e. �� S reen-shot translation: Before eating ho olate, Mario rumbles it thin. With respe t to the referen e senten e, how do you judge this senten e? �� S reen-shot translation: When Mario moves do uments, he sta k them haoti ally. With respe t to the referen e senten e, how do you judge this senten e? ase of ������� �� ������������� ��������� ����� 96 Figure 4.8: Referen e senten e, s reen-shot (Magnitude Estimation). Figure 4.9: Example of experimental senten e judgement (PR), s reen-shot (Magnitude Estimation). When Mario has many banknotes, (he) piles them verti al. (275) Quando Mario ha When M. ammonti sg. le pl., a -pl.f. many banknote- hia verti almente. 3.sg. mount- have-3 molte ban onote, verti al- adv When Mario has many banknotes, (he) piles them verti ally. The onstru tion of experimental items must pay attention to the of lexi al material. In fa t, not all adje tives whi h PR onstru tion present a (276) an be employed in the orresponding adverb. Quando Daria riordina When hoi e i libri, Daria riarranges the books, li a .3m.pl. ammassa alti. sta ks high. ���� ������� ��� ��������� Figure 4.10: 97 Example of experimental senten e judgement (ADV), s reen- shot (Magnitude Estimation). When Daria riarranges the books, she sta ks them high. (277) # Quando Daria When Daria ammassa high- adv. riordina i riarranges the altamente libri, books, a .3m.pl. li sta ks . When Daria riarranges the books, she amasses them highly. In examples above, the adverb rived from the adje tive alto, altamente, `highly', morphologi ally de- `high', generates a semanti ally odd senten e. This is due to the fa t that it does not involve the meaning of the meaning of thorough. height, rather I assume that this asymmetry is due to idiosyn rati lexi al gaps, in other words, some adverbs are built on se ondary meanings of adje tives. To avoid semanti and adverbs has been oddity, an a urate sele tion of adje tives ondu ted in the preparatory phase. 4.5.2 Parti ipants 15 native Italian speakers (9 female) parti ipate in this experiment, dierent so ial aspe ts are reported in table 4.4 (page 98). 4.5.3 Results Ea h parti ipant evaluates senten es on the base of her personal s ale, determined by the value she assigns to the modulus senten e. This implies that results obtained by dierent parti ipants are not immediately sin e they are based on dierent s ales. omparable, For this reason, answers of ea h ������� �� ������������� ��������� ����� 98 Edu ation (%) Age (%) master degree PhD 93,33 18-25 13,33 6,67 26-32 53,33 33-40 20,00 41-60 13,33 Table 4.4: Informants' age and edu ation (Magnitude Estimation). informant are normalized on the base of the value she assigned to modulus senten e. Results onrm the experimental hypothesis: when the modi ation is adverbial, the senten e is judged with higher values gure 4.11, page 98 t (24) = 0.2926, p > 0, 05 t2 (20) = 0.0018, p > 0, 05) by ( 1 ea h subje t gure 4.12, page 99. ��� � ��� ��������� ������ ����� � ��� � ����� Figure 4.11: Graph of the a nitude Estimation). eptability rate for ea h experimental item (Mag- ���� ������� ��� ��������� 99 � � � ��������� ������ ����� � � � � ������������ Figure 4.12: Graph of the a eptability rate for ea h subje t (Magnitude Estimation). Results of this test point out that in the synta ti of impli it ontext of modi ation reation verbs, adverbs are preferred to adje tives. In the next se tion, the role of adverbs is analyzed. 4.5.4 Dis ussion The ME shows that in the same synta ti ontext, adverbs are preferred to adje tives in PRs. Moreover, adverbials ICVs su h as (278) an entertain two readings when impilare. Maria a atasta i libri aoti amente. Mary sta ks books haoti ally a. Chaoti manners. b. Chaoti sta k. ombined with ������� �� ������������� ��������� ����� 100 As usual with resultative verbs, the adverb an entertain two readings in relation to the part of the derivation it modies. It part (rP), books, books in a pile, an modify the resulative obtaining narrow s ope; or the v proje tion, pile obtaining wide s ope. In the ase of (278), the adverb an interpreted as modifying the rP, having narrow s ope. (279) ... a atasta i libri aoti amente. ... sta ks books haoti ally. ( int. the sta k is haoti ) vbe (280) vbe ome ' ome rP rP AdvP aoti amente DP r i libri a- √ atasta In the same ontext the adverb the adverb predi ates of event of (281) ... a atasta i libri an re eive a wide s ope. sta king aoti amente. In this ase, (282). ... sta ks books haoti ally. ( int. the event of sta king is haoti ) ���� ������� ��� ��������� 101 vbe (282) vbe vbe ome P ome P ome AdvP rP aoti amente DP r i libri a- √ atasta To summarize, the two interpretations that the adverb pseudo-resultative an re eive with onstru tion are generated by a dierent s ope. Resulta- tive adverb interpretation arises be ause the adverb has narrow s ope and it is interpreted as modier of the r proje tion. Manner adverb interpretation arises be ause the adverb has wide s ope and it is interpreted as modier of the vP. Now that the properties of adverbs are laried, I will argue for the higher preferen e of speakers to use the adverb rather than the adje tive in PR onstru tion. (283) Quando Daria ompra le s arpe nuove, le ammassa aoti he. When Daria buys new shoes, she sta ks them haoti . (284) Quando Daria ompra le s arpe nuove, le ammassa aoti amente. When Daria buys new shoes, she sta ks them haoti ally. We have seen that PR adje tive modies the impli it entity but it synta ti ally agrees with the dire t obje t, produ ing a split between syntax and semanti s. On the other hand, the adverb does not produ e in any ase a split between syntax and semanti s. As usual with resultative verbs, the adverb is read either as a modier of the result proje tion or as a modier of the v P. I argue that this higher transparen y between syntax and semanti s of adverbs determines their higher a than the orresponding adje tives. eptability in relation to PR onstru tion ������� �� ������������� ��������� ����� 102 4.6 This Con lusions hapter analyses the properties and the grammati ality of the pseudo- resultative onstru tion (Levinson 2007) in Italian. Several experiments demonstrate the grammati ality of PR. However, results open some questions about the higher rate of PR a eptability or- related with the presen e of a pronominal dire t obje t, and on the other hand, the higher preferen e of adverbs over their synonymous adje tives. Dieren e in a eptability rate for Italian PR is due to two readings gen- erated by the adje tive: as pseudo-resultative or as the internal obje t modier. The grammati ality rate of PR with senten es with pronominal dire t obje t depends on the impossibility by an adje tive to modify a pronominal DP, leaving only one reading for the adje tive, that of a predi ate of impli it entity. I argue that adverbs are preferred to synonymous adje tives in PR on- stru tions be ause the former do not generate a split between syntax and semanti s. In parti ular, this is possible in relation to the nature of the verb whi h is resultative, and it allows an adverbial modi ation either with narrow s ope, modifying the Pr, or with wide s ope, modifying the v P. Chapter 5 Pseudo-resultatives in Fren h 5.1 This Introdu tion hapter analyses Fren h pseudo-resultative interpretation task reveals that, less a 5.2 onstru tions. ontrary to Italian, this A semanti onstru tion is mu h epted in Fren h. Fren h pseudo-resultatives Fren h is a Roman e language where parasynthesis is a produ tive verbformation pro ess. (285) Jean a amon elé es aaires sur le bureau. John has sta ked his belongings on the table. (286) Jean a empilé les oreillers. John has piled the pillows. (287) Jean a émietté le pain. John has rumbled the bread. Senten es above express ussed in ausative events, as the Italian senten es dis- hapter 4. In other words, taking (285) as example, the event be paraphrased as Jean did something to sta k on the table. This interpretation is whi h the ause his stu to be arranged in a learer in the following examples in ausation is made expli it with verbs disposer `arrange' and `redu e'. (288) Jean a disposé es aaires dans un mon eau sur le bureau. John arranged his belongings in a sta k on the desk. 103 an faire ������� �� ������������������� �� ������ 104 (289) Jean a disposé les oreillers dans un pile. John arranged the pillows in a pile. (290) Jean a fait le pain en miettes. John redu ed the bread in rumbles. In this regard, Italian and Fren h ICVs seem to be perfe tly alike. However, the question about the a eptability of Fren h PR needs further explo- ration. In hapter 4, I have shown that Italian pseudo-resultative reading of ad- je tives is the only available reading when the dire t obje t is pronominal. We an assume that if Fren h native speakers do not a ept PR in this on- text (291), pseudo-resultatives are not grammati al in Fren h. (291) Quand Jean essaye de ranger When John tries il to organize his belongings, he ammon elle hautes sur le a-sta k- high 3.sg. ses aaires, on les l- obj.pl. bureau. the table. When John tries to organize his belongings, he sta ks them on the table. The question is interesting be ause Italian appears to be parti ular in the treatment of se ondary predi ates (adje tival resultatives and depi tives). The eld of se ondary predi ation is pretty large and in ludes strong resultatives (292) and weak resultatives (293), depi tives (294), small lauses (295). (292) = Sandra ki Sandra ki ked the door open. ked the door and as result of this a tion the door is open. (293) Giulia ha rotto il tavolo in pezzi. Giulia broke the table in pie es. (294) a. Sandro ha guidato la ma hina ubria o. Sandro drove the ar drunk. b. Sandro ha mangiato la arne ruda. Sandro ate the meat raw. (295) Maria onsidera Carla una buona ami a. Maria onsiders Carla a good friend. It is well-known that Roman e languages disallow strong resultatives (Talmy 1991, 2000; Washio 1997; Folli 2001). In this regard, Italian and Fren h behave in the same way, ex eption made for strong resultatives in whi h the predi ate is dupli ated or modied by an adverb (Folli 2002) (297). ���� ������ ������������������� (296) 105 a. *Maria martellò il metallo piatto. (Italian) b. *Marie martela le metal plat. (Fren h) Mary hammered the metal at. (297) Giovanni ha piallato il tavolo sottilissimo. (Italian) John planed the table ultra-thin. (298) ??Jean a raboté la table bien ne. (Fren h) John planed the table ultra-thin. Even though Italian and Fren h belong to the same linguisti family, their behavior with respe t to se ondary predi ation are not always the same. For this reason, I ondu ted a study of semanti in order to investigate the a de ision task for Fren h speakers eptability of PR. In the following se tion, I will present design and pro ess of experimental item onstru tion. 5.2.1 Semanti de ision task This se tion reports methodology and results of a semanti de ision task ondu ted on Fren h native speakers with the aim to investigate the a ept- ability of PR. The task was administered via pen il and paper. The experiment was omposed of three parts: (i) presentation of the resear her and the so iolinguisti questionnaire; (ii) sample senten e; (iii) linguisti task. The presentation of the resear her briey des ribes her as a student enrolled in a PhD program in Linguisti s at Université Paris 8 and does not mention the purpose of the resear h in order to avoid possible bias of resear h expe tan y respe t ( hapter 1). The so io-linguisti questionnaire asked for age, edu ation level, origin, residen e, spoken languages and mother tongue. A sample senten e was in luded in order to test whether instru tions were lear. Instru tions were verbally given in Fren h by the resear her and written in the instru tions part of the test. Presentation, so io-linguisti ontained in the rst page. questionnaire, instru tions and example were Parti ipants were kindly requested not to turn the page before having understood instru tions and having lled the required information in. The linguisti both part was ategories being ten e had two omposed of 8 experimental senten es and 8 llers, onstru ted on denominal ausative verbs. Ea h sen- onditions, as the Italian version of the same experiment: (i) full dire t obje t; (ii) pronominal dire t obje t. Parti ipants never judged 106 ������� �� ������������������� �� ������ two onditions of the same senten e. Senten es below report examples of the two onditions. (299) Pour la préparation des sandwi hs, Marie tran he le salami n. For the preparation of sandwi hes, Mary sli es salami thin. (300) Marie travaillait à la bibliothèque où elle empilait les livres hauts. Mary worked at the Library where she piled books high. There were two dierent versions of the questionnaire with two dierent random orders. Parti ipants were asked to identify whi h interpretation they assign to senten es by hoosing between the two proposed under ea h senten e. Namely, (i) adje tive modies dire t obje t, a lassi DP modier within the obje t DP; (ii) adje tive modies the impli it entity, the pseudo-resultative interpretation. Table 5.1 at page 106 reports an experimental item and the task � to be performed on it . Senten e Pour la préparation des sandwi hs, Marie tran he le salami n. Reading 1 A partir du salami, Marie fait des tran hes nes. Reading 2 To prepare sandwi hes, Mary sli es the salami thin. From the salami, Mary reates some thin sli es. A partir du salami n, Maria fait des tran hes. From the thin salami, Mary reates some sli es. Table 5.1: Condition 1. Example of experimental senten e (Semanti inter- pretation task FR). Two lasses of undergraduate students of Linguisti s and two lasses of students of an Italian language private s hool, in whi h the resear her worked at the time as Italian tea her, de ided to parti ipate in the task. were administrated in the Parisian region. All tests � I have tested 44 Fren h native speakers (37 female) , of dierent age groups, as reported in table 5.2 at page 107. All speakers possess at least a high-s hool formation: 72.73% of infor- mants have an undergraduate level, the 18.18% of informants are graduate and 2.27% possess a PhD, as shown by table 5.3 at page 107. All informants are Fren h native speakers and do not have other mother tongue, furthermore, they do not speak other languages at high pro ien y levels. � Experimental items and instru tions grammati ality have been he ked by a native speaker. � Experiments of non-native Fren h speakers were eliminated from the analysis ���� ������ ������������������� 107 Age group % 18-25 65.91 26-32 6.82 33-40 6.82 41-60 13.64 60+ 6.82 Table 5.2: Parti ipants age (Semanti interpretation task FR). Level of edu ation % Ba helor 9.09 Ba helor +1 18.18 Ba helor +2 29.55 Ba helor +3 4.55 Ba helor +4 11.36 Ba helor +5 15.91 PhD 2.27 Table 5.3: Parti ipants edu ation level (Semanti interpretation task FR). 5.2.1.1 Results for ondition 2 (pronominal dire t obje t). Adje tives within senten es with pronominal dire t obje t are interpreted as pseudo-resultative modiers with a rate of 92.29% (s= 11.34). An example of experimental items in the se ond ondition is reported in table 5.4 at page 107. Senten e Pour la préparation des sandwi hs, Marie a hète le salami et elle le tran he n. Reading 1 A partir du salami, Marie fait des tran hes nes. Reading 2 To prepare sandwi hes, Mary buys salami and she sli es it thin. From the salami, Mary reates some thin sli es. A partir du salami n, Maria fait des tran hes. From the thin salami, Mary reates some sli es. Table 5.4: Condition 2. Example of experimental senten e (Semanti pretation task FR). This result is perfe tly aligned with the results obtained in Italian. inter- ������� �� ������������������� �� ������ 108 5.2.1.2 Results for ondition 1 (expli it dire t obje t). Data are mu h more interesting regarding the interpretation of adje tive when ontained in senten es with an expli it dire t obje t. Results are not homogeneous, the adje tive re eives a pseudo-resultative interpretation only in senten es built on tran her, empiler, `to pile', tresser, `to braid', `to sli e', as reported in graph 5.1 at page 108. � ��� ��� ��� ��� ��� �� �� ����� ��� ��� ��� ��� � Figure 5.1: Condition 1. Per entage of interpretation (Semanti interpreta- tion task FR). Senten es built on these three verbs re eive a signi ant higher rate of PR interpretation. Applying a Chi-square test among these two groups, we obtain a hi value of 9, 852˙10−8, showing that the dieren e within these two groups is statisti ally signi ant. I assume that it an be as ribed to a phonologi ee t. ���� ������ ������������������� 109 5.2.2 Dis ussion Contrary to expe tations, Fren h pseudo-resultative eptable for most of the part of the tested verbs. signi ant dieren e in PR a pile', tresser `to braid', tran her onstru tion is not a There is a statisti ally eptability rate of three verbs: empiler `to `to sli e'. I argue that the dieren e in PR a eptability for these verbs is due to more transparent phonologi al relationship between the verb and the base. In other words, the phonologi al form of the base in the verb is as ribable to the phonologi al form of the base when it fun tions as an independent noun. (301) (302) empiler, b. to pile pile, [pil] pile a. tresser, b. (303) [Ãpile] a. a. b. [trese] to braid tresse, [trEs] braid tran her, [trÃSe ] to sli e tran he, [trÃS] sli e Other verbs employed in the experiment do not entertain a dire t phonologi al relationship with their bases as pointed out by the following examples. (304) (305) a. entasser, b. to sta k tas, [ta] sta k a. b. [Ãtase] amasser, [amase] to amass amas, [ama] heap I argue that the derivational nature of verbs is not always a essible to speakers when the base is phonologi ally distin t to the verb. For this reason, in verb (su h as amasser ) ase of phonologi al ina merges within going a morpho-synta ti as shown in (307). v, essibility of base noun, the sin e it is not per eived as under- pro ess of derivation from the base noun ( amas ), ������� �� ������������������� �� ������ 110 (306) [...℄ amasser les livres hauts. ... to pile the books high. (307) Voi eP ... vP v DP amasser les livres hauts This pro ess prevents the grammati al formation of pseudo-resultative onstru tion. Sin e the verb is not per eived as derived from a base, the pseudo-resultative adje tive annot modify it, and the only possible inter- pretation for the adje tive is to fun tion as the dire t obje t modier. 5.3 This Con lusions hapter reports results of a semanti Fren h native speakers on erning the a interpretation task ondu ted on 44 essibility of pseudo-resultative on- stru tion in senten es built on denominal verbs. Results show that pseudoresultative reading is a essible only for 3 verbs out of 8. I argue that this is due to the tighter phonologi al relationship between these verbs and their bases, whi h onsequently guarantees synta ti tion in whi h the base root is merged separately from v and deriva- an hen e be modied by the PR adje tive. Thus, I propose that the lower a eptability rate for PR in Fren h is due to a lower phonologi al transparen y of verbs. In Italian, where a higher degree of phonologi al transparen y exists between the root and the derived verb, PR interpretation is more readily available. Part II Ambiguous verbs 111 Introdu tion The se ond part of the present dissertation investigates the relationship between stativity and ausation with spe ial attention to a lass of deadje tival Italian verbs whi h generates a double aspe tual reading. We will see that stativity and phenomena, but that they an o ausation are not two opposite linguisti ur in the same stru ture of ausative stative verbs, su h as assumed for obje t-experien er verbs (Pylkkänen 2000). order to propose a uniform theory about to ausation, a for e-dynami In approa h ausation (Copley & Harley 2015) is presented and applied, with due modi ations. In parti ular, on eptual energeti for es that we nd in the world are demonstrated to be linguisti ally signi ant. Stative verbs do not involve on eptual energeti reason, I argue that ausative eventives stay at energeti verbs stay at abdu tion. of on epts su h as ausation, sin e no for e is involved. This a base. This for es as stative ount is possible thanks to the separation ausation and Con erning Italian deadje tival that they For this hange. ausative parasyntheti an be divided into three dierent verbs, we will see ategories depending on their hapter fo uses on two of them. One lass regroups verbs formed from adje tives of form, namely those adje tives whi h involve a physi al (and onsequently energeti ) The other lass hange, su h as not ne essarily involve a physi al the obje t whi h takes pla e a the latter lass grande, pesante, hange, but only a presumed `heavy'. ording to the speaker. hange of Verbs belonging to an have both an eventive and a stative stru ture whi h is ree ted by the (in)anima y of the subje t. spe i `big', and ontains verbs formed from adje tives of surfa e, whi h do Typology is built by means of stativity diagnosti s. The denition of pre ise diagnosti s for stative verbs is problemati (i) many of the tests presented in the literature are language-spe i annot be transposed sin e and ross-linguisti ally; (ii) some of them sele t for epiphe- nomena that are often (but not always) linked to stativity. Chapter 6 analyses some of the most popular stativity tests for Italian and shows that synta ti tests (agrammati ality in progressive and imperative) are not reliable, sin e 113 114 stative verbs an be easily for ed in parti ular stru tures. On the other hand, we will see that tests based on semanti interpretations (interpretation under modal verbs, interpretation under temporal adverbials and temporal narrative ontribution) are more reliable both in Italian and English and are likely to be v ross-linguisti ally valid as well. Chapter 7 proposes a syntax for ausative stative verbs in whi h avor of is responsible for establishing the ausal link between Causer, Theme and properties of the Theme. v Contrary to usual statives for whi h a relational is responsible for the identity relation between Holder and Rheme (Harley 2005), we assume that stative ausation is brought about by the presen e of a rP in the lower part of the derivation. tion A virtual for e alled abdu - introdu ed in the system by the speaker (this is the speaker's opinion) an be thought as the stative & Harley 2015) of eventive orrespondent of the energeti ausative verbs. opinion is further pointed out by a pragmati 2005; Stephenson 2007) whi h for e (Copley The importan e of speaker's judge parameter (Laherson an refer to dierent parts of the stru ture in relation to the eveneutality of the verb. Chapter 8 reports results of a pra ti al appli ation of stativity diagnosti s presented in hapter 6. It is the out ome of a wider proje t held by Bridget Copley (CNRS) and Phillip Wol (Emory University), whose nal aim sists in the automati will see, temporal on- interpretation of temporal senten e orientation. As we onstraints involved by stative and eventive verbs play a big role in the determination of temporal orientation of senten es. tions about strategies for the automati Ree - identi ation of statives are reported, with parti ular attention to pro edural stages employed for the stativity gradient of English verbs extrapolated from a orpus. reation of a Chapter 6 Stativity diagnosti s in Italian 6.1 Introdu tion Stativity seems to be a sort of lexi al feature asso iated to parti ular stru tures and prohibited in others. There are usual stative verbs, whi h are more di ult to be for ed in eventive stru tures, su h as love, own, be. Other verbs seem more plausible in ambiguous readings, su h as deadje tival parasyntheti s ( f. We hapter 7). an imagine that this propensity for ambiguity resides in extra-linguisti fa tors, probably in a high ognitive fa ility for stative verbs of being read as eventive if for ed into a proper stru ture. The fa t that stative verbs an (almost) always be oer ed into eventive stru tures makes it important to have reliable stativity tests whi h are not inuen ed by external fa tors. One issue tivity. onsists in the la k of a pre ise and ee tive denition of sta- Consequently, the design of diagnosti s is empiri . In other words, the fa t that stative verbs do not benet from a positive denition, and are dened negatively with respe t to eventive verbs (i.e. nami , not teli , not of statives are not dy- hange, do not introdu e an agentive subje t, ...), makes the job of nd diagnosti s hard. Diagnosti s are not planned to pi k up spe i properties, but to not pi k up properties of other aktionsarten. The risk of pi king up epiphenomena is high, be ause the exa t nature of the phenomenon investigated, namely stativity, is vague. We will see that tests whi h seem to yield better results ti on ern seman- interpretation, rather than stru tural ill-formedness. At least in Italian, progressive periphrasis and imperative do not dierentiate between eventives and statives, sin e these tests appear not to produ e agrammati al senten es with statives. However, statives and eventives re eive dierent interpreta- 115 ������� �� ��������� ����������� �� ������� 116 tions under modal verbs; and they produ e dierent temporal and intera t dierently with temporal adverbs like gia, onstraints `already'. It appears that statives and eventives dier with respe t to impli ations in temporal domain, the former being an hored to the present and the latter to the future. Tests whi h sele t epiphenomena should be avoided. In fa t, even though epiphenomena are tightly ould be onne ted to the target to be investigated, there ases in whi h epiphenomena appear in isolation. This issue is illustrated by the appli ation of adverb voluntarily to dete t eventivity (Lako 1966; Dowty 1979). In fa t, the adverbial modi ation of the predi ate does not dete t eventivity (the phenomenon), but agentivity (the epiphenomenon). (308) John losed the window voluntarily. (309) *The wind losed the window voluntarily. This leads to the in orre t taining a non-eventive verb, ategorization of senten es like (309) as on- ontrary to the fa t. Another example of the di ulty to dene reliable diagnosti s for eventualities for-X-time test, whi h was performed to dete t ateli ity. to identify  operation involving a series of small hanges  onsists in It a tually seems (Erteshik-Shir & Rapoport 2004: 76). (310) Giovanni ha hiuso la nestra per 10 minuti. John losed the window for 10 minutes. (311) Giovanni ha rotto il bi hiere per 3 minuti. John broke the glass for 3 minutes. Normally, adverbial totypi al ausative teli for-X-time is expe ted to be agrammati al with pro- verbs, therefore we expe t it to be una examples above, but it is not the eptable in ase. If the test does not identify ateli ity, it doesn't identify the duration of the nal/resultant state, but it identies the duration of the pro ess. Sin e the pro ess of breaking glasses is a pro ess of breaking glasses only if it attains the nal/resultant state of having glasses broken, it implies that the adverbial generates a reading where dierent small events of breaking a unique single glass take pla e. The present hapter will analyze some attested stativity diagnosti s in order to see if they are valid, in parti ular for Italian. Se tion 6.2.1 presents imperative and progressive periphrasis as diagnosti s for stativity in Italian, se tion 6.3 dis usses diagnosti s where no agrammati ality is expe ted but where dierent readings are systemati ally asso iated to onstru tions involving stative or eventive predi ates. The last ���� ����������������� 117 two se tions report experimental tests on behavioural responses to stative and eventive predi ates, suggesting that experiments an be useful for the determination of stativity. 6.2 Agrammati alities This se tion dis usses stativity diagnosti s whi h are based on synta ti tures, in other words on agrammati alities. fea- Namely, these tests are aimed to identify dierent eventualities on the basis of the (a)grammati ality of a predi ate in a parti ular stru ture. We will see that these diagnosti s are not ross-linguisti ally valid and are not always reliable. 6.2.1 Spe i They Imperative and progressive periphrasis Italian tests to dete t stativity are identied by Bertinetto (1991: 30). onsist in the impossibility to appear in the imperative form and in the impossibility to appear in the progressive form. In this by stare, (312) hapter we will analyze one Italian progressive periphrasis formed `to be', and the gerund of the lexi al verb. Maria sta M. stay- 3.sg. ballando. ger. dan e- Mary is dan ing. A huge dieren e in reading between Italian and English progressive periphrasis is determined by the fa t that the Italian simple present a progressive reading, while its English (313) Maria balla. (314) Mary dan es. The Italian example (313) ounterpart an re eive annot. an re eive both a habitual reading and a progressive reading (Bertinetto 2000: 565), while the English (314) generates only a habitual reading. ounterpart This, among other fa tors, an inuen e the range of meanings generated by Italian progressive. Progressive as diagnosti s for stativity does not always work as expe ted. On the one hand, verb possedere, `possess', whi h is unanimously judged as stative, is agrammati al in the progressive form, as expe ted. (315) *Sta possedendo inque ase. He's possessing ve houses. ������� �� ��������� ����������� �� ������� 118 It is worth noting that some highly stative verbs do not li itly parti ipate in progressive periphrasis, su h as (316) *Maria sta avendo sei have or be. ase. Maria is having six houses. (317) *Maria sta essendo grassa. Maria is being fat. Although, Grossmann (2004: 347) points out that this is not true for all presumed stative verbs su h as (318) sorire, `suer'. Sta sorendo. He's suering. � [t℄he use of the progressive periphrasis is in expansion, parti ularly regarding the type of verbal a tion of verbs to whi h it applies . This statement is supported by a orpora analysis onBeretta (1993: 220) observes that  du ted by Beretta (1993), who reports one ex erpt: (319) [...℄ sai he non mi refl.1sg. [...℄ (you)-know that not se remember- ger. io al lavoro ho if I at sto ri ordando am las iato i work have miei zo left the my oli. logs You know, I don't remember if I left my logs at work. Consequently, the reliability of the progressive periphrasis in order to dete t stativity in Italian is dubious. Many other verbs behave ontrary to expe tations, and progressive periphrasis, as the following examples show. (320) Maria sta amando questo aè. Maria is this oee. lov- ger. Mary is loving this oee. (321) Il aè sta pia endo The oee is ger. like- a Maria. to Maria Maria is liking the oee. (322) Giovanni sta avendo un atta Giovanni is a ger. have- o di atta k uore. of heart John is having a heart atta k. � Translation of the following paragraph is mine. an appear under ���� ����������������� (323) 119 Sto avendo sete. be- have- thirstiness pres.1.sg. ger. I'm thirsty. It is worth noting that all senten es des ribe a non-habitual eventuality whi h is limited in time. For example, questo aè, `this oee', is parti ular il aè, ` oee' in (324), whi h and it is limited in time. The same way for is not parti ular per se, but re eives a parti ular reading when it fun tions as the subje t of a progressive periphrasis. (324) Il aè pia e a Maria (habitually). The oee pleases to Mary (habitually). Italian progressive periphrasis for es a parti ular and time limited read- [...℄ the Italian dia hroni data show that at the beginning the progressive refers to purely durative situations and only later it has spe ialized as an aspe tual form, not expressing purely durativity, but imperfe tivity  (Squartini 1998: 102). Time limited reading implies ing, whenever this is available, in fa t  that the state holds for a limited time period: a tivities that in lude the entire whole life of the subje t annot be expressed by Italian progressive. may be employed only in ases of stri t fo alization [...℄ where the speaker is only on erned with what is going on at a parti ular point in time  (Bertinetto 2000: 564). Italian progressive periphrasis  (325) Maria sta lavorando a s uola. Mary is working at s hool. Senten e (325) annot be intended, for the same ontext, as the Spanish progressive (Squartini 1998: 110) where Mary would be interpreted as working habitually in a s hool. In Italian, Mary is working in a s hool only for a ertain period and not habitually. This an be extended to statives sin e, as while the author assumes that statives  the progressive in general, he are not admitted  (ibid : 104) in ontends that only permanent statives (ILPs in international terminology) are statives (SLPs) are more a laimed by Squartini (1998), ategori ally ex luded, but non-permanent eptable in the progressive. On the other hand, Bertinetto (2000: 583 .) re ognizes that statives are not systemati ally agrammati al under progressive periphrasis, irrespe tive of the type of state involved. This is as ribed to a possible double reading of the lexi al entry, whi h an be either stative, when used in present tense, or eventive, when used with progressive periphrasis. We already saw that this last assumption is (320) and (321). ontradi ted by examples It is hard to arm their eventive interpretation, be ause ������� �� ��������� ����������� �� ������� 120 the subje t is rather in a state of loving the loving. oee, than in a pro ess of I agree with both Bertinetto (2000) and Squartini (1998) that not all statives are feli itous under progressive periphrasis, but I do not agree with them on ategorization of verbs that allow the progressive. Squartini argues that the agrammati ality of statives under progressive is due to the permanent vs. temporal nature of the state involved, while Bertinetto argues that only eventive verbs in lexi al an appear under progressive, determining a ategory in the We saw that verbs su h as o avere, ur in the progressive, while verbs su and avere sete, `be thirsty', by two fa ts: subje ts of hange ase of statives. essere, `to be', annot h as amare, `to love', pia ere, `to like', `to have', and an. I argue that this asymmetry is determined amare, `to love', pia ere, `to like', and avere sete, `to be thirsty' are in a dire t relation with the state, in the sense that the subje t is the Experien er and the state is neither durative nor habitual ( ontrary to avere, `to have', and essere, From these examples, I stare `to be'). an assume that Italian progressive periphrasis + gerund is not an e ient diagnosti s for stativity, be ause it does not systemati ally ex lude all stative verbs. Another diagnosti s of stativity that is often put forth along with progressive periphrasis is the use of imperative. However, doubts are raised about its reliability in pi king out only eventive verbs (Grossmann 2004). In the Roman e panorama, imperative is a proper verbal mode, even though morphologi al syn retism with Indi ative and/or Subjun tive is present in dierent languages, su h as in Fren h and Italian. perative presents spe i (namely A However, Italian im- morphologi al marks in one of three onjugations -are ). ording to Squartini (1990) and Levin (2007), the ungrammati ality of stative verbs under imperative is probably due to their la k of agentivity. Eviden e omes from verbs, su h as in (326) and (327), that are a eptable in the imperative only if the subje t is an Agent, and are agrammati al when the subje t is a Patient. (326) Vola! Fly! (327) Giovanni è orso a lavoro. John is run at work (328) Corri a lavoro! Run at work (329) *Arriva! Arrive. ���� ����������������� 121 A good eviden e for the use of imperative in onjun tion with statives is represented by the Italian translation of Ten Commandments: (330) Ri ordati di santi are le feste. Remember the sabbath day, to keep it holy. (331) Onora il Padre e la Madre. Honour thy father and thy mother. (332) Non desiderare la donna d'altri. Thou shalt not ovet neighbor's wife. (333) Non desiderare la roba d'altri. Thou shalt not ovet neighbor's belongings. Clearly, in none of these perien er. This way, a ases the subje t is an Agent, rather it is an Ex- ording to Squartini and to Levin, previous senten es should be agrammati al, and this is not the ase. Grossmann (2004) also observes that statives are sometimes a eptable in the imperative mood independently of agentivity: (334) Sperate pl. Hope-2. di essere promossi! of be passed Hope to pass the year! (335) Dimenti ami! Forget me! In both senten es above, the person to whom they are addressed is dire tly involved in the a omplishment of the requirements. In other words, pupils (whom the rst senten e is addressed to) have the power to improve their notes; likewise the person to whom the se ond senten e is addressed has the power/possibility to forget someone. This way, the subje t seems to be a sort of Agent, be ause it has the possibility of enabling the eventuality des ribed by the imperative. And the main hara teristi of agentivity onsists in the fa t that an individual has the possibility to a t in an eventuality. the the It appears lear that the grammati ality of imperative does not reside on aktionsart of the verb, but in the possibility of the addressee to inuen e oming into being of the eventuality itself. Consequently, Italian impera- tive is not a good diagnosti s for stativity be ause it seems to be onditioned by other fa tors. To summarize, neither progressive nor imperative are good stativity diagnosti s in Italian. andidates for ������� �� ��������� ����������� �� ������� 122 We will see in hapter 8 that progressive in English is a tually a perfe t andidate to identify stativity when applied to Natural Language Pro essing. We should ask then why English progressive is sensitive for stativity while Italian progressive is not and how the two dier in the intera tion with eventualities. 6.3 Semanti interpretation tests In previous se tions we saw that synta ti statives tests useful to identify English annot be applied to Italian. It is the ase of progressive periphrasis and imperative. This language-spe i behavior ould be as ribed to numerous su h as dierent temporal impli ations of spe i auses, stru tures (su h as pro- gressives), dierent aspe tual prohibitions due to the wrong ombination of lexi al verbs and grammati al stru tures. However, there are some tests of stativity whi h do not involve grammati ality, but whi h involve semanti and in Italian: poral interpretation, reliable both in English interpretation under modal verb (epistemi /deonti ); tem- onstraints (present/future); possible ontribution in narrative hain. These tests are interesting be ause they seem to rely on a general and fundamental feature of stativity. We will use these tests in Italian deadje ti al verbs hapter 7, in order to show that a group of an re eive two readings tightly onne ted to the base adje tive and showed by the (in)anima y of the subje t. 6.3.1 Interpretation under modal In this subse tion, I will analyze dierent interpretations that a stative verb an generate under modal verb. I will report results of an experiment du ted on Italian native speakers whi h on- onrms this dieren e in interpre- tation. Giorgi & Pianesi (1997) des ribe dierent modalities assumed by modal verbs in dierent languages: they identify dierent interpretations of prediates under omplement of dovere, `must'. When a verb is embedded under `must', the stru ture readings: deonti a an generate two and epistemi . The deonti /obligational reading expresses ommand about an a tion that must be realized. The epistemi reading on erns a spe ulation about a present state of aairs. These two readings entertain also a temporal onstraint: deonti /obligational reading generates ���� �������� �������������� ����� + stative dovere - epistemi - present +eventive a future and deonti reading - future onstraint onstraint, while the epistemi reading onstraint - deonti Table 6.1: Dierent readings and 123 onstraints of modal a present dovere. onstraint. A summary is present in table 6.1 at page 123. The two interpretations arise in relation with the eventuality of the predi ate. Eventive predi ates predi ates both an epistemi an re eive only a deonti and a deonti that the preferred one is the epistemi reading, while stative reading, even though tests show (refer to se tion 6.3.1.1). In senten es below we use two predi ates under dovere, `must': amare Matteo, `love Matteo' (stative), and orrere questa maratona, `run this marathon' (eventive). Completions show that they entertain two dierent readings. (336) Maria deve amare Matteo... Mary must love Matteo. a. per fare una s elta osì s io a. (epistemi ) in order to make su h a stupid hoi e. b. per essere una buona moglie. (deonti ) in order to be a good wife. (337) Sandro deve orrere questa Maratona... Sandro must run this Marathon, a. # per onsumare le s arpe in questo modo. (epistemi ) # in order to use shoes that way. b. per dimagrire. (deonti ) in order to lose weight. The same is appli able to English, showing the possible ross-linguisti validity of the diagnosti s. (338) (339) Mary must love Mar o. a. in order to make su h a stupid b. in order to be a good wife. hoi e. (epistemi ) (deonti ) Sandro must run this Marathon. a. *in order to use shoes that way. (epistemi ) ������� �� ��������� ����������� �� ������� 124 b. in order to lose weight. (deonti ) We saw that senten es involving the stative verb an epistemi and a deonti amare an produ e both reading of the modal. To summarize, I reported stativity diagnosti s linked to the dierent interpretation of modal verbs. Namely, stative verbs and deonti an generate both epistemi interpretation, while eventives have only a deonti The following se tion reports results of an experiment reading. ondu ted on 188 Italian native speakers. They judged the interpretation (deonti /epistemi ) of senten es ontaining stative or eventive verbs. Results show that this dif- feren e in interpretation uts a ross statives and eventives and it is perfe tly per eivable in everyday language. A similar test has been ondu ted on 25 English native speakers in the NLP proje t and its results are reported in Chapter 8. 6.3.1.1 Semanti interpretation task under modal This se tion reports all steps of a semanti de ision task ondu ted on 188 Italian native speakers about the interpretation of eventive and stative verbs under modal dovere, `must'. We will see that results onrm the hypothesis about the deonti tation of eventives and the possible double deonti /epistemi interpre- interpretation for statives. The experiment is divided in two parts: (i) the so iolinguisti naire; (ii) the linguisti part. The formulation of instru tion hes been a deli ate task. ti and epistemi question- are s ienti Terms deon- terms, to make this experiment trustworthy they must be translated in everyday language. In important role in the experimental su hapter 1, we saw that an ess is held by the larity of instru - deonti is translated with omando, ` ommand', and translated with osservazione/opinione, `assumption/opinion'. tions. For this reason, epistemi is The so iolinguisti questionnaire investigates for age, sex, edu ation and origin of parti ipants with an usual format. The linguisti part is omposed of 56 senten es: eventives; 14 senten es built on tives. 28 senten es built on ausal statives; 14 built on non- ausal sta- Subje ts of both groups of eventive and stative verbs were equally divided into animate and inanimate DPs (ref. table 6.2, page 125). Senten es ontain the modal dovere, `must', at the present tense. senten es were built with subje t + verbal omplements. Sin e generi All omplex + dire t obje t and obje ts inuen e eventuality, omplements are quantized obje ts, as shown by following examples reporting one senten e per ���� �������� �������������� ����� 125 Stative Causative Animate Eventive Non- ausative Inanimate Animate 7 7 7 14 Inanimate Animate 7 14 14 Inanimate 14 28 Table 6.2: Condition distribution (Interpretation under modal ITA). ondition: (340) stative with animate subje t; (341) stative with inanimate subje t; (342) eventive with animate subje t. (340) Carla deve onos ere il ontenuto del testamento di Maria. Carla must know the ontent of Mary's will. (341) Il libro sulla storia d'Italia deve interessare Maria. The Italian history book must interest Mary. (342) Sandro deve s iogliere del burro. Sandro must melt some butter. Three native speakers who did not parti ipate in the experiment tested all senten es for plausibility before the experiment administered. Other three native speakers tested the experiment in its beta version before its on-line version was laun hed via so ial networks and e-mails. Parti ipants were asked to judge all 48 senten es, whi h were presented in random order (determined by IbexFarm internal algorithm). 188 Italian native speakers (133 female) parti ipated in the experiment, aged of 32,18 years (minimum 20 and maximum 62) with a high edu ation � degree (91 PhD, 86 Degree, 11 High-S hool) distributed in the three main Italian varieties (North, Center, South) with a predominan e of the North variety. These data are reported in table 6.3 at page 125. Edu ation Origin Gender Phd 91 North 122 Female 133 Degree 86 Center 22 Male 55 High-S hool 11 South 53 Table 6.3: So iolinguisti s of parti ipants (Interpretation under modal ITA). � It is worth noting that this edu ation rate is not representative of the national mean. ������� �� ��������� ����������� �� ������� 126 Although parti ipants were asked to they hoose one or both interpretations, hose only one interpretation as expe ted. This is a well-known problem of naive speakers, who are not used to jump from one reading to another for one single senten e (like linguists do). For this reason, results do not show the expe ted predominan e of both answer for stative verbs. eventives and statives were � However, learly interpreted dierently (as reported in table 6.4 at page 128 ). � For translations of senten es, please refer to appendix. (Continue on the next page) Command Both 171 178 156 157 164 56 145 166 167 168 181 181 102 124 180 172 182 142 175 124 158 164 145 149 170 85 176 133 4 6 14 5 11 53 10 10 7 5 3 3 51 23 2 6 2 13 4 22 11 7 14 15 6 41 4 16 13 4 18 26 13 79 33 12 14 15 4 4 35 41 6 10 4 33 9 42 19 17 29 24 12 62 8 39 ����� �������������� �������� ���� Sn01 - Carla deve adorare il suo nuovo ollega Sn02 - Giulio deve amare il gelato al io olato Sn03 - Maria deve ammirare la nuova trasmissione televisiva Sn04 - Questo bell'anello deve appartenere a Maria Sn05 - Giulio deve apprezzare le anzoni di Battisti Sn06 - Carla deve onos ere il ontenuto del testamento di Maria Sn07 - Maria deve redere alle bugie di suo marito Sn08 - Giulio deve desiderare quelle s arpe in vetrina Sn09 - Maria deve detestare quel divano marrone Sn10 - Carla deve invidiare Maria Sn11 - La presenza del sole deve man are a Giulio Sn12 - Sandro deve odiare il aè ma hiato Sn13 - Sandro deve possedere quella ma hina sportiva rossa Sn14 - Sandro deve temere il ane del suo vi ino di asa S 01 - I brutti sogni devono angos iare il bambino di Maria S 02 - Questa tisana deve agitare Carla S 03 - Il on erto deve annoiare Sandro S 04 - Lo spetta olo del mago deve divertire Giulio S 05 - La giostra del par o deve impaurire Maria S 06 - Maria deve infastidire Carla S 07 - Le bolli ine sulla pelle di Carla devono inquietare Giulio S 08 - Il libro sulla storia d'Italia deve interessare Maria S 09 - La puntura del alabrone deve intimorire Giulio S 10 - L'assenza del presidente deve stupire gli impiegati S 11 - Maria deve preo upare sua mamma S 12 - La musi a ubana deve rallegrare la festa S 13 - L'aumento del prezzo dei bus deve s o iare molti utenti S 14 - Il olore di apelli di Sandro deve stupire Carla Assumption 721 Senten e E01 - Giulio deve agitare bene lo s iroppo 9 E02 - Maria deve porre delle ondizioni pre ise 14 E03 - Il essibile deve spezzare la atena della bi i letta 70 E04 - L'aumento delle tasse del 2017 deve azzerare le dierenze so iali 56 E05 - Sandro deve s iogliere del burro 12 E06 - Maria deve diventare una dottoressa 44 E07 - Giulio deve guadagnare il suo primo stipendio 26 E08 - Carla deve vendi are la morte di suo fratello 36 E09 - La erimonia di apertura deve intrattenere gli spettatori oreani 32 E10 - Giulio deve lavorare alla sua tesi 12 E11 - Carla deve attare la sua asa in ampagna per un mese 36 E12 - Maria deve pesare il pros iutto 3 E13 - L'azienda deve importare 8 ontainer di pezzi di ri ambio 11 E14 - Sandro deve sostituire la sua ve hia automobile 36 E15 - Maria deve votare il nuovo delegato sinda ale 11 E16 - Carla deve riferire la notizia a Giulio 4 E17 - Giulio deve rubare mille euro dalla assaforte di suo papà 28 E18 - La pro edura dis iplinare deve de lassare Sandro 24 E19 - La medi ina deve guarire Sandro 73 E20 - La manovra nanziaria deve azzerare il debito pubbli o 47 E21 - La disinfestazione deve eliminare metà delle zanzare 42 E22 - La legge deve abolire la s hiavitù 10 E23 - Il ris aldamento autonomo deve rimpiazzare quello entralizzato 18 E24 - La ristrutturazione deve allontanare i due muri portanti 17 E25 - Il dibattito televisivo di stasera deve ontrapporre gli avversari 50 E26 - La nuova giunta omunale deve distruggere il ve hio entro ommer iale 23 E27 - Sandro deve avvelenare tutti i topi he abitano nel suo granaio 21 E28 - La ala di antidoto per il veleno deve risvegliare Giulio 91 Table 6.4: Experimental items (Interpretation under modal ITA). Command Both 170 135 94 79 154 79 113 107 107 121 101 160 158 86 157 152 129 124 73 92 98 139 138 152 87 125 128 57 9 39 24 53 22 65 49 45 49 55 51 25 19 66 20 32 31 40 42 49 48 39 32 19 51 40 39 40 821 Assumption ������� �� ����������� ��������� �� ������� Senten e ���� �������� �������������� ����� 129 Referen es Sn01 to Sn14 point to stative non- ausative verbs, S 01 to S 14 point to stative ausatives, E01 to E28 point to eventive verbs. Senten es from Sn01 to S 14 re eive a predominant assumption interpretation, onrming an epistemi of predi ates involved. reading and onsequently the stative nature No dieren es are registered among dierent tions (subje ts' (in)anima y and ausality). I a role in the distin tion between epistemi ondi- on lude that they do not play vs. deonti interpretation. Senten es from E01 to E28 show the prevalent  ommand answer, rming their deonti on- interpretation and thus the eventiveness of predi ates involved. It is worth noting that some of the (presumed) eventive senten es re eive less sharp results (su h as E03). I argue that these senten es were easily interpreted as habituals, making them statives. This properly allows a higher rate of assumption answers. Results of this task show that the interpretation re eived by verbs under modal dovere, `must', is a good diagnosti anima y of the subje t and s for stativity, independently of the ausal semanti s of the verb. 6.3.2 Future/Present onstraint Dierent interpretations of modal are linked to issues of temporal nature (Katz 2003: 6) whi h are in turn onditioned by the eventuality of the pred- i ate. Imagine two senten es headed by you must, whose omplements are re- spe tively a stative and an eventive predi ate; the eventive one requires the eventuality to be realized in the future in order to make the senten e true; the stative one requires the statement to be realized in the present in order to make the senten e true. Note that the temporal interpretation of the omplement in [a stative senten e℄ is present-like, while in [an eventive senten e℄ is future-like. [The stative one℄ means that given what we know now it follows that you love Lin now, while [the eventive one℄ means that to be in line with requirements you need to kiss Lin sometime in the future (Katz 2003: 6). The type of eventuality expressed by the lexi al verb determines the temporal orientation of the senten e (Condoravdi 2002: 69). Dierent orientations of modals depend on the  ating eventualities to the referen e time  temporal relation for lo� (Condoravdi 2002: 70) . Whether modals and onditionals follow the same temporal onstraints is matter of debate, refer to Copley (2008, 2014). � 130 ������� �� ��������� ����������� �� ������� Condoravdi assumes that the AT relation (translation of temporal operators) varies in its interpretation depending on the eventuality involved. This an be represented by the following equation (Condoravdi 2002: 19). (6.1)   ∃e[P (w)(e) & τ (e, w) ⊆ t] AT (t, w, P ) = ∃e[P (w)(e) & τ (e, w) ◦ t]   P (w)(t) The property P is instantiated in world ( depends on the type of eventuality. w) if if if P P P is eventive is stative is temporal t at time ( ) in a way that If the eventuality is a state, P is a property of states and there is temporal overlap; if the eventuality is eventive, P is a property of events, and there is time in lusion; if P is a property of times, the property P holds at A t in w. ordingly, there is temporal overlap when some part of an eventuality overlaps with the time of utteran e, and temporal in lusion when the starting point is in luded in the segment of time identied by the time of utteran e. In other words, temporal overlap is obtained when the state started at some point in the past, before the time of utteran e. Temporal in lusion is obtained when the event starts at some point in luded in the time of utteran e and nishes some time after (Condoravdi 2002: 73). Modals expand the lo al time of evaluation, in the absen e of ontextual eviden e: modals for the present with statives determine that the temporal tra e of state in ludes time of utteran e ( ibidem ), sin e they involve temporal overlap, as exemplied in (343) and (344), where temporal adverbials spe ify � the temporal interpretation of senten es . (343) He might be here (now). (344) He might run (now). This same temporal onstraint applies in other ontexts, su h as the present (simple) tense, both in Italian and in English. The following examples show that present tense senten es (without a habitual interpretation) generate a dierent temporal onstraint depending on the eventuality of the verb. (345) Daria odia questo aè adesso/*domani. Daria hates this oee now/*tomorrow. It is worth noting that this an be due to ontrollability of the eventuality by the subje t (Copley, p. .). Thus, if an event is un ontrollable, it results in agrammati ality for statives too, as in Daria gets si k *tomorrow. � ���� �������� �������������� ����� (346) Daria va a 131 asa adesso/domani. Daria goes home now/tomorrow. (347) Mary likes this up of oee in this very moment/*tomorrow. (348) Mary plays the third game *in this very moment/tomorrow. Noti e that Italian and English eventive examples dier be ause Italian allows a progressive reading with the present tense, while English does not, making hen e impossible to use the temporal adverbial but this is a language-spe i in this very moment, property of Italian tense. Stative and eventive verbs present dierent temporal modals and with present tense. This dieren e onstraints under an be employed for the determination of the eventuality of ambiguous verbs. 6.3.3 Contribution in narrative dis ourse In this se tion we study the impossibility of moving forward the narration time in a narrative dis ourse that involves stative verbs. Stative verbs annot inuen e the narrative hain, i.e. they do not tribute to its temporal progress (Dry 1983; Katz 2003), on- ontrary to eventive verbs whi h trigger a narrative advan ement. If we look at examples below, we see that, in (349) ea h verb des ribes an a tion whi h takes pla e after the previous one. On the other hand, in (350), eventualities take pla e at the same time. (349) Mary arrived. Her daughter sat down on the ou h and her dog felt asleep. (350) Mary arrived. Her daughter was sitting down and her dog was sleeping. In (349), there is a narrative daughter sitting on the hain starting with Mary arriving home, her ou h and then the dog felt asleep. When the order of verbs in (349) and (350) is s rambled, the of eventive verbs versus stative verbs appears more ase the order of episodes in narration ontribution learly. In the previous hanges (351), in the latter no hanges are involved (352). (351) Mary arrived. Her dog felt asleep and her daughter sat down on the ou h. (352) Mary arrived. Her dog was sleeping and her daughter was sitting on the ou h. ������� �� ��������� ����������� �� ������� 132 We will use this test in hapter 7, on Italian deadje tival verbs, and we will see that it is parti ularly useful to identify stativity for verbs whi h present ambiguous readings. The non-temporal ontribution to a narration hain seems to be linked to general temporal properties of stative verbs. The same fashion as general present orientation with present tense and the epistemi interpretation under modal verbs. I on lude that these three tests are reliable and (probably) ross-linguisti ally valid be ause they are based on general and fundamental properties of stative verbs. 6.4 Experiments involving involuntarily responses This se tion reports results of a self-pa ed reading test ondu ted on English native speakers by Gennari & Poeppel (2003), hen e GP (2003). It shows that stative verbs are pro essed dierently from eventive verbs and onsequently the test ould be used as stativity diagnosti s. we show its short omings in disambiguation However, ase of ambiguous verbs, where the pro ess of ould inuen e experimental results. more than one meaning is pro essed slower sin e the Usually a verb with ognitive ost of inter- pretation is higher. Pro essing verb meaning is inuen ed by lexi al semanti s of verbs, as showed by dierent studies (Brennan & Pylkkanen 2010; mati and argument stru tures have pro essing to the type of event the verb is expressing. predi ates is onne ted to lexi al semanti inter alia ). orrelates whi h are linked Furthermore, eventuality of omplexity. Several studies have provided pro essing eviden e, whi h show that lexi al semanti su h as themati The- roles and argument stru ture, are qui kly a properties, essed by the pro essor when the verb is re ognized (GP 2003: B27). GP (2003) ondu t an experiment whi h aims at investigating whether ausal stru ture of a verb has pro essing orrelates. They base their study on Dowty (1979)'s and Ja kendo 's (1990, 1991) lexi al semanti s: a state is a single event, while events involve two dierent sub-events: → (353) love (354) break (355) arry x love y → → x CAUSE(BECOME y be broken) x's ACT(CAUSE(BECOME y be displa ed) ���� ����������� ��������� ������������� ��������� 133 They use an event-de omposition approa h in order to show that the ausal part of eventive verbs determines a slower reading for eventive prediates than of stative predi ates whi h do not possess it. [...℄ the agent in a breaking event is the initiator of a ausal hain ae ting the patient. This sort of information is required to semanti ally distinguish verbs su h as love and dis over, whi both asso iated with the <experien er, theme> themati ture. The stru - riti al property distinguishing these verbs is whether they denote a GP (2003) h are hange of states (GP 2003: B28). ondu t two experiments: a self-pa ed reading task and a visual de ision task in order to avoid the possibility that signi ant results are due to the pro essing of previous information on the verb. Stimuli of the rst task were rian lasses omposed of eventive (of all three Vendle- on erned) and stative verbs. Verbs were length, argument stru ture, frequen y synta ti of senten es were alike in the (356) The retired musi ian he ked for frequen y, frames and plausibility. Pair riti al segment, ex ept for the verb. built his se ond house from s rat h. (event, GP: B30, ex.1) (357) The retired musi ian loved his se ond hild very mu h. (state, GP: B30, ex.1) Results are statisti ally signi ant, as reported in gure 6.1 (page 134). This test is interesting in itself, however the on lusions are mu h more so. The authors take a depart from the psy holinguisti tradition that onsiders orrelates between verb type and rea tion times due to questions of themati pro essing of event stru ture properties are a tivated during pro essing, and that these properties subsume those of themati roles and argument stru ture  (Ibid.: 34). In fa t, orrelates roles and parti ipant slots. They on lude that  are dierent between eventives and statives that have the same number of parti ipants and the same argument realization. Does the semanti empiri al task. omplexity assumed by lexi al semanti s have some orrelates? They resort to one rst experiment of self pa ed reading The pool of experimental items was omposed of senten es-pairs of stative-eventive, diering for the verb and when ne essary for the internal omplements, he ked both for word-frequen y and for plausibility. They dis over that stative verbs are read 27 ms faster than eventive verbs, [r℄epeated measure ANOVAs omparing reading times at the verb position revealed a signi ant word type ee t (F1(1,29) as reported by GP (2003: 31):  ������� �� ��������� ����������� �� ������� 134 Figure 6.1: Gennari and Poeppel (2003: g. 1): verb reading times. = 10:66, P = 0.003; F2(1.43) = 8,9, P = 0.004). Eventive verbs took 27 ms longer to pro ess than stative verbs . An open question on erns the universal validity of these pro essing lates, sin e statives are not monolithi whether orre- in nature. In parti ular, we should ask ausative statives are pro essed in a signi ant dierent way than non- ausative statives employed by GP's (2003) experiment. Results of Brennan & Pylkkänen's (2010) experiment, whi h analyses (among in hoative er ion ontribution to pro essing o- orrelates) the pro essing of psy hologi al verbs, show that statives involve dierent pro essing osts depending on their ategory. By means of a self-pa ed reading task they analyse whether obje texperien er verbs are pro essed dierently from subje t-experien er verbs. These two ategories are argued to have dierent l-semanti Namely, obje t-experien er verbs involve a ausative omplexities. omponent (Pylkkänen 2000). Results show that obje t-experien er verbs require a higher pro essing ost and are treated slower than subje t-experien er verbs, onrming results of previous studies (Cupples 2010; Gennari & Ma Donald 2009). We now know that non- ausative stative verbs (of the type taken into a ount by GP (2003)) present a dieren e in pro essing with respe t to eventive verbs and that they present a dieren e in pro essing with respe t to ausative sative verb. A joining link is la king, namely the one whi h links eventuality and ausation. Is it possible to isolate a gradient in l-semanti omplexity (i.e. non- ausative stative < ausative stative < eventive < ausal ���� ����������� eventive)? Some l-semanti 135 omponents derive a higher pro essing others (i.e. eventivity more than ost than ausativity)? These questions will not nd an answer in this dissertation, but hopefully they will be addressed in future resear h. 6.5 Con lusions In this hapter we explored dierent stativity diagnosti s reported in the literature for both English and Italian. In parti ular, we saw that some of them dis riminate for epiphenomena whi h often (but not always) arise in onjun tion with stativity. In parti ular, diagnosti s of stative verbs whi h involve ill-formedness in imperative and progressive not ongurations appear to be unreliable, or at least ross-liguisti ally exportable. Diagnosti s whi h seem to work better involve dierent semanti pretations These onditioned by the eventuality of the predi ate in inter- ontexts. ontexts are, for example, the interpretation under modals and the fu- ture or present onstraints. Higher trust-worthiness of semanti is due to the use of spe i ertain synta ti diagnosti s ore features of stativity, rather than the properties of a stru ture, whi h is not always We saw that behavioral tests show some ross-linguisti ally valid. onvin ing eviden e in favor of the dierent treatment of statives and eventives. 136 ������� �� ��������� ����������� �� ������� Chapter 7 Deadje tival parasyntheti verbs 7.1 Introdu tion Any attempt to dene the stru tural element responsible for stative-eventive readings of predi ates is a hallenging task. In this respe t, verbs whi h give rise to two readings are parti ularly interesting for underlining the stru tural spe i ity that is fundamental in triggering this ambiguity. In this hapter, I will analyze a lass of deadje tival prexed verbs with ausative semanti s whi h entail two readings; they are made expli it by the semanti (358) � ontent of the subje t . Giovanni abbellis e la stanza. John makes-beautiful the room. (359) Le fotograe abbellis ono la stanza. Pi tures make-beautiful the room. These verbs are interesting for two parti ular issues. First, they alternate between a stative reading and an eventive reading, this alternation seems to orrelate with the subje t role: when the subje t is animate (Causer), the � verb is eventive , when the subje t is inanimate (Sour e), the verb an be either eventive or stative. Se ond, both stative and eventive readings appear to be ausative (se tion 7.7). Although eventive ausation is not problemati , sin e it has been treated at length in the literature, stative ausation has not re eived mu h attention, and it raises some theoreti al problems. To understand these verbs, we will have to prove that they involve two readings, one stative and one eventive, whi h are both ausal. Additionally, � Not by (in)anima y alone, as we will see in se tions 7.5 and 7.6 . � If the subje t is not read as an inanimate, i.e. John is making the room beautiful with his smile, equal The smile of John is making the room beautiful. 137 ������� �� ������������ ������������� ����� 138 we will have to a ount for the ausal nature of stative verbs. In se tion 7.6, I will show by means of semanti and synta ti tests that verbs under study re eive two distin t interpretations stative and eventive . As we will see, deadje tival parasyntheti verbs behave dierently with respe t to a number of phenomena, in luding epistemi dovere, interpretation under `must' (Giorgi & Pianesi 1997), interpretation under (Mittwo h 2014), ontribution in a narrative già, `already' hain (Dry 1983; Katz 2003) and treatment of adjun ts. In se tion 7.7, I demonstrate that the stative reading, as well as the eventive reading, is ausal (Fabregas & Marìn 2014; Martin & Tovena 2012). I will argue that the (in)anima y of the subje t alone is not su ient to dis riminate between the two eventualities, and that the relationship of the subje t with the property lexi alized by the verbal base plays an important role. Se tion 7.3 is dedi ated to the des ription of morphologi al of the omponents lass of verbs in question. Se tion 7.8.1 des ribes for e-dynami approa h to ausation (Talmy 1985a, 1985b, 1988; Croft 1998, 2012; Copley & Harley 2015; Copley $ Wol 2014b; inter al.) and develops it in order to a je tival parasyntheti ount for stative ausative dead- verbs (hen e DPVs). Se tion 7.9 proposes l-syntax of ausative-eventive DPVs, ausative-stative DPVs and regular statives. Se tion 7.11 fo uses on a pragmati of adje tives (or in our ase of roots). parameter related to spe i This parameter is judge parameter (Laherson 2005; Stephenson 2007) and shows that it applied dierently in relation to type of 7.2 types alled personal an be ausation involved by the verb. Stativity-Eventivity puzzle When we talk about the stative-eventive alternation, we refer to dierent types of meaning shift. In fa t, some lexi al statives in ertain synta ti statives o an be for ed to have an eventive reading environments. This happens, for example, when lexi al ur in the progressive in English. (360) Daria is having one of her ba ka hes. (361) I'm loving it. On the other hand, verbs whi h are usually ategorized as eventive an be interpreted as stative predi ates when they sele t non-quantized obje ts, involving a meaning shift. ���� ������������ �������������� (362) Daria runs marathons. (363) Daria breaks windows. A third ase embellish (Derived states) onsists in a lexi al ambiguity of some verbs whi h is not triggered by synta ti and 139 environments. It is the that, all synta ti ase of verbs su h as elements being surround onstant, generate two even- tualities. (364) (365) a. Daria surrounds this b. Trees surround this a. Daria embellishes this table by means of those owers. b. Flowers embellish this table with their In this astle, with her army. astle. olors. hapter, we are interested in the last type of alternation be ause, ontrary to the other two, is not determined by external synta ti means (su h as tense). Rather, it is determined by elements whi h are internal to the lexi al stru ture of the verb itself (l-syntax or l-semanti s or on eptual module). Contrary to what has been stated or left impli it (Rappaport Hovav & Levin 1998; Harley 1995; Ram hand 1998) stative verbs are neither a homogeneous group nor aspe tually simplex (Pylkkänen 2000; Rothmayr 2006). Consequently dierent stru tures ould be asso iated to the more general label of stativity. In the next se tion, I will des ribe a group of Italian verbs whi h systemati ally parti ipate in the eventive-stative alternation. 7.3 Deadje tival Parasyntheti s I identied a homogeneous group of Italian verbs whi h share morphologi al omposition and synta ti -semanti theti � verbs formed from adje tives I am not interested in the whole behavior. (Ia obini 2004). lass of deadje tival parasyntheti I only examine those verbs whose paraphrases A, make the obje t more A, where A � I do not verbs. orrespond to make the obje t orresponds to the base adje tive. Following these morphologi al and semanti verbs (hen e DPV), among them: These are Italian parasyn- parameters, I identied 221 a e are (`to blind'), addol ire (`to sweeten', ommit myself at the moment on the adje tival or root nature of the base. It's for exposition onvenien e that, until dierently spe ied, I will all the base element adje tive. ������� �� ������������ ������������� ����� 140 `to alleviate'), irrigidire (`to stien'), sgrezzare (`to make rough'). The full list is reported in appendix. The base adje tive remains a essible in the verb semanti s, as expli itly reported in 7.1 (page 140). bello `beautiful' brutto `ugly' giallo `yellow' bian o `white' nero `bla k' grande `big' stupido `stupid' > > > > > > > > a-bell-ire `make (more) beautiful' im-brutt-ire `make (more) ugly' in-giall-ire `make (more) yellow' im-bian -are `make (more) white with an addition of white s-bian -are `make (more) white with a loss of another a-nner-ire `make (more) bla k' in-grand-ire `make (more) big' in-stupid-ire `make (more) stupid' Table 7.1: Morphologi al onstituents of DPVs. The next se tion dis usses the three morphologi al omponents of DPVs. 7.3.1 Morphologi al omponents of DPVs In this se tion, I will fo us on two re ognizable morphologi al of DPVs: the base adje tive and the prex. omponents I will provide only the formal des ription and statisti s about their distributions, leaving aside for the moment the dis ussion about the synta ti and semanti ontribution of ea h part to the whole predi ate. For ontroversies about the nature of parasynthesis, refer to Chapter 3. 7.3.1.1 Base adje tive or base root? This subse tion shows that the base element is a root and not an adje tive. We will see that the degree of the resultant state, a hieved by the ae ted obje t, is left unspe ied. If the base were a its s ale, open or ategorized adje tive, we should expe t it to lose, in the derivation. ontribute Sin e the s ale gradient of the resultant stat is not determined, I argue that the base is not a ategorized adje tive. Several synta ti -semanti tools exist in order to test whether the base is a root or an adje tive, namely: modi ation, agreement and s ale. When the base element is an adje tive, for the fa t of having been narrowed down, it is hara terized by a pre ise s ale; where s ale is dened as olor' olor' ���� ������������ �������������� a pair < S, � δ > onsisting of a set of obje ts and an asymmetri relation along some dimension � Gradable 141 δ ordering (Kennedy and M Nally 2002: 8). adje tives are divided in two groups depending on the presen e or absen e of a limit point of their s alar stru ture (Kennedy and M Nally � 2002: 9): open s ale and losed s ale . As Kennedy and M Nally (2002) point out, adje tives show dierent behavior when they are modied by degree adverbs, su h as tially. Open s ale adje tives do not o ompletely or par- ur with su h modiers (366), while losed s ale adje tives do (366), sin e they present a terminal end-point to whi h the adverb refers. (366) (367) a. ? ompletely tall/short/... b. ?partially tall/short/... a. b. ompletely full/ awake/... partially full/awake/... These adverbs, alled proportional modiers (Kennedy & M Nally 2002: 10), require adje tives that map to s ales with dened end-points or startingpoints. Intuitively, an adje tive s ale that does not possess a maximal or minimal end-point annot be modied by an adverb that works as identier of this maximal or minimal end-point. Applying this test to base adje tives of deadje ti al parasyntheti s, we see that they fall in both lasses open s ale and losed s ale adje tives, as graph 7.1 (page 142) shows. (368) ?? ompletamente freddo/muto/pesante/bian o ompletely old/dumb/heavy/white (369) ?? parzialmente ri o/ruvido/vi ino/nero partially ri h/ oarse/near/bla k (370) ompletamente fradi io/sordo/mollo ompletely soaked/deaf/weak The s ale type does not orrelate to the prex, sin e prexes distribute similarly among the two s ale types (table 7.2, page 142). Adje tives bian o, `white', and nero, `bla k', an be modied by degree adverbs only if they are impli itly referring to an extension of surfa e. this In ompletamente is better translated as the English adverb entirely, than ompletely, making evident the idea of a surfa e being modied. ase, rather � A gradable adje tive is a predi ate that takes an obje t and returns a measure of degree to whi h the obje t possesses some gradable property (Hay et al. 2002). � I keep aside the distin tion among upper bounded and lower bounded s ales. ������� �� ������������ ������������� ����� 142 Figure 7.1: Distribution of adje tives s ale (deadje tival parasyntheti s). Prex open-s ale (%) losed-s ale (%) ains- 71,19 18,64 75,70 22,43 85,71 14,29 Table 7.2: Distribution of prexes among adje tive-base (371) La asa era lasses (DPVs). ompletamente nera. The house was entirely bla k. (372) *Il pa o era ompletamente pesante. The pa kage was ompletely heavy. On the other hand, if we apply the same test to deadje tival parasyntheti s, we an see that results are not sharp. from (373) to (376) to the (373) Il ghia io ha Compare the DPVs in examples orresponding adje tives from (368) to (370). ompletamente infreddolito i bambini. The i e has ompletely got the hildren old. (374) La pioggia ha infradi iato pazialmente i panni stesi. The rain has dren hed partially the laundry. (375) Il sole ha ompletamente arrostito Giovanni. The sun has ompletely roasted John. ���� ������������ �������������� (376) 143 La vin ita al Lotto ha parzialmente arri hito Maria. The lottery win has partially enri hed Mary. In examples above, we see that no signi ant dieren e is present between verbs onstru ted on open or the verb infreddolire losed s ale adje tives. For example, in (373), freddo, ` old' infradi iare is is supposed to be built on the adje tive (368) whi h is an open s ale adje tive; in (374), the verb supposed to be built on the adje tive fradi io, `soaked' (370) whi h is a losed s ale adje tive. If DPVs were derived from ategorized adje tives, we would expe t some dieren es between those derived from open or There are two logi al possibilities to a losed s ale adje tives. ount for it: (i) adje tival s ale is not available to adverbial modi ation; (ii) the base is a root, onsequently la king s ale. In order to arm that the base is a root, two diagnosti s agreement and modi ation. an be employed: The rst one is not fully available in ase of Italian deadje tival verbs sin e the nal agreement morpheme in adje tives is systemati ally severed in verbs. (377) (378) rosso - rossa - rossi - rosse red- - red- - red- - red- m.sg. bello m.sg. beautifulbelle f.pl. f.sg. - bella m.pl. f.sg. - beautiful- f.pl. - belli - arross-ire m.pl. - beautiful- - - abbell-ire beautiful- Regarding the se ond test, we know that roots, even though they proje t their own phrase (Harley 2005; Levinson 2010), by morphemes reserved to spe i grammati al an annot be modied ategory. Deadje tival parasyntheti s are all built on non-derived bases (Ia obini 2004), this means that we (379) annot nd verbs bello - bellissimo - *abbellissimare beautiful - very beautiful (380) ontaining superlative adje tives. - making very beautiful grande - grandissimo - *ingrandissimire big - very big - making very big In addition to synta ti linguisti lues, roots are ognitive obje ts whi h be ome obje ts when narrowed down in the syntax. If we analyze languages that have a mu h more (su h as Semiti languages), dierent meanings lear denition of root an be attributed to the same ������� �� ������������ ������������� ����� 144 onsonant ombination in dierent morphologi al paradigms. For example, in Hebrew, the same root an reate a set of words whi h share only a sort of ore meaning (Arad 2003). This indi ates that the root, as possesses a non-narrowed be omes a linguisti (381) √ root: ore meaning whi h is further spe ied when it obje t, i.e., a word. btx batax a. CaCaC b. CiCCeC biteax hiCCiC hivtiax . ognitive obje t, `trust' (Arad 2003: 742 ex.5) `insure' `promise' We will see in further se tions that the division of DPVs in two dierent lasses (verbs of surfa e and verbs of form) is supported by the assumption that the base is a root. In fa t, sin e roots are not synta ti purely obje ts, but are on eptual elements, the distin tion made further relies on rather than linguisti on eptual fa ts. To summarize, the resultant state of the ae ted obje t is not dened, it ould be  ompletely A or more A be ause the predi ate is not narrowed by a ategorizer; for this reason I argue that deadje tival parasyntheti s are root-derived, √ instead of Adj. I assume that roots that feed deadje tival parasyntheti s generally form adje tives, this is why they 7.3.1.2 an often be onfused. The prex The other morphologi al building blo k of deadje tival parasyntheti s is the prex. In this sub-se tion, I will report statisti al analyses about the distri- a in /rin-, s ). bution of the three possible prexes ( , Within the analyzed 221 DPVs, prexes distribute with per entage reported in table 7.3 (page 144), and onrm statisti s reported in Ia obini (1999). Prex a in s % 28,37 61,54 10,10 Table 7.3: Per entages of prex distribution (DPVs). ���� ������������ �������������� 145 There are some pairs of verbs whi h are formed from the same root by means of two dierent prexes. The feren e in reading, as for hange of prex does not abbellire/imbellire ause any dif- (`make beautiful') and the other examples below. Very rarely it ae ts the meaning, su h as for sbian are, loosing means of putting fa t that prex (382) s- triggers blea h ), while where prex olor (as in a. s- olor. imbian are- a pro ess of making white by means of prex As mentioned in im a pro ess of making white by hapter 3, this is interpreted as a privative an be due to the s-. abbellire - imbellire make beautiful b. addol ire - indol ire sweeten . aggentilire - ingentilire make gentle d. ammiserire - immiserire make miserable/poor e. arruvidire - irruvidire make rough f. asserenare - rasserenare alm g. infreddare - rareddare ool h. sbassare - abbassare shorten/lower i. sbian are - imbian are whiten The fa t that prexes do not trigger semanti dieren es tends to sup- port the idea that they are vestiges of former Latin prepositions/prexes, whi h have gradually lost their semanti spe i traits. However, the la k of distributional dieren es does not lead to the la k of semanti s. They ontribute in making the verb ontribution to verbal ausative, sin e they are head of the relation proje tion r. It remains unexplained why parasynthesis is being repla ed by suxes like -izzare/-i are suxes whi h are repla ing parasynthesis as produ tive derivational me hanism in modern Italian (Ia obini 2004). Probably suxes izzare/i are have been preferred under the impulse of Fren h in XVIII and ������� �� ������������ ������������� ����� 146 XIX Centuries and they are more adaptable sin e they an be atta hed to derived adje tives. Produ tivity of parasynthesis is de reasing in ontemporary Italian (Ia- obini 2004) in favor of a derivative pro ess involving the sux following examples are from tre ani.it -izzare. The that re ords neologisms used in web versions of Italian newspapers. (383) lombardo - lombardizzare Lombard - to make Lombard (384) virtuale - virtualizzare virtual - to virtualize (385) illombardire (expe ted) to make Lombard (386) invirtualire (expe ted) to virtualize The derivational sux responsible for: (i) the -izzare is learly ausative, this means that it is hange in ategory of the base; (ii) the introdu - tion of a rP whi h determines the ausative meaning. Examples (385) and (386), whi h are not attested but plausible and expe ted forms, represent the parasyntheti parasyntheti x. We ounterparts of examples (383) and (384). In the ase of verbs, there are two derivational elements: a prex and a suf- an imagine that the verbal sux is responsible for the hange in ategory of the base. On the other hand, I assume that the prex is responsible for proje ting a relational proje tion (rP), whi h is responsible for the ausative meaning. The lower part of l-syntax for abbellire, `to make (more) beautiful' is given in (387). vbe (387) vbe ome P ome rP DP il bambino r a- √ bello I am aware that some issues about the mirror prin iple arise (A edoMatellan 2006: 12). This seems to be an issue for all theories of parasynthesis. ���� ������������ �������������� 147 7.3.2 Dierent types of external arguments roles In se tion 2.7, we saw that the eventuality of the lexi al verb and the eventuality of the fun tional head introdu ing the external argument must a ord in order to get a well-formed EI: dynami eventualities are onne ted with Agents and Causers and stative eventualities with Holders (Kratzer 1996: 123). involve We will see another external role for stative verbs when they ausative semanti s. In this sub-se tion, I will report a synta ti approa h to the distin tion between Agents and Causers in eventive predi ates. developed to explain DPV behavior in This will be further ase of stative reading. For this reason I will leave aside theories of underspe i ation of external argument roles � (Ram hand 2008) . Theories that pla e the external roles distin tion within syntax that Agent/Causer distin tion is not only a represented in linguisti Voi e. onsider on eptual distin tion, but it is stru ture as dierent semanti hara terizations of In the last years new eviden e in favor of a distin tion between the verbalizing little v and the introdu er of external argument Voi e has been provided (Pylkkänen 2002; Alexiadou, Anagnostopoulou and S häfer 2006; Harley 2014). External arguments of dynami predi ates � v roles: Agent and Causer . Two kinds of by Folli & Harley (2005), who do not separate for the li ensing of Agents, while v ause an hold two dierent semanti vdo heads ( v and v from Voi e. ause ) are assumed vdo is responsible for (inanimate) Causers. These two avors are related to the presen e of resultative semanti s (S häfer 2008). (388) Giovanni spazza il pavimento. John sweeps the oor. (389) *Il ume spazza l'argine. *The river sweeps the dam. (390) Il ume spazza via l'argine. The river sweeps the dam away. While v ause is asso iated with resultative semanti s, pointed out in exam- ples above by the parti le via (`away'), vdo o urs when ausative semanti s In order to establish the right event de omposition, Ram hand denes primitives responsible for the identi ation of parti ipants in events/sub-events. One of those onsists in ausation. Causation is ree ted in the argument domain with initiator role, whi h denotes an individual whose properties/behavior are responsible for the eventuality to oming into existen e (Ram hand 2008: 24). � Refer to Alexiadou & S häfer (2007) for the assimilation of Instruments to one or the other role. � ������� �� ������������ ������������� ����� 148 is absent. From a semanti perspe tive, Causers and Agents are distinguished by possibility [of the subje t℄ of generating an event on [its℄ own, from start to nish  their teleologi al apability (Higginbotham 1997), whi h is the  (Folli & Harley 2005: 200). In this work, I will put forth some pie es of eviden e whi h will diversify semanti roles of external arguments of stative predi ates. As for eventive verbs, I will assume that also for stative verbs, two types of external arguments are ne essary and ea h type (Holder/Sour e) is determined by the presen e or absen e of a resultative part. 7.4 Are inanimate subje ts a We presented DPVs as verbs that an a essible in DPVs? ept both animate and inanimate subje ts. It is worth verifying whether Italian speakers allow both types of subje ts with these verbs. This se tion reports design and results of a lexi al lling test on 55 Italian native speakers and inanimate subje ts are equally a The experiment ondu ted onrms the hypothesis that animate and essible for DPVs. ontains three parts: (i) so io-linguisti (ii) instru tions and example; (ii) linguisti The experiment was administered via IbexFarm. ea h presented in a single s reen-shot; the linguisti s reen-shot per senten e. So io-linguisti questionnaire; task. Part (i) and (ii) are part is omposed of one questionnaire asks for gender, age, edu ation and origin of parti ipants. Instru tions and example part explain the exa t task and make the linguisti ontrolled. In the linguisti register of referen e expli it: middle part, parti ipants are asked to hoose between an animate or inanimate subje t for 40 senten es. Experimental items onsist of 20 senten es built on DPVs; llers are 20 morphologi ally derived verbs. In order to avoid automati responses, llers are equally divided between pronominal verbs and transitive verbs. 7 out of 10 pronominals require an inanimate subje t. Experimental items and llers are presented in random order, produ ed by IbexFarm's internal algorithm. An example of the task is reported in the following example. (391) hanno abbellito la stanza. have made the room beautiful. � Mar o e Giulia Mar and Julie � I quadri Paintings ���� ��� ��������� �������� ���������� �� ����� Parti ipants are expli itly told to on their own opinion. 149 hoose one or both subje ts, depending Consequently, possible answers are: (i) animate and inanimate (hen eforth, ANIN); (ii) animate (hen eforth, AN); (iii) inanimate (hen eforth, IN). 55 Italian native speakers of dierent regional varieties are tested. Table 7.4 (page 149) reports the distribution of subje ts along so iologi al features of sex, age, edu ation and origin. Speakers are equally distributed for age, sex and edu ation, they are not equally distributed along regional variety, a prevalen e of Northern variety is registered. Age Sex Edu ation Origin M 22 18-26 9 College 10 North 46 F 33 26-32 35 Graduate 39 Center 3 33-40 6 PhD 6 South 6 41-60 3 60+ 2 Table 7.4: Parti ipant so iologi al features (Lexi al lling). Results show that both animate and inanimate subje ts are a even though dierent rates are observed, depending on spe i The minimum rate of ANIN answer is 18% (senten es 7 and 18, ` oarsen', and 6, rallegrare, ` rimbe illire, essible, senten e. irruvidire, `be ome stupid'), and maximal is 61% (senten e heer up'), with a global mean of 40,44%, as reported in gure 7.2 (page 150). It is worth noting that IN option, after INAN option, is the most This fa t is ounter-intuitive, but it an be explained for pragmati inanimate subje ts are highly lexi ally plausible with ea h verb. must have fa ilitated the an be as ribed to three dierent A. Informant This fa t hoi e of the reading with inanimate subje ts. Fur- thermore, some speakers have the tenden y to This hosen. reasons: hoose only one possibility. auses: onsiders only one answer orre t (wished possibility). B. Informant is not able to pass qui kly from one reading to another, then he/she marks only the most preponderant. C. Informant does not understand the methodology and marks only the rst answer he/she reads. ������� �� ������������ ������������� ����� 150 Figure 7.2: Answer means (Lexi al lling). Possibility C should be dis arded be ause of results obtained in senten es: 3, 7, 11, 13, 17, 19, where inanimate subje t has been hosen foremost even if presented later; and senten es 12, 18, where animate subje t has been foremost even if presented as se ond hosen hoi e. This test eliminates a plausible predi tion about a peripheral use of inanimate subje ts for DPVs. The fa t that inanimate subje ts are sometimes preferred to animate subje t depends on the prototypi ality of lexi al subje ts (for ee t of frequen y and prototypi ality: Gordon, Hendri k and Johnson (2004), Doyle and Levy (2008), among others ). To summarize, DPVs are equally produ tive with animate and inanimate subje ts. 7.5 Classi ation of DPVs This se tion analyses a lassi ation of DPVs based on the semanti s of the base root whi h determines whether the same lexi al entry so iated with two eventuality stru tures, stative or eventive. an be as- We will see that the property des ribed by the base root is fundamental, along with the (in)anima y of the subje t, in order to identify the eventuality of the prediate. DPVs divide into three groups, depending on the semanti s of the base: form; surfa e; psy hologi al. ���� �������������� �� ���� The rst group llarg-are, 151 � onsists of DPVs based on root of form , su h as `broaden, widen', a-ppiatt-ire, `atten', and rim-pi iol-ire, a- `make smaller'. (392) a. Giovanni ha allargato il muro. G. widened the wall. b. L'umidità ha allargato il muro. Humidity widened the wall. (393) a. Giovanni ha appiattito il us ino. G. attened the pillow. b. I ollant hanno appiattito il sedere di Giovanna. Sto kings attened Jeanna's behind. (394) a. Il sergente ha rimpi iolito il plotone. The sergeant made the squad smaller. b. Lo stu o ha rimpi iolito il bu o. The stu o made the hole smaller. The se ond group bian -are, � ima-nner-ire, onsists of DPVs based on roots of surfa e , su h as `whiten, whitewash', in-sozz-are, `dirty, tarnish', and `bla ken'. (395) a. Il pittore ha imbian ato la tela. The painter whitened the anvas. b. La pittura ha imbian ato la tela. The painting whitened the anvas. (396) a. Un delinquente ha insozzato la porta. A delinquent made the door dirty. b. Il fango ha insozzato la porta. the mud made the door dirty. (397) a. Maria ha annerito il sotto. Mary bla kened the eiling. b. Il fumo ha annerito il sotto. Smoke bla kened the eiling. The third group is based on a psy hologi al base, su h as `stun, daze', rimbe illire, `make stupid' and � This must be intended as a label. � This must be intended as a label. intristire, in-stupid-ire, `make sad'. ������� �� ������������ ������������� ����� 152 (398) a. Il professore instupidis e i suoi studenti. The professor stunned his students. b. La droga ha instupidito i ragazzi. The drug stunned boys. (399) a. Il fratello ha rimbe illito la bambina. The brother made the girl stupid. b. Il rumore ha rimbe illito i pazienti. The noise made patients stupid. (400) a. Sandra ha intristito l'ami a. Sandra made her friend sad. b. Il de esso del nonno ha intristito i nipoti. Granpa's passing made grand hildren sad. In the rst group, the hange is physi al sin e the base root denotes a ore quality of an individual. In these verbs, the subje t undergoes a hange of one of its dimensions, of its inner properties. On the other hand, verbs of surfa e involve a the individual itself. hange that is external to For example, a wall does not see its inner properties hanged if it is painted red: if it was 2 feet high it remains 2 feet high, if it was 1 in h deep, its depth remains un hanged. However, a wall does its inner properties if it is widened. One an without even tou hing the obje t, but one obje t without hange the annot olor of an obje t hange the shape of an hanging the obje t itself. I will assume that the presen e or the la k of a of inner properties physi al hange ontributes to the o hange is produ ed, the verb there is no physi al hange involving readings are possible, whi h are Δ (delta, i.e. a urren e of two hange) aktionsarten. If a an only have an eventive reading. If onstitutive parts of the Theme, two learly ree ted by the (in)anima y of the subje t. In se tion 7.8.2.1, I will analyse the nature of hange and produ e its denition. We have seen that DPVs are divided into three groups (surfa e, form, psy hologi al) and we will investigate the rst two. ture resear h psy hologi al predi ates, sin e they independent group whi h shows spe i literature (Belletti & Rizzi 1988; We will see in the next We will leave for fu- onstitute a pe uliar and properties, as often shown in the inter al.). hapter that a systemati dieren e in eventuality is mostly produ ed when an inanimate subje t appears as external argument of DPVs of surfa e, making the senten e stative. On the other hand, DPVs of form are interpreted as eventive independently on the (in)anima y of the subje t. ���� ��������� ������������� �� ���� 7.6 153 Dierent eventualities in DPVs This se tion reports eviden e about the two possible readings of DPVs of surfa e, on the one hand, and the unique eventive reading for DPVs of form, on the other hand. We will resort to four tools: the epistemi under dovere, interpretation under bution and adjun ts (refer to 7.6.1 già interpretation `already', temporal narrative ontri- hapter 6). Interpretation of dovere This subse tion presents data useful to the identi ation of two eventualities expressed by DPVs of surfa e with modal The dovere omplex dovere, `must'. an generate two interpretations: deonti /obligational and epistemi . The deonti /obligational reading on erns the future and expresses a mand about an a tion that must be realized. The epistemi on erns a spe ulation about a present state of aairs. related to the eventuality of the predi ate. only a deonti 7.6.1.1 interpretation Interpretations are Eventive predi ates reading, while stative predi ates om- an re eive an re eive both. Verbs of form I have already shown that verbs of form do not generate a stative reading, sin e they involve a Δ in inner physi al properties of the Theme. In this subse tion we will see that DPVs of form do not generate a stative reading in subje ts. verbs. onjun tion with dovere, neither with animate nor with inanimate I start by presenting eviden e about future In order to highlight it, I will resort to adverbial onstraint of these entro domani, `by tomorrow'. (401) a. Giovanni deve allargare il muro entro domani an hé il lavoro sia nito. G. must widen the wall by tomorrow, in order to get the job nished. b. L'umidità deve allargare il muro entro domani an hé il lavoro sia nito. Humidity must widen the wall by tomorrow, in order to get the job nished. (402) a. Giovanni deve appiattire il us ino entro inque minuti per an- dare a letto. G. must atten the pillow in ve minutes in order to go to bed. ������� �� ������������ ������������� ����� 154 b. I ollant devono appiattire il sedere di Giovanna in un'ora an hé possa andare alla festa. Sto kings must atten Jeanna's behind within one hour so that she an go to the party. (403) a. Il sergente deve rimpi iolire il plotone in tre giorni per parte i- pare all'eser itazione. The sergeant must redu e the squad in three days in order to parti ipate in the training. b. Lo stu o deve rimpi iolire il bu o in un minuto an hé il la- voro sia nito. The stu o must redu e the hole within one minute in order to get the job nished. We saw that senten es involving DPVs of form entail a future independently of the (in)anima y of the subje t. onstraint, Both animate (Giovanni and the sergeant) and inanimate (humidity, sto kings and stu o) give rise to a tions whi h must take pla e in the future in order to get the statement true. 7.6.1.2 Verbs of surfa e DPVs of surfa e give rise to two readings, whi h are made evident by the (in)anima y of the subje ts A �� . ordingly, the modal verb subje t is animate a future dovere yields dierent interpretations. If the onstraint arises, if inanimate a present onstraint. �� However, animate subje ts an be interpreted as the orrespondent inanimate subje ts and serve as subje t to a stative predi ate when they are not Agents but Sour es. Then in the next se tions I will refer to animate subje ts uniformly as Agents. Animate subje ts an generate a stative reading of DPVs of surfa e whenever they are read as inanimate. (1) Daria abbellis e la foto. Daria embellishes the pi tures. a. Daria makes the pi ture beautiful by painting it. b. Daria makes the pi tures beautiful by her presen e on it. Inanimate subje ts an generate a stative reading of DPVs of form whenever there is no physi al hange in the Theme. (2) Il divano ingrandis e la stanza (se ondo Daria). The sofa enlarges the room (in Daria's opinion. ���� ��������� ������������� �� ���� entro /in I resort to adverbial ` X-time' to make the reading ten es with animate subje ts are a (404) a. lear. Sen- eptable, while the ones with inanimate subje ts are not. However, the purpose if intended as 155 lause is a eptable in b. examples must nish to. Il pittore deve imbian are la tela entro domani per nire il lavoro. The painter must whiten the anvas until tomorrow in order to nish the job. b. *La pittura deve imbian are la tela entro domani per nire il lavoro. The painting must whiten the anvas until tomorrow in order to nish the job. (405) a. Il delinquente deve insozzare la porta entro due minuti an hé il lavoro sia nito. The delinquent must make the door dirty until two minutes so that the work is done. b. *Il fango deve insozzare la porta entro sabato an hé il lavoro sia nito. The mud must make the door dirty until Saturday so that the work is done. (406) a. Maria deve annerire il sotto entro sabato prossimo. Mary must bla ken the eiling by next Saturday. b. *Il fumo deve annerire il sotto entro sabato prossimo. Smoke must bla ken the eiling by next Saturday. Consider the following ontext: It has been long time I haven't ome to Giulia's. However, I remember the disposition of the furniture and the olor of the walls. When I ame in today I saw something dierent and I said: (407) Il muro è nero! La verni e deve s urirlo. The wall is bla k! Paint must make it dark. At the moment of utteran e of (407), the wall is dark, the speaker states his/her surprise for this state of aairs. This means that the painting must have darkened the wall before the moment of utteran e, no future is involved. Furthermore, it easily re eives an epistemi is not sure about the a tual ause of the wall state. onstraint reading if the speaker ������� �� ������������ ������������� ����� 156 On the other hand, a ates a misunderstanding. ording to the same ontext, senten e (408) gener- In fa t, it is interpretable only if John is making the wall darker by means of his body (i.e. hanging on the wall). (408) Il muro è nero. # Giovanni deve s urirlo! The wall is bla k! John must be making it dark. These readings arise be ause eventive predi ates under a future onstraint whi h is in ompatible with the dovere reated generate ontext. At the moment of utteran e the state of aairs expressed by the statement is already present. A ording to the same a future (409) ontext, we an produ e a senten e whi h generates onstraint, and entails an eventive reading, su h as (409). Il muro è bian hissimo! Giovanni deve s urirlo. The wall is bone-white. John must make it dark. In (409), the dovere omplex verb is feli itous be ause it generates a future reading of the verb and this is not in bian hissimo. Bone-white and onstraint of dovere is at work. adje tive future ontrast with the frame-senten e dark are in ontrast and thus the We see that DPVs of surfa e generate two readings under the modal `dovere': (i) in presen e of an inanimate subje t they involve a present straint; (ii) in presen e of animate subje t they generate a future A on- onstraint. ording to previous assumptions, DPVs of surfa e with inanimate subje ts are stative, while with animate subje ts are (mostly) eventive. In this se tion, we analyze the behavior of DPVs of both fa e and form) under modal dovere, tations as expe ted. DPVs of surfa e ategories (sur- `must', whi h shows dierent interprean trigger both deonti and epistemi reading, thus they are stative. DPVs of form generate only deonti reading, thus they are eventive. 7.6.2 Interpretation under già. It has been noted by Mittwo h (2014) that the adverb `already' an ombine only with derived and lexi al statives, su h as progressives and perfe ts. (410) a. Daria orre già. Daria already runs. b. *Daria orre già la Maratona di NY del 2016. *Daria already runs the NY Marathon 2016. ���� ��������� ������������� �� ���� . Daria sta già 157 orrendo la Maratona di NY del 2016. Daria is already running. d. Daria ha già orso la Maratona di NY del 2016. Daria has already run the NY Marathon 2016. In the following se tions I will apply this test to dierent ategories of DPVs. 7.6.2.1 DPVs of form In this subse tion, I will explore the behavior of DPVs of form when used with già, `already'. Senten es below report DPVs of form with animate and inaminate subje ts. (411) a. *Giovanni allarga già il bu o del salotto. G. already widens the hole in the living room. b. *L'umidità allarga già il bu o della u ina. Humidity already widens the hole in the kit hen. (412) a. *Il sergente rimpi iolis e già il plotone della sesta armata. The sergeant already redu es the Sixth regiment squad. b. *Lo stu o rimpi iolis e già il bu o del muro del salotto. The stu o already redu es the hole in the living room wall. Examples (411) and (412), where the presen e of a quantized obje t prevents the possibility of interpreting them as habituals, onrm that DPVs of form are eventive with both types of subje ts. 7.6.2.2 DPVs of surfa e With DPVs of surfa e, the (in)anima y of the subje t is a ree t of the eventuality of the senten e. a Consequently, with già we should nd dieren e in eptability of senten es whi h are linked to the (in)anima y of the subje t. (413) a. ??Il pittore imbian a già la tela del Caravaggio. The painter already whitens Caravaggio's anvas. b. La pittura imbian a già la tela del Caravaggio. The painting already whitens Caravaggio's anvas. (414) a. ??Un delinquente insozza già la porta del ivi o 33. A delinquent already makes the 33rd door dirty. b. Il fango insozzato già la porta. The mud already makes the 33rd door dirty. ������� �� ������������ ������������� ����� 158 Senten es with animate subje ts are not a eptable with già �� . This se tion analyses dierent behaviors of DPVs in relation with their semanti s and the adverb già. statives are grammati al in This test The test reveals that only lexi al and derived onjun tion with già, while eventives are not. onrms our hypothesis that DPVs of form are always eventive. 7.6.3 Temporal narrative ontribution This se tion presents the role of DPVs of form and DPVs of surfa e in building temporal hains in a narrative dis ourse. It is a well known property of statives (Dry 1983; Katz 2003) that they do not ontribute to the temporal progress of a narrative dis ourse (415), ontrary to eventive verbs (416). (415) Mary arrived. Her daughter was sitting and her dog was sleeping. (416) Mary arrived. Her daughter sat down on the ou h and her dog fell asleep. In the following subse tion I will apply this test to DPVs' 7.6.3.1 ategories. DPVs of form DPVs of form ontribute to the narration progress, this means that they are interpreted as being part of a hain of onse utive events whi h take pla e one after the other. Examples below point out that DPV of form ingrandire, `in rease', on- tribute to the progress of the narration. In example (417) Daria rst arrives, then makes the hole bigger and then sits down. In example (418), the mold sprang, then made the hole bigger and then died. (417) Daria è arrivata, ha ingrandito il bu o e si è seduta sul divano. Daria arrived, (she) made the hole bigger and (she) sat on the ou h. (418) La mua si è formata, ha ingrandito il bu o ed è morta. The mold formed, (it) made the hole bigger and died. 7.6.3.2 DPVs of surfa e In this subse tion we analyze the temporal ontribution of DPVs of surfa e to the narrative progress. We will see that they inuen e the narrative progress when the subje t is animate, but fail to do so when the subje t is inanimate. �� It is worth noting again that the dire t obje t must be quantized in order to generate an eventive reading, otherwise it generates an habitual reading whi h is stative. ���� ��������� ������������� �� ���� (419) 159 Daria è arrivata, ha imbian ato la tela del Caravaggio e si è seduta sul divano. Daria arrived, whitened Caravaggio's anvas and sat down on the ou h. (420) La verni e è stata stesa, ha imbian ato il muro e ha s hiarito la stanza. The painting was painted, whitened the wall and brightened the room. In (419), a narrative with her sitting on the hain whi h starts with Daria's arrival and nishes ou h is des ribed. Example (420), on the other hand, does not entertain a narrative The other two verbs do not hain; there is only one event: the painting. ontribute to narration progress. 7.6.4 Adjun ts Animate subje ts (of both surfa e and form DPVs) an o ur with adjun ts denoting instruments, i.e. individuals whi h belong neither to the obje t nor to the subje t. (421) Il bambino rallegra la festa on i pallon ini. The hild lightens up the party with balloons. (422) Giovanni s hiaris e il té on il limone. John makes the tea learer with lemon. With inanimate subje ts the pi ture appears more mate subje ts with DPVs of surfa e a ept on -`with' ompli ated. Inani- adjun ts, albeit with some restri tions, while DPVs of form do not. Senten es built on DPVs of surfa e with inanimate subje ts tain adjun ts denoting independent instruments, they annot on- an only denote a proper part of the subje t. In other words, individual denoted by the adjun t and the individual denoted by the subje t are in an inalienable possession relationship. This fa t is pointed out by the possessive adje tive present in adjun ts. appear in Senten es with animate subje ts, built on DPVs of surfa e, an ombination with adjun ts lexi alizing the inalienable possession relationship between the subje t and the inner ause. In this ase, however, the senten e re eives a stative reading, the animate subje t being treated as an inanimate. (423) Giovanni imbian a la stanza on il suo sorriso. G. whitens the room with his smile. (424) Giovanni insozza l'atmosfera on il suo muso. G. dirties the atmosphere with his fa e. ������� �� ������������ ������������� ����� 160 (425) La musi a rallegra la festa on il suo ritmo in alzante/* on lo stereo. The musi lightens up the party with its insistent pulse/*with the stereo. I suspe t that the restri tion is due to the fa t that subje ts in (423) to (425) an ontrol se ondary tools (Nielsen 1973; S hlesinger 1989) and annot ontrol instruments, unless they are in an inalienable possession relationship with them. Senten es built on DPVs of form with inanimate subje ts annot ontain adjun ts denoting properties or parts responsible for the eventuality in an on -adjun (426) a. t, but in a ausa di -`be ause-of ' adjun ts. ??La mua ha allargato il muro on le sue spore. The mold enlarged the wall with its spores. b. La mua ha allargato il muro a ausa delle (sue) spore. The mold enlarged the wall be ause of its spores. (427) a. ??La nebbia ha allungato la rotta on la sua densità. *The fog lengthened the route with its density. b. La nebbia ha allungato la rotta a ausa della (sua) densità. The for lengthened the route be ause of its density. In example (427), the fog would be per eived as being voluntarily thi k. The subje t is an agent, but it is still inanimate and annot have full ontrol on other instruments. 7.6.5 To sum up This se tion presented dierent behavior of DPVs depending on the semanti s of the base, whether of surfa e or of form. Their interpretation under their interpretation under già, their ontribution to the narrative dovere, hain and possible adjun ts. The (in)anima y of the su je t pf DPVs of surfa e helps in highlighting dierent readings. A summary of stativity tests results whi h have been dis ussed is reported in table 7.5 (page 161). ���� �� ������������� ��������� ���� 161 Dovere Già Temp. Contr. Adjun ts Animate deonti * � on, instruments Inanimate deonti * � a ausa, instruments DPVs of surfa e Animate deonti * � on, instruments Inanimate epistemi � on, not instruments Table 7.5: Re ap of stativity tests results (DPV). DPVs of form ������� �� ������������ ������������� ����� 162 I argue that DPVs of form are eventive, while DPVs of surfa e alternate between a stative and an eventive readings. This alternation is made expli it by the (in)anima y of the subje t, as reported in table 7.6 (page 162). DPVs of form DPVs of surfa e Animate eventive eventive Inanimate eventive stative Table 7.6: Eventualities of DPVs. We will see that subje ts of DPVs of form have tenden ies to do, to a t, thus they an ombine with a dynami /energeti stru ture. Subje ts of DPVs of surfa e with stative reading have tenden ies to be. We will point out that the two argument stru tures dier in one point: the presen e of energeti DPVs are energeti for e (Copley & Harley 2015) in little ausatives. Stative DPVs are stati In next sessions, we will dis uss the v. Eventive ausatives. ausative nature of both DPVs types in details. We will see that DPVs have a pe uliar property that dierentiate them from other ausative statives, su h as Obje t Experien er Psy hologi al verbs (Pylkkänen 2000), due to the presen e of personal judge parameter (Laherson 2005; Stephenson 2007). 7.7 Are all DPVs ausative? In this se tion, I will The semanti Tovena (2012) �� onsider whether all DPVs present ausative semanti s. role of Roman e prexes has been investigated by Martin & . It is worth noting that Roman e languages do not present a produ tive prex system of predi ates, like Slavi apable to inuen e lexi al and grammati al aspe ts languages. 7.7.1 DPVs of form DPVs of form are shown to be eventive. issues whi h need to be a Therefore, there are not spe i ounted for, sin e eventive ausative verbs do not pose problems in any theory of verbal lexi al semanti s (Copley & Harley 2015; Ram hand 2008; Borer 2005). �� They analyze deadje tival Fren h verbs and investigate dierent semanti s linked to dierent derivational morphologi al means in asso iation with one single adje tival base. Fren h possesses dierent ways of deriving a verb from an adje tive: sux -izer /-ier or prex en-/an-. ���� ��� ��� ���� ���������� 163 Analyzing the following senten es by means of we see that they are usual (428) a. ases of ausative verbs. → Giovanni ha allargato il bu o. ausare il fatto orrespondent paraphrases, G. ha fatto qual osa per he il bu o sia più largo di prima. G. widened the wall. → G. did something to ause that the hole is larger. b. L'umidità ha allargato il muro. per ausare il fatto → L'umidità ha fatto qual osa he il bu o sia largo. Moisture widened the wall. → Moisture did something to ause that the hole is large. (429) a. Il sergente ha rimpi qual osa per → iolito il plotone. ausare il fatto Il sergente ha fatto he il plotone sia più pi olo di prima. The sergeant redu ed the squad. → The sergeant did something to ause that the squad is smaller. b. Lo stu per o ha rimpi ausare il fatto → Lo stu iolito il bu o. he il bu o sia pi o ha fatto qual osa olo. The stu o redu ed the hole. → The stu o did something to ause that the hole is smaller. Paraphrases are parti ularly interesting be ause they on eptual parts into whi h we an suggest dierent an (informally) divide the event des ribed by the predi ate. In ase of DPVs of form, we see that the rst of a dynami on ept, do something. on eptual part The subje t a tion whi h leads to the result. In (428a), we onsists arries out an undened an imagine that the subje t performs an a tion of demolition or an a tion of renovation whi h auses the moisture (although result. The same way, in (428b), we inanimate) performs an a tion that an presume that auses the result of being rot of the Theme. Even though I argue that both environments are ausative and eventive, two distin tions must be drawn between animate and inanimate subje ts whi h both seem to depend on world-knowledge. First, the smaller range of possible a tions performed by inanimate subje ts inuen es plausibility. Clearly, the moisture - ontrary to John- annot perform an a tion su h as "hammering" in order to get the result. Animate subje ts, by their intrinsi �� nature , an perform a large set of dierent a tions. Se ond, animate sub- �� When I use the term "intrinsi nature" I refer to our ommon knowledge about the world and the handling power of individuals. ������� �� ������������ ������������� ����� 164 je ts, ontrary to inanimate, subje ts ause an exer ise ontrol over the a tions. Inanimate ir umstan es without will and without ontrol. The presen e of prex and the behavior within periphrases lead to the on lusion that DPVs of form are ausative. 7.7.2 DPVs of surfa e DPVs of surfa e have been shown to be ambiguous between an eventive and a stative reading. This ambiguity is made expli it by their subje t's anima y. While inanimate subje ts ex lusively generate a stative reading, animate subje ts mostly generate an eventive reading. Resorting to paraphrases, we will see that the rst on eptual part hanges in relation to subje t's anima y. (430) a. Il pittore ha imbian ato la tela. per ausare il fatto → Il pittore ha fatto qual osa he la tela sia (più) bian a. The painter whitened the anvas. → The painter did something to ause that the anvas is (more) white. b. → L'esistenza della verni La pittura ha imbian ato la tela. tela ha ausato il fatto e sulla he la tela sia bian a. The painting whitened the anvas. → The existen e of the painting on the anvas aused that the anvas is white. (431) a. Un delinquente ha insozzato la porta. fatto qual osa per ausare il fatto → Un delinquente ha he la porta sia (più) spor a. A delinquent made the door dirty. → A delinquent did something to ause that the door is dirtier. b. Il fango ha insozzato la porta. ha ausato il fatto → L'esistenza del fango sulla porta he la porta sia spor a. The mud made the door dirty. → The existen e of the mud on the door aused that the door is dirty. These paraphrases dier ausing ir umstan es. onsiderably in their rst part, whi h is about Animate subje ts perform a tions, they do some- thing, and these dynami events provoke the result to ome into existen e. On the other hand, inanimate subje ts do not perform a tions, they do not parti ipate in dynami events. One should then ask how they an ause a result. Inanimate subje ts of DPVs of surfa e are in a parti ular state that is pereived by the speaker as the immediate verbs ause for the result. While eventive an be paraphrased by the subje t has made be ause an eventive part is responsible for the inner ash of for e in the system, stative verbs (whi h ���� ������ �������� 165 by denition are not energeti ) annot be so paraphrased by subje t has done, sin e no for e is introdu ed in the system (se tion 7.8.1 for details). The ausative part of paraphrases above is onstituted by  ause.... The fa t that this type of paraphrase is allowed for both animate and inanimate subje ts suggests that all senten es are ausal, with no distin tion to their eventuality. DPVs of surfa e are ausative, as morphologi ally shown by prexes and paraphrases. 7.7.3 To sum up DPVs of form do not present parti ular issues about their Prexes and paraphrases orroborate this ausative nature. on lusion. Eviden e from prexes and paraphrases prove that DPVs of surfa e are ausative. However two kinds of and a stati ausation. Stati ausation seem to be at stake: a dynami ausation presents some puzzles. I will on- sider it in se tion 7.8, whi h reports previous studies about the existen e of ausative reading among 7.8 ertain types of stative verbs. Causal relation Human languages systemati ally employ dierent means in order to dis riminate between ausatives and non ausatives s enarios. Some languages resort to dedi ated morphologi al means, su h as the presen e of ausative axes within the verbal part (433) (Walla e 1981); other languages resort to synta ti means su h as periphrasti to expli it morpho-synta ti Dierently put, we spe i reate means (437). an nd languages that express ausation by means of spe ial ausation by means of mã onstru tions. Usually, in the literature, ausation- reating strategies are identied, syntheti or periphrasti ausation. kan gar- hu. prs.1sg. (Nepali) 1sg. work do- I do the work. (433) ??); some others do not resort morphemes in the verbal domain or other languages that are able to two dierent (432) ausatives ( mã 1sg. gar-au- hu work do- aus-prs.1sg. kam I have the work done. (Nepali) ausation ������� �� ������������ ������������� ����� 166 (434) Daria mangia Daria eat- una 3.sg. det.f.sg. mela. apple. Daria eats an apple. (435) Daria fa mangiare 3.sg. Daria make- una eat- mela inf. det.f.sg. (a Maria). apple (to Mary). Daria feeds Mary with an apple. (436) John eat pizza. (437) John feeds Mary with pizza. The expression of ausation is not independent from the wider dis ussion about argument realization, sin e adding ausative semanti s generates some hanges in argument pattern as it is visible in examples above. For example, a dieren e in argument patterns of (436) and (437) is lear sin e the internal obje t is Goal in the previous example and an Experien er in the latter. We will see that this is due to a general pattern about As we already dis ussed in ausal hapter 2, linguisti hains (Wol 2007). theory has investigated how human beings lexi alize dierent real-world events; that is, how the parti ipants in an event are expressed as arguments of a verb. ferent approa hes theoreti al Many dif- an be identied, depending on the number and kind of onstru tions used. In this se tion, we will fo us on a for e dynami approa h to ausation (Talmy 1985a, 1985b, 1988; Croft 1998, 2012; Copley & Harley 2015; Copley & Wol 2014b), sin e it an su unexplained, we will see that it essfully solve some puzzles otherwise an be extended to a ount for ambiguous deadje tival verbs with two eventive readings, namely DPVs of (438) olor. Giovanni abbellis e la stanza ( on i quadri). John embellishes the room (with pi tures). (439) Le foto abbellis ono la stanza ( on i loro olori). Pi tures embellish the room (with their olors). A ording to eviden e presented in previous se tions, senten es (438) and (439) dier in their eventuality, the former is eventive and the latter is stative. We have also seen that ausal stative reading the subje t hains they represent are dierent sin e in the annot ontrol an external instrument argument, but only an inherent possessed part. (440) Il bambino rallegra la festa on i pallon ini. The hild lightens up the party with balloons. ���� ������ �������� (441) 167 La musi a rallegra la festa on il suo ritmo in alzante/* on lo stereo. The musi lightens up the party with its insistent pulse/*with the stereo. It is worth noting that for e-dynami linguisti s framework, but that it approa h was born in a ognitive an be easily translated into a more formal approa h to language (Copley & Harley 2015; Copley & Wol 2014; Copley 2015). In fa t, this approa h seems to have identied the blo ks responsible for dieren es in ognitive building ausation expressions, and these building blo ks seem to be dis riminated in language expressions too. The fa t that a on ept is eviden e dull ognitively dis riminated from others does not per se olors are for its linguisti onstitute an importan e: for example, vivid olors and ognitively dis riminated, but linguisti ally they are not (at least in English and Italian). However, when a on ept is dis riminated both ognitively and linguisti ally, it is worth investigating it. The most widely dis ussed theory of terfa tual, ausation in linguisti s, has been proposed by Lewis (1973). ausation belongs to the more general ommon denominator of these theories alled lass of dependen y theories. onsists in the fa t that A ontexts; i.e. ontexts in whi h a possible The auses B i B depends on A in some sense. These theories have problems in the of emption oun- Counterfa tual theory of ase ause is not the real ause due to an emption event, introdu ing another possible auser (early pre-emption). In order to a ount for ausative stative verbs, I will propose a small extension to the analyses proposed by Copley & Harley (2015). larly, I will assume that eventive energeti energeti Parti u- ausation is involved whenever an for e enters the system, as expe ted. On the other hand, stative ausation arises when the system does involve only a virtual for e ( alled abdu tion) introdu ed by the speaker (in DPVs) who is responsible for establishing the ausal link between individuals, between Sour e In the next se tion, I introdu e the approa h to �� and Theme. onstitutive parts of the for e-dynami ausation. 7.8.1 For e-dynami approa h For e-dynami approa h to ausation is rooted in ognitive linguisti s, par- ti ularly in Talmy (1976, 1985, 1988, 2000) and Croft (1991, 2012). Here, Sour e is the role of external arguments of stative ausatives. In Copley & Harley's terminology Sour e is applied to all external arguments whi h are responsible for introdu ing energeti for e in the system. �� ������� �� ������������ ������������� ����� 168 This approa h has been on eived in order to provide explanation of dif- ferent patterns of argument stru ture realization; parti ularly to nd whi h ognitive prin iple regulates them. been assumed to reside in the parti ipants of an event. tween parti ipants; It The ognitive prin iple responsible has ausal stru ture of events whi h links the an be dened as the transmission of for e be- ausation in a for e-dynami approa h is an asymmetri intera tion between entities (Croft 2012: 198). Talmy (1972, 1976) identies four kinds of ausation hains, a ording to the physi al or mental nature of the two entities involved, namely initiator and endpoint (Croft 1991: 166): (i) physi al ausation hara terized by a physi al obje t a ting on another physi al obje t (physi al initiatorphysi al endpoint); (ii) volitional ausation where a volitional entity a ts on a physi al obje t (mental initiator-physi al endpoint); (iii) ae tive tion ausa- hara terized by a physi al obje t a ting on a volitional entity (physi al initiator-mental endpoint); (iv) indu tive ausation where a volitional entity a ts on a volitional entity ae ting her mental state (mental initiator-mental endpoint). The entities involved, both physi al or volitional, have a parti ular for e tenden y (Talmy 1998, 2000). They tenden y to stasis. the ball, an have the tenden y to motion or the This means that in a state of aairs the ball has a tenden y to fall whi h is �� like John stops ontrary to the tenden y of John to a t on the ball. The event produ ed is the result of the addition of the two for es brought about by parti ipants, the same way as in physi s the ve tor sum of for es is responsible for equilibrium. Psy hologi al physi alist theories of dynami s, share some basi ausation, to whi h belongs for e- assumptions, su h as the hypothesis that the ausal nature of an intera tion is due to internal fa tors (Wol 2007: 85). Considering mines the  ausal relationships as physi al deterministi intera tions deter- lo al level of granularity on the analysis  (ibidem ), follows that, when two events are not temporally from this it ontiguous, a linking ausal hain must be assumed. Translated in a linguisti theory, this means that a verb, in a parti ular argument realization pattern, has a spe i part of the ausal hain, for example in the �� The term onsists in the hain ( ausal segment) it represents (Croft 2012: 205-206; Ram hand 2008). Prepositions too prole is able to a verbal prole that an prole ausal segment of the ausal ase of oblique arguments or benefa tives. Verbal ount for argument realizations patterns, establishing a situation assumes a spe i meaning in the for e-dynami approa h. For this reason, I am not using it in ontexts whi h require it. I will resort to state of aairs when I want to refer in a naive sense to situation. ���� ������ �������� 169 link between role designation and realization and verbal semanti s, by means of a relatively small range of rules, reported in 442 from Croft (1998a: 24). These linking rules: apply to any an a ausal hain; are ross-linguisti ally valid; ount for oblique arguments. (442) a. The verbal prole is delimited by Subje t and Obje t (if any); b. Subje t is ante edent to Obje t in the ausal obj . hain: subj → An ante edent oblique is anteedent to the obje t in the ausal hain, a subsequent oblique is subsequent to the obje t in the ausal d. hain: a. obl. → obj → s. obl. In orporated arguments are between subje t and obje t in the ausal hain: subj → in orp. → obj However, these linking rules are valid only for those verbs that are nonneutral for e-dynami ally, i.e. for verbs that involve a verbs with ausative meaning remain una ausal hain. Stative ounted for in this approa h (Croft 2012: 235). The formal linguisti power of for e-dynami approa h onsists in the keep, that are hardly fruitful analysis of maintaining verbs, su h as stay a These verbs are eventive, as ountable for in an event-based approa h. or shown by their well-formedness in the progressive; they are also something events auses something else to be/do), yet they ausal (sin e annot be des ribed as ausing events. (443) John keeps the door open. (444) John is keeping the door open. In example (443) there is no a t in fa t even without movement John ondu ted by the subje t on the door, would still be keeping the door open. Thus, verbs of maintaining do not involve a tions, but are eventive sin e they allow progressive forms. If we resort to for es, we In the ase of keep, an easily unify the analysis of eventive verbs. for example, a for e introdu ed by the subje t is a ting ontrary to the disposition of the obje t: losed, John the door has a tenden y to be applies an opposite and stronger for e, with the result that the door is kept open. 7.8.1.1 Denitions in a formal framework (Copley & Harley, 2015) The for e-dynami its roots in approa h to event-stru ture and argument realization has ognitive linguisti s. In these last years, some resear hers imple- ������� �� ������������ ������������� ����� 170 mented this approa h in a formal syntax-semanti framework, parti ularly Copley & Harley (2015), Copley (2015) and Copley & Martin (2014). In order to apply this model in a formal theory of syntax, we must assume that for es are linguisti language, and not only elements, whose presen e is dis riminated by the on eptual elements. These pie es of eviden e of ausation in pure ome from a produ tive ex hange between theory ognitive dis iplines and linguisti dis iplines. able to show that there is a re urrent and regular link between types of ausation and linguisti a knowledge that linguisti part from our If we riphrasti ognitive per eption of ognitive representation of the We ausation should follow at least in ausal lassi ation of onstru tions, they ognitive means implied in their expression. expression of onsider the usual Ea h is hains. ausatives in lexi al and pe- orrespond to a dieren e in the (in)dire t ausal hain (Fodor 1970; Cruse 1972; Shi- batani 1976; Smith 1970). In the following examples, (445) denes a dire t ausation hain in whi h the subje t must himself open the door; while (446) denes an indire t ausation hain in whi h the subje t must provoke some- thing/someone else to open the door. (445) John opened the door. (446) John made the door open. This statement has been put to experiment by Wol (2003) by means of a 3D reality models verbally des ribed by English native speakers. The in both dire t and indire t ausation [...℄ an entity an be viewed as an intermediary only if it is fully independent of the auser and ausee , (Ibidem : 6). This means that when the ausative relationship experiment shows that  is mediated, it is linguisti ally dierentiated with dierent stru tures. The formalization of the for e-dynami (2015) shares with the model made by Copley & Harley ognitive for e-dynami on epts that nds a linguisti reex. approa h the set of basi Furthermore, it has the virtue of requiring a very small number of denitions. There are two main obje ts, for e and situation, from whi h all for edynami A event stru tures an be derived. linguisti For e (f ) is spatially and temporarily lo ated and it arises a fun tion from an initial linguisti situation s to the ( eteris paribus, linguisti ) nal situation s', whi h orresponds to a on eptual net for e ϕ. The latter is a (mental representation of) an input of energy that arises from all the individuals and their property attributions in a on eptual situation σ  from individuals in the situation and their properties. It is dened as:  ���� ������ �������� 171 (Copley & Harley 2015: 15). It is a fun tion of type hs, si, from situation to situation. A linguisti Situation (s) is formed by obje ts and their properties (Barwise & Perry 1983: 7 .), is delimited by the speaker in her/his linguisti a t and it is primary lo ated in spa e and time. It is dened as  orrespond[ing℄ to a on eptual situation σ , whi h is a spatio-temporally bounded annotated snapshot of individuals and their property attributions  (Copley & Harley 2015: 14). It is of type situation (hsi). For es and situations are building blo ks of a for e dynami ausation and they are related to ea h other in a ausal approa h to hain. From these building blo ks with a small number of denitions, all the for e-dynami approa h to ausation omes alive. The net for e (fn or net(f )) is the sole and unique for e that arises from a spe i situation, i.e. it arises from all the individuals and their properties in that parti ular situation (by denition). (447) net(f ) =: net f orce of s Applying the inverse of the net for e fun tion (net−1 ) we an derive the initial and the nal situation. The initial situation (init(f )) is the situation of whi h f is net for e. (448) init(f ) = net−1 (f ) The nal situation (f in(f )) is the situation that results when f takes s as its argument, i.e. it is the situation that results when net for e applies to s. (449) f in(f ) = f (net−1 (f )) The su essor situation of s (suc(s)) is the situation that results when the net for e takes s as its argument. (450) suc(s) = f in(net(s)) Prede essor situation of s (pred(s)) is the situation of whi h s is su situation. (451) essor pred(s) = suc−1 (s) A situation is e a ious when no external for e intervenes, in a (`all the rest being equal') ase. The opposite ase is the ase eteris non in whi h a non-attended for e intervenes. paribus eteris paribus (`all the rest not being equal'), ������� �� ������������ ������������� ����� 172 Copley & Harley (2015), with the formal apparatus in pla e, propose a representation for the main eventuality types from a for e-theoreti point of view. They adopt ore assumptions of the general approa h to argument and event stru ture, parti ularly the one whi h sees the VP synta ti ally de omposed into dierent phrasal levels, isomorphi the verb, dominated by a vP node ( ibid : to the eventuality stru ture of 18). Copley & Harley (2015) begin with the analysis of the of ausative-in hoative alternation, basi one. ommon lass onsidering the in hoative form as the Usually, these verbs are treated in the literature as having two subevents: a ausing eventive subevent and a stative result subevent. fa t is pointed out by dierent s opes of (452) Daria is (453) Daria is again again This adverb. losing the door again. losing the door. In example (452), the adverb takes low s ope over the resultative subevent, originally over a SC. Thus, losed. Daria is losing a door whi h has been previously In example (453), the adverb takes high s ope over the subevent, namely over for e. The orresponding reading is that ausative Daria is losing the door another time. Therefore, ausative verbs synta ti ally involve at least two phrasal pro- je tions. In for e-dynami approa h, the whi h is applied to a situation. not to hold, ausing subevent is repla ed by a for e The for e is responsible for the situation onsequently yielding a situation where the result state holds (Dowty 1979). In other words, a for e applies to a situation where the result state does not hold, this for e eteris paribus yields the result state. Synta ti ally, the result state is represented by a SC (Harley 2005; Ramhand 2008) in the lower part of the verbal proje tion; being a state (a situation), it is a predi ate of situations (type <s,t>). Upwards, the v head introdu es a for e, assuring the right output as predi ate of for es (whi h is needed by aspe tual fun tional head), taking a predi ate of situation as its [and it℄ introdu es a for e f and asserts that p holds of the nal situation of that for e, that is, it identies n(f) as a p situation. The v◦ head of a hange of state predi ate further imposes the requirement that the initial situation of the for e is a p situation  (Copley & Harley 2015: 24). Therefore little v expresses an energeti for e, whi h orresponds to the ausative input:  fun tional head, and Voi e introdu es the Sour e of the for e �� �� . We will use Sour e as label for external argument of ausative statives, while Causer ���� ������ �������� 173 In (454), we report the graphi stru ture derivation assumed for the in hoative the door opened : (454) (Copley & Harley 2015, ex 20) vPhf ti vhst,f ti SChsti (be ome) DPhei √ openhe,sti open the door The semanti (455) �vbe ontribution of ome � vbe ome head is the following: = λp λf. p(f in(f )) The transitive form of the ausative verb is assured by the addition of a Voi e fun tional head, responsible for the introdu tion of the external argument. The external argument is dened as the Sour e of the for e generated in the event, as reported in (456). (456) �V oicea tive � = λπ λx λf. π(f ) & source(x, f ) (Copley & Harley 2015, ex 22) More generally, the Sour e role is not divided into dierent ategories, su h as Agent, Causer or Instruments. The Sour e argument is an individual that, be ause of its inner properties or be ause of its intentions to a t, is responsible for the ausing subevent. The appli ation of for e-dynami of form or DPVs of model on eventive DPVs, su h as DPVs olor with animate subje ts, does not present parti ular issues. We have seen that the base is a root and the prex proje ts a relational proje tion responsible for the Harley (2015), I ausative semanti s. all Causer the external argument of verbs and Sour e the external argument of (457) Contrary to Copley & Daria appesantis e la bar a. Daria add weight to the boat. for external argument of ausative eventive verbs. ausative eventive ausative stative verbs. ������� �� ������������ ������������� ����� 174 V oice (458) auser P DP Daria V oice vbe auser vbe ome P ome rP DP r √ a- pesante la bar a It is worth noting that in (458), semanti s of Voi e is dynami , sin e it is of type hft, he,ft ii, it is a fun tion from for e to truth-value to a fun tion from individual to for e to truth-value. We have seen in previous se tions that it is the semanti s of Voi e that mat hes with v. A big puzzle remains unexplained in (458), namely the respe t of the mirror prin iple by the prexes (A edo-Matellan 2006). This appears to be an issue for all morphologi al theories of parasynthesis, sin e it appears to be a morphologi al derivational me hanism whi h not allowed in other reates ir umxes, whi h are ontexts in Italian, and more generally in the Roman e panorama. With these formal means we are still in no position to a stative reading of ambiguous DPVs, whi h still involves a In fa t, in (458), ausation involves an energeti stative verbs do not involve energeti ount for the ausative semanti s. for e; and by denition for es. In next se tions, I will propose an extension to for e-dynami that an a ount for stative ausative verbs. eviden e for the distin tion (linguisti and ognitive) between hange. Furthermore, I will produ e a denition of the for e-dynami nature of stative approa h Parti ularly, I will provide ausation and hange; I will investigate ausation and the linguisti reality of non-physi al perspe tives in the prepositional domain. 7.8.2 Causation of stative verbs It has already been demonstrated that inanimate subje ts ular ir umstan es, parti ipate in 2008): an, under parti - ausative stru tures (Alexiadou & S häfer ���� ������ �������� 175 (459) The sti k breaks the window. (460) Il vento rompe la nestra. The wind breaks the window. However, even in this regard there are ature. ontrasting judgments in the liter- For example, Folli (2001: 85), arguing against Reinhart's hypothesis that in hoative is derived from transitive ausative by elimination of the ex- ternal theta-role, assumes that (460) is ungrammati al and that the lexi al verb should be repla ed by periphrasti ausative Far rompere thing break'). In my opinion, (460) is well-formed, (`make some- ontrary to Folli's (2001) opinion. Monolithi nature of statives has been questioned by other resear hers (Pesetsky 1996; Grimshaw 1990) who assume that psy h-verbs do not stitute a homogeneous on- lass, but are divided into individual level predi ates (hen e ILPs) and stage level predi ates (hen e SLPs). (461) Firemen are available. (462) Firemen are altruisti . Example (461) represents a ase of SLPs. It involves an individual in a Firemen denite moment of its life. have the hara teristi of being available now. Example (462) represents a ase of ILPs. A general property of an indi- vidual is predi ated whi h does not pertain to a spe i moment. Firemen are generally altruisti . SLP/ILP distin tion is supported by experimental data by Hartshorne, O'Donnell et al. (2010). The authors onsider dierent argument patterns of psy h-verbs (subje t-experien er or obje t-experien er), in order to investigate if onsistent dieren es in their semanti s are found. Building the experiments on Pylkkänen (2000), the authors design a series of experiments whose informants (English or Japanese) were asked to de ide whi h novel verbs (referring to argument realization frame) would be used in a senten e, depending on the event des ribed. Results onrm Pylkkänen's assumption: speakers are more likely to sele t obje t-experien er psy h verbs to des ribe short mental states, whi h onstitute SLPs. These results provide strong eviden e for the non-uniformity of statives and the la k of on lusion that orresponden e between aspe t and ausality, leading to the ausation is independent from spe i fa t that some aspe tual lasses are linked to tenden y. I suppose that the fa t that aspe tual lasses. The ausation must be onsidered a ausation is more likely to be related to eventive predi ates is due to the higher likelihood that as involving energeti ongoings for ognitive reasons. ausation is per eived ������� �� ������������ ������������� ����� 176 Furthermore, other studies are providing new data against the oversimpli ation of the lass of statives. Irimia (2015) �� presents some interesting data from Mandarin Chinese, where resultative se ondary predi ates o ur with stative predi ates. Previous data from satellite frame languages, su h as English and I elandi , show that resultative se ondary predi ates were generally ungrammati al with stative verbs. (463) John walked the shoes at (int. As result of John's walking, the shoes be ame at). (464) *John loves Mary tired (int. As result of John's love, Mary be ame tired). Data from Mandarin Chinese open a new perspe tive in the investigation about resultativity, (465) ausation and stativity. Ta yige xiaoshi jiu He one hour kàn lèi right-after see le. tired perf. (Irimia 2015) As a result of his seeing, he be ame tired in an hour. The data from Indo-european languages have led to an in orre t generalization about se ondary predi ates ombined with stative verbs. on lusions are interesting for three reasons: stative predi ate spe ied by se ondary predi ations, predi ates, ausation and stativity In order to a approa h to dynami approa h to ur together. ausative verbs, the for e- ausation needs some new tools. ausation an be further ausation is not ex lusive for eventive an o ount for the stative reading of dynami Irimia's an give a Parti ularly, for e- ount only to energeti ausation, for e being an input of energy. Energeti ausation is dened as the ausal relationship generated by physi al for es. As pointed out by Copley & Harley (2015), energeti tion does not imply su h as keep ausa- hange but for es. This means that for verbs of stasis, (466), no hange is per eivable but a for e is applied by the stone on the door. The type of ausation involved in (466) is energeti , sin e energy is involved in the situation. (466) The stone keeps the door open. We saw that ausation an be involved in stative predi ates, whi h  by denition do not involve energy. We must ask ourselves whi h type �� Seminare Ontologie et Typologie des Etats, 23/03/2015, Laboratoire SFL, Paris. ���� ������ �������� of ausation this is. namely stati 177 I will propose to introdu e another type of ausation. Stati in the absen e of energeti ausation, ausation is the relationship whi h appears for es, and in presen e of a lear subordination of one individual to some resulting state made by the speaker. Before implementing this new type of better distin tion between hange and ausation it is ne essary to draw a ausation. In the next se tion, I will provide better eviden e for it. 7.8.2.1 Change is not ausation In the present study, we onsider that there is hange when the same indi- vidual is not in the same state at two dierent times, Change is tightly onne ted to time, ality by means of time development. t1 and t2 . onsequently it is linked to eventu- Eventive predi ates are the only ones apable to make time progress in language. ��������� ��������� �� �� (467) We will assume that stative reading of DPVs of surfa e in ludes ausation sin e a relationship is built between the external argument and the small lause involving the internal obje t. This relationship does not involve a hange on the Theme, but a state. This in turn implies that hange is not dire tly involved by the presen e of a result proje tion with the Theme. Relation proje tions represent states, hange is introdu ed by an upper eventive head. The eventive head is re- sponsible for the progress of referen e time, it an hors an initial state to an initial time and a nal state to a nal time. for e introdu ed by expression of vbe ome I argue that the energeti is responsible for the progress from t1 to t2 and hange. Despite the fa t that hange and ausation appear frequently asso iated, they represent two dierent phenomena as underlined by Copley & Harley (2015). Change and ausation appear together be ause we normally per eive sation by means of are hanges, and we advo ate hanges that we are not able to explain. ausation in au- ases where there ������� �� ������������ ������������� ����� 178 We an imagine that human beings are used to see physi al ee ts of ausal hains, for example one individual inuen ing (somehow) another one in a predi table sense. If we see someone tou hing a button and suddenly a light nearby turns on, we an say that Someone has turned the light on, even though the button is broken and the light turns on be ause of a temporary ele tri ity pi k. Obviously, not all events and individuals in a ausal relation sin e, in order to have an be put ausal inferen e, some pragmati restri tions need to be respe ted, namely: temporal priority, temporal tiguity, spatial We ontiguity and on- ovariation (Hume, 1739/1969, 1748/1955). an still easily re all dierent false beliefs whi h, being based on ultural stereotypes, build a ausal link between events that are not ( ausally) related. For example, in Italy one false belief states that women should not tou h plants during their periods, otherwise plants will die. a In this false belief ausal link between the tou h of a women and the death of a plant is established. It is di ult to re all what kind of event (if there were one!) would have possibly given birth to this fabri ation. We an suppose that a woman during her period on e tou hed a plant and it died, sin e then a ausal link was established between the two �� . Furthermore, human beings often re ur to myths and reate ausal links in order to explain state of aairs that otherwise would be in omprehensible. Many dierent examples are available in dierent an ient ivilizations. One example in the Roman mythology, based on the Greek one, onsists in the myth responsible for the expli ation of the Earth rotation period. Romans refer to Apollo who, by means of his Thus, Romans re urred to a An ient art, drags the Sun in the sky. ausal link started by an anthropologi al god in order to explain a natural phenomenon. All these examples are useful in order to point out that organisation of state of aairs into a ausal relations is ommon in dierent situations where hange is produ ed in the real world. Change is linked to With the same line of reasoning, we an see that ausation. ausal links are also established in order to explain states of aairs whi h do not involve Cosider for example the ommon superstition about the devil eye ditionally, it is invoked whenever someone is in a parti ular state. sterility �� hanges. �� . Tra- Female depends on the devil eye. A parti ular female individual is in a ondition whi h prevents her to have hildren; nothing is said about her pre- vious ondition, in other words there is no �� Or it is only a matter of sexual dis rimination. hange involved in the woman's �� It seems that this superstition is spread all over Europe with dierent names and dierent remedies. �� Again a matter of sexual dis rimination! ���� ������ �������� ondition. 179 However, a responsible is found: auses the woman to be sterile, but no the devil eye. hange in the woman is involved. Myths and beliefs help us to per eive the role that ways of The devil eye ausation plays in our on eptualize the world. The human tenden y to build ausal links, even though a dire t relationship is missing ( ausal illusion or per eption of ausal ausation) has been investigated by Thorstad & Wol (2016) with a series of experiments. In the Jedi Experiment, a man in an elevator moves his hand in orre- sponden e with the opening of the elevator door. People inside the elevator are the experimental subje ts. After having inadvertently assisted to the Jedi's power, they are asked to ll a questionnaire, both qualitative and quantitative des riptions of the event. In the qualitative part (verbal free des ription), informants massively re ur (91% of them) to ausal linguisti The man in the elevator kept ausing the door to stay open on the wrong oor, like magi (Ibid.: 920). In the quantitative part (Likert s ale), informants stru tures. rely to a They des ribe what they have seen with senten es like: ausal  hain, even though they signi antly per eive it for a moment. After a rst moment in whi h they establish a ausal relation between the Jedi and the opening of the doors, informants are able to re ur to world knowledge in order to reassess the state of aairs, namely that Jedi's powers do not exist in real world. This experiment provides s ienti beings usually re ur to eviden e to the ausation even though no ausal laim that human hain exists in nature. It is su ient to per eive a dire tion and an out ome in order to them to a for e, and to establish a This means that opposite hange leads to ausal relationship between elements. on eptualization of ausation, but the laim is not valid. What is retained is that human beings re ur to ausation in order to explain states of aairs, regardless to words, ausation is a hange but the opposite is not true, ausation. Causation without Linguisti ally, we is parti ularly important in The fa t that hange. hange is not hange does exist. an imagine verbs that do not denote mination point su h as ateli already argued for hange. In other essible in the presen e or in the absen e of Causation is implied in implied in orrelate verbs to involve a hange or ausative semanti s. ulThis ase of stative verbs; Copley & Harley (2015) ausative semanti s in a tivities. hange and ausation are independent on epts is a funda- mental starting point for further se tions, as it allows us to theorize for the two on epts separately. 180 7.8.2.2 ������� �� ������������ ������������� ����� Stative ausation In a for e-dynami by an eventive non-neutral state of aairs (for example the one des ribed ausal verbs), a net for e results from the initial situation. The net for e is the sum of the for es produ ed by all the on erned individuals in the situation (468). ��� � �� ��� ��� ��� ��� (Copley & Harley 2015, (468) ex.17) Sin e for e is a ve torial measure (des ribed with dire tion and magnitude), the net for e of a situation is the result of the sum of for es in that situation, as represented in gure (469). (469) �� For es are either real obje ts obje ts, as in formal linguisti s. simpli ation of the , as in ognitive linguisti s, or abstra t Figure in (469) must be interpreted as a on ept of energeti for e. It or abstra t, but it remains a ve torial measure. �� Present in the world. an be interpreted as real ���� ������ �������� 181 In one single situation ( ognitively and linguisti ally signi ant) there are dierent individuals and dierent for es whi h arise from them. Consider two individuals in a situation s1 : individual x produ es a for e fx , individual y produ es a for e fy . The net for e of s1 is the sum of fx and fy . The net for e of s1 leads to a spe i nal situation. If dierent individuals were in s1 , or dierent for es were produ ed by the same individuals, we would obtain a dierent nal situation. In the model proposed by Copley & Harley (2015), a for e onsists in an input of energy whi h is responsible for the transfer from one situation to the next. How is input of energy dened? Until now, the input of energy has been impli itly dened by means of situation: there is an input of energy when situations I refer to this type of The notion of showed that rent. I ausation as energeti ausation. ausation is linked to the notion of hange and hange. For this reason, hange, but I previously ausation are separate, even though often on ur- onsider that we are missing one possibility, namely the one in whi h ausation takes pla e without hange. We must now dene another formal means to get from the initial situation to the nal situation. Consider the possibility we want to analyze: ausation without for e. The la k of for e would lead to the impossibility of assuming the transfer between an initial situation and a nal situation, and no formal means guarantee the translation from s0 to s1 . At this point the main question is whether stative ausation is plausible within a for e-dynami linguisti lues are ne essary to a approa h, and if so, whi h ount for it. In the next se tions, I will try to answer whether stative ausation falls within for e-dynami s. 7.8.2.2.1 Is stative ausation for e-dynami ? answer to this question is no, sin e it I will show that the annot be generated by energeti as assumed by the general for e-dynami for e, approa h. We refer to the 2000 version of Talmy's approa h, where for e is assumed ibid : 409), to be a linguisti primitive with dire t grammati al representation ( parti ularly for ausation. In Talmy's formulation, the notion of for e an be extended to psy hologi al predi ates, although they do not involve physi al ibid : 430). Psy (the divided self ) for e sin e a psy hologi al for e is involved ( hologi al for es are supposed to a t within an individual hara terized by dierent dire tion and magnitudes depending on the predi ates involved. However, in Talmy's approa h to for e-dynami s the denition of psyhologi al predi ates diers from the one assumed in the present work. fa t, in Talmy's a In ount, the behavior of all sentient individuals is driven by ������� �� ������������ ������������� ����� 182 psy hologi al for es (2000: 433). In senten es su h as (470), the subje t is supposed to use her psy hologi al for e in order to physi ally a t. (470) An attendant restrained the patient. In the analysis proposed in this neither stative nor psy hologi al (Talmy 2000, ex. 20) hapter, predi ates su h sensu stri tu. restrain are I do not deny the inuen e that psy he has on sentient individuals, but I assume that it is not always linguisti ally relevant. only in the psy It is linguisti ally relevant when ausation happens he of an individual without expli it referen e to the outside world. We depart from Talmy's on eption of for es, sin e he does not that two kinds of for es are at stake onsider ognitively and linguisti ally. In the approa h I want to put forth, we must assume the presen e of two types of for es: energeti for e and abdu tion. I propose that these two kinds of for e onstitute two dierent linguisti dierent obje ts and are visible, to some extent, in onstru tions. 7.8.2.2.1.1 perspe tive. Linguisti eviden e for the reality of a non-physi al Besides the spe i ase of ausation, languages seem to make distin tions in the domain of referen e of expressions: whether physi al or mental. Con retely, this happens in the verbal domain by means of morphologi al elements. Italian possesses some verbs that generate two readings: one applied to the physi al domain and the other applied to the psy hologi al domain. The two readings are expressed by means of a dierent argument whi h an be onguration, hara terized by the presen e or absen e of a preposition, by dierent prepositions, or by (in)transitivity. For example, the verb ing, has a physi al and a psy hologi al mean- orrelated with the presen e or absen e of the preposition translated in English as ` the preposition as in idere ondire, su arve ' a. olpire, `hit' and posare, `lay'. Giovanni ha in iso la orte ia dell'albero. John arved the tree's bark. b. Giovanni ha in iso sull'edu azione di Maria. John ae ted Mary's edu ation (472) a. It an be ae t ' if is present. The same behaviour is attested for verbs su h `season', The following verbs are extra ted from LeMonnier (2014). (471) su. if no preposition is present, or as ` Giovanni ha ondito l'insalata on l'a eto. John seasoned the salad with vinegary. ���� ������ �������� b. Giovanni 183 ha ondito ha olpito il ha olpito Maria il dis orso di/* on stupidaggini. John spi ed the dis ourse with nonsense. (473) a. Giovanni on/*per ane il bastone. John hit the dog with the sti k. b. Giovanni per la sua gentilezza. John stru k Mary with his kindness. (474) a. Giovanni posa ome modello. John poses as model. b. Giovanni posa a gentiluomo. John a ts the gentleman. These senten es demonstrate that language makes distin tions between a tions that have a orrespondent energeti ounterpart in the real world (physi al), and a tions that do not have an energeti ounterpart in the real world but have only psy hologi al reality. We see that natural languages dis riminate between physi al and psy hologi al reality. We must on lude that both are per eivable ognitively and expressible linguisti ally as two distin t phenomena. Consequently, we adopt Wol 's physi alist approa h to energeti ausa- [whose℄ basi idea [...℄ is that su h relationships [between obje ts℄ an be redu ed to physi al quantities in the world, su h as energy, momentum, linear and angular momentum, impa t for es, hemi al for es, and ele tri al for es, among others  (ibid.: 85). tion  As stative verbs do not represent situations ties, they do not involve that stative physi al energies. hara terized by those quanti- We are therefore for ed to assume ausative verbs are not as ribable in a traditional for e-dynami model whi h makes use only of energeti mal means in order to a for es. Thus we need dierent for- ount for the presen e of Assuming the neutral for e-dynami prevent us to employ theoreti al status for ausality in stative verbs. ausative statives does not on epts of for e-dynami resear h. Rather, the use of su h means helps us in proposing a sort of unied/symmetri a ount for both 7.8.2.2.2 an a �� ases . Causation without ount for stative hange In this se tion, I will propose ausative verbs that involves the on ept of stati ausation and slightly redenes the notion of situation. �� A possible obje tion to this on lusion ould be that stative ausative verbs do not parti ipate in parti ular argument realization ongurations, rather they enter in usual patterns, ex eption made for una usatives. ������� �� ������������ ������������� ����� 184 In senten e (475), we annot assume that there is a hange sin e, as shown in Se tion 7.7, it is stative. (475) The drape darkened the room. The person who pronoun es this senten e expresses a link between individuals in a parti ular situation. This relationship is not only Figure-Ground of the type assumed for ILPs by Ram hand (2008: 55) and reported in (476). vP (476) v DPholder v This means mainly two things about ′ DP/NPrheme hange; and about for e. For the property of Experien er is persistent, and no hange, hange is involved in its sta- tus. Namely, previous state/property of the obje t is linguisti ally relevant: the room is dark now, will be dark in the future and was dark sometimes in the past �� . There is no salient hange expressed linguisti ally, of ourse, being SLP the properties of individuals are alterable, but the potential moment of hange is not expressed. Regarding for e, no for e is involved, sin e the predi ate is stative, no input of energy is involved. Consequently, no transition from an initial situation to a next situation takes pla e as shown by (477) where there are only situations and no for e arises. (477) Sin e no hange and no for e are involved in senten e (475), we should assume that there would be only one (initial) situation, if we sti k to a lassi the for e-dynami approa h. This would make impossible to a ount for ausal meaning of these senten es. �� For persisten e in time of stative predi ates, refer to Altshuler & S hwarzshild (2012). ���� ������ �������� In the ase of stative 185 ausation, we assume that a situation is a single individual and its properties. ut around Propositions are sets of possible sit- uations, rather than sets of possible worlds. Possible situations are parts of possible worlds. Lumping is the operation that assures the right truth values to propositions. Lumping A proposition p lumps a proposition q in a world w if and only if (i) and (ii) both hold: (i) p is true in w (ii) Whenever a situation s is part of w and p is true in s, then q is true in s as well [Kratzer (1989: 611)℄ A ording to examples of ontexts proposed by Kratzer (1989: 608) and reported in (478) and (479), we an see that a same state of aair in the world (naively speaking) ut o in dierent ways su h that all are an be true. (478) Pedant: Paula: Dialogue with a pedant. What did you do yesterday evening? The only thing I did yesterday evening was paint this still life over there. Pedant: This annot be true. You must have done something else like eat, drink, look out of the window. Paula: Yes, stri tly speaking, I did other things besides paint this still life. I made myself a up of tea, ate a pie e of bread, dis arded a banana, and went to the kit hen to look for an apple. (479) Lunati : Paula: Dialogue with a lunati . What did you do yesterday evening? The only thing I did yesterday evening was paint this still life over there. Lunati : This is not true. You also painted these apples and you also painted these bananas. Hen e painting this still life was not the only thing you did yesterday evening. 186 ������� �� ������������ ������������� ����� Imagine that the world is a room and that in this room dierent obje ts exist. You an ask your friends to des ribe the state of aairs. One of them It is a room, another one It is a warehouse and a bun h of obje ts in the same pla e. All these senten es an say another one It is refer to the same room, but ea h underlines something dierent of this same room. It is what happens for situations. Both in for e-dynami s and in stati salient individuals. ausation, a situation However, for e-dynami s and stati the size of situations involved. In for e-dynami s, a situation ent individuals (480), in stati ontains all ausation dier for ontains dier- ausation only one (486). (480) In the ase of stative ausative verbs involving two arguments, we are in presen e of two situations, ea h of whi h ontains one individual (denoting the argument) and its salient properties. If in a for e-dynami model, the situation is the derived net for e, in stati ut around individuals and ausation the situation is individual and its properties. Thus, in an energeti and in a stati ut around one ausation there is for e, ausation there is property. We saw that a situation a state. Sin e stati ontains a state of aairs that an be des ribed by ausation involves two situations, there are two states in relation: the existen e of the external argument and the state of the internal one. Whi h element is responsible for the introdu tion of ausal meaning? The link between existen e of the external argument and the state of the internal one is not reated by an energeti for e, but it is brought in the system by a human being (the speaker). She is responsible of the establishment of relation by means of her abdu tive ausal apa ity. In order to dene abdu tion, the following ontext is useful. Imagine [o℄ne morning you enter the kit hen to nd a plate and up on the table, with bread rumbs and a pat of butter on it, and surrounded by a jar that  ���� ������ �������� 187 of jam, a pa k of sugar, and an empty arton of milk. You on lude that one of your house-mates got up at night to make him- or herself a midnight sna k and was too tired to lear the table. This, you think, best explains the s ene you are fa ing. To be sure, it might be that someone burgled the house and took the time to have a bite while on the job, or a housemate might have arranged the things on the table without having a midnight sna k but just to make you believe that someone had a midnight sna k. But these hypotheses strike you as providing mu h more ontrived explanation , (http://plato.stanford.edu/entries/abdu tion/ , Douven: 2011). The abdu tive way of reasoning is the following: a. All peas of this box are green. b. These peas are green. . Then, these peas belong to this box. The truth of the third senten e is only probable and not these peas ertain, sin e an belong to another box. It is important noting that abdu tion is a way of reasoning used in the presen e of energeti ausation too, sin e it is the sole way of reasoning able to raise our level of knowledge of the world. What I dene here as abdu tion is a sort of extreme of the philosophi al on ept. That is, abdu tion arises when the speaker establishes a link between two entities, assuming that a general property of one of these is responsible for the property of the other, in the absen e of expli it and visible is the reasoning about hain. In other words, abdu tion auses by ee ts. If in a for e-dynami whi h generates ausal model there is the produ tion of physi al energy ausation, in a non for e-dynami model there is abdu tion of a sentient individual, as summarized in Table 7.7 (page 187). Model For e-Dynami Stati ausation Generator of ausation Preferen e physi al energy abdu tion Table 7.7: Distin tive traits of eventive and stati ausation (provisional). It is worth noting that abdu tion is the sole for e �� able to reate stati ausation. In the absen e of a sentient individual who as ertains a between two states, so who inserts abdu tion, no �� If it an be onsidered a for e. ausation ausal link an be assumed. ������� �� ������������ ������������� ����� 188 Furthermore, for establishing a stati ausal link between two obje ts in the world, both obje ts must be present. In other words, the presen e of the Sour e in the situation is ne essary so that a speaker establishes a ausal link between it and the Theme's state. For energeti state ausation, imagine a an be due to dierent violently or the ontext in whi h an egg is broken. This auses, for example, John opened the fridge too at played with an egg left on the In ea h of these ounter-top. ases, if we enter the kit hen we an onstru t some hypotheses about the breaking of the egg, even though John or the not present. The main point is that a person ar are an assume that someone or something is at the origin of the egg's a tual state, even if the responsible is not present at the time utteran e. (481) John broke the egg. (482) The egg is broken, it must be John. Imagine, on the other hand, a state of aairs in whi h you see a wall. You annot suppose that it is yellowish be ause of the sofa, if the sofa is not there. This means that senten e (484) annot be pronoun ed if the sofa is not under our eyes. The presen e of the Sour e is mandatory for assuming Theme's state. If the Sour e is not physi ally present, a relation between it and the Theme annot be established. (483) The sofa yellowed the wall. (484) The wall is yellow, it must be the sofa. In energeti ausation, the result alone is su ient to allow the expression of a link between it and a possible Causer, even though the latter is no more present in the state of aairs. speaker On the other hand, in stati ausation the annot adbu t about the Theme's state if the state of the Theme annot be led ba k to the Sour e, i.e. if the Sour e is not present. We an apply stati ausation approa h to ausative statives, su h as fear nightmares and in (485). (485) Nightmares frighten John. The two individuals present in the argument stru ture, John are linked by means of John's state of mind. John is the sole person in the world who an feel that the sole individual who nightmares s are John. an establish a For this reason, John is ausal relation with nightmares. ���� ������ �������� 189 Why do nightmares have this ee t on John? We don't know from senten e (485), sin e it only states that one or more properties of nightmares are responsible for one property of John, namely the one of being s ared, whi h is lexi alized by the verb. This way, what is dened as stati two properties of two individuals. sense that no physi al energeti is ausation The is a onne tion between onne tion is not physi al, in the for e intervenes to establish it, rather it reated/per eived by the psy he of an individual, thanks to abdu tion. There are properties of an individual whi h are per eived to be responsible for properties of another individual. These are onsequently lexi alized as a ausal link. � � �� �� �� (486) Figure (486) represents the ognitive situation: two individuals x and y belong to two distin t situations sa and sb whi h are not in relation to one the other, we 1 an imagine that they belong to another wider situation s . When an organism (Barwise & Perry 1983: 10) established a �� between sa and sb , she's stating that in her opinion ausal link some impre ise property of x is responsible for a property of y whi h is exempliable with p(y), without generating a hange. I assume that sentient individual an establish this link by means of ten- den ies possessed by obje ts. Just as in for e-dynami uals possess preferen e to to be. a t, in stati model, where individ- ausation, individuals have preferen e Individuals have tenden ies to possess parti ular properties that are interpreted as possible ausal ante edents. A property belonging to an individual is part of its inherent qualities. This means that proper parts, proper features of an obje t properties. same obje t dierent onstitute its The denition of property is not obje tive, sin e one and the an be assigned dierent properties by dierent individuals or ontext, be ause In this regard, we de gustibus non disputandum est. an use the following examples and see that they are grammati ally ne, but pragmati ally odd. �� For what she knows about the world. ������� �� ������������ ������������� ����� 190 (487) ??Il emento ingiallis e l'albero. The on rete yellows the tree. The oddity of (487) is due to the fa t that the tenden y towards yellow trees. If we substitute on rete does not have a on rete in (488), the senten e be omes more plausible. We with guano, su h as guano an imagine that has a tenden y towards yellow trees. (488) Il guano ingiallis e l'albero. Guano yellows the tree. Thus, traits that distinguish energeti ausation and stati ausation (as summarized by table 7.8, page 190) are: (i) dierent generators of the ausal trend, energy for the former and abdu tion for the latter; and (ii) the tenden y of involved individuals, to Model for the former and to Generator of For e-Dynami Stati a t ausation essarily for the latter. Tenden y physi al energy to a t abdu tion to be ausation Table 7.8: Distin tive traits of eventive and stati Stati be ausation does not involve energeti ompletely time-overlapping, ausation (denitive). for e, thus situations are ne - ontemporaneous, represented by the wider s1 , in (486). The assumption of a wider situation onsistent intra-linguisti an explain the high ross-linguisti ally variability in argument stru ture of stative verbs, su h as for psy hologi al verbs. All ongurations lexi alize the same dierent linguisti in argument patterns. ognitive and linguisti They are symptoms of a and (490). a. Le foto sono sul muro. Pi tures are on the wall. b. Le foto abbellis ono il muro. Pi tures embellish the wall. (490) a. hanging nature of the bond between individuals: lo ative stative relation, su h as (489) and (490), or (489) ognitive situation by means of Giovanni teme gli in ubi. John fears nightmares. ausative stative, su h as (489) ���� ������ �������� b. 191 Gli in ubi spaventano Giovanni. Nightmares frighten John. The la k of energeti for e allows humans to re ur to dierent argument patterns to express the stati situation they per eive, taking advantage of this vagueness in order to put dierent elements in prominen e. In eventive predi ate this is not allowed, as seen in se tion 7.8.1, sin e argument stru ture mirrors the ausal hain. To summarize, in stati ausation there are two individuals (x and y) with their properties (p and q). There is not just one possibility to rely p(x) to q(y). In speaker's opinion, it is the ase that p(x) what she knows, for what she sees, she auses q(y), i.e.: an abdu t that p(x) is in a for ausal link with q(y). Thanks to abdu tion, x and y are ordered and this has immediate ree ts on the argument stru ture: x is prominent and y is lower. x is Sour e and y is Theme (generally), x is subje t and y is obje t. We an apply these new improvements to stative DPVs. We have al- ready seen that they are root derived and that the prex is expression of the ausative head. This means that synta ti ally, stative DPVs do not dier in this extent to their eventive en e ounterparts. The main and substantial dier- onsists in their la k of eventive proje tion. are generated by vbe ome , stative (491) �vbe (492) �vrelation� = λp λs. p(s) ome � v relation, whi two elements. energeti Eventive ausative DPVs ausative DPVs are generated by vrelation. = λp λf. p(f in(f )) h is a predi ative head, establishes a ausal relation between (491) and (492) dier in the nature of the transfer involved: ausation involves transfer of energy from Causer to Theme, stati �� ausation involves a virtual The head vrelation and non permanent transfer of property. assures the ausal relation between a property of the subje t and a property of the obje t. A stative sub-event auses a stative result, this is possible be ause of the speaker building that link whi h is not otherwise physi ally present. This means that no eventive sub-events are assumed in the derivation. Dierent semanti s of v0 assures dierent Voi e heads whi h are responsi- ble for the introdu tion of external arguments. In stative DPVs, the subje t is not responsible for the introdu tion of for e, while in eventive DPVs it is. �� In the sense that it is not really per eivable by means of a physi al hange. ������� �� ������������ ������������� ����� 192 (493) V oice (494) V oicesour auser e = λf.causer(x, f )p(f in(f )) = λs.source(x, s)p(suc(s)) In this se tion we showed that stati for e-dynami model of ausation. nitive element responsible for ausation does not belong in a It involves a single situation. The og- ausation's generation is abdu tion, whi h is introdu ed by the a sentient individual (the speaker). If the Sour e is not present in the state of aairs, no stati arise. Subje ts of stati knowledge about their properties I propose that stati alled vrelation ausation an relation have tenden y to be, and general worldan generate oddity. ausation is stru turally built by a fun tional head whi h is responsible to establish a link between the presen e of the Sour e and the state of the Theme. The ausative meaning arises be ause of the presen e of a lower rP. 7.9 Synta ti derivations 7.9.1 Causative eventives We have seen that eventive DPVs do not dier from other verbs. This means that their ausative semanti s is ausative eventive ompatible with the presen e of a lower predi ative stru ture (here rP, elsewhere SC) (Hoekstra 1988; S häfer, 2008) and the eventive/ ausative semanti s is introdu ed by vbe ome . vbe ome is responsible for establishing a relation between the external ar- gument and the new state of the internal one by means of energeti The fun tional head Voi e v, auser, a ording to the eventive semanti s of introdu es the external Causer argument. (495) Giovanni anneris e la stanza. John bla kened the room. for e. ���� ��������� ����������� V oice (496) 193 auser Pheti V oice DP auserhe,f ti ' Giovanni Voi e v Phf ti auserhf t,he,f tii v be omehst,f ti rPhsti DP la stanza r √ a- nera In se tion 7.11, we will see that another element must be taken into a ount in the derivation. This is a judge parameter introdu ed by the base root in relation to its nature of predi ate of personal taste. 7.9.2 Causative statives We have seen that statives an be ausatives due to abdu tion, whi h is introdu ed in the system by a sentient individual. Abdu tion is not an energeti for e. The speaker's intelle t is able to establish a ausal relationship between the presen e of the external argument and the state of the internal argument. We have onsidered the verbal part, whi h in not involve any energeti ase of stative verbs does for e. We represent this by means of vrelation whi h is responsible for the relational link between the lower relational proje tion (the state of the internal obje t) and the existen e of the external argument. In parti ular, vrelation denotation is as the following: (497) vrelation = λpλs.p(s) (498) Il divano anneris e la stanza. The sofa bla kens the room. ������� �� ������������ ������������� ����� 194 vP (499) v rP vrelation DP stanza r a- √ nera In ase of stative reading of DPVs, the Voi e head introdu es a Sour e external argument. (500) V oicesour e V oicesour e 's derivation is: = λs.source(x, s)p(suc(s)) The external argument (Sour e) is per eived by the speaker as responsible for the ausal link between it and Theme, it has a property responsible for the state of the internal argument. I will on lude that the ausal relation between external and internal argu- ments is brought into the system by the speaker. This is semanti ally further supported by the presen e of a judge parameter introdu ed in the derivation by the base root, whi h represents speaker's opinion (se tion 7.11.3). The ausal meaning of these verbs is determined by the presen e of a r head in the lower part of the derivation whi h proje ts a rP, just as in eventive verbs. (501) Il divano anneris e la stanza. Sofa bla kens the room. ���� ��������� ����������� 195 V oicesour e Phsti (502) V oice′sour e he, sti DP il divano Voi esour v Prelationhsti ehst,he,stii v relationhst,sti rPst DP r la stanza Note that the result of the derivation is of type verbs to hsti, whi a- √ nera h prevents stative ombine with the progressive aspe t in English; the progressive sele ts predi ates of for es of type hf ti. A dilemma remains open, and it involves the relationship between real world and real world, it an be ognitive situations. In energeti ognition and language for stati an be established: a for e is in the world, ognitively per eived, hen e a and a mat hing linguisti stru ture ognitive hain is produ ed, Does the same hold ognitive situation means. In fa t, if orresponds to what happens in the real world, we must a knowledge two dierent operations generating verb's aktionsart. ausal an be employed. ausation? We must ask what ognitive situation ausation, a parallelism between suc(s) in relation to the When a for e is produ ed in the world, it is ognitively per eived (whenever it is a tually per eived) as involving an energeti Linguisti ally, a ausal dynami verb an represent the and this is represented by the operation in the world, an energeti suc(s) for e is not suc(s). ognitively per eived. to possible real world stati linguisti ausal hains. Consequently ausation is involved, ognitively, in the speaker's mind, happens and, by means of abdu tion, order to solve this issue, studies ognitive situation, When a for e is not produ ed should not be involved. However, sin e must be a li it operation. Thus, an be suc(s) for e. an be su suc(s) ausation essfully applied. In ondu ted about how human rea tions situation and how they establish ognitive and ������� �� ������������ ������������� ����� 196 7.10 Causative statives and statives The dieren e between statives and ausative statives is the same as the one observed between eventive verbs and ausative eventives. The dieren e redu es to the presen e of a non-eventive relational link between the dire t obje t and the verbal base in the lower part of the derivation, namely a rP. Causative eventives involve a resultative reading by means of the presen e of energeti v for e in head, whi h is onsequently hara terized by for e. Causative statives do not yield a resultative reading, sin e no is involved, the hara terization of their eventive one: no energeti We for e is v (vrelation) regular statives, sin e both present a ausative statives must dier from the ontained. an ask then what is the dieren e between for eventives, hange vrelation ausative statives and verbalising head. Exa tly as ontain a rP, while regular non- ausative statives present a simple individual. In senten e (503), there is a relationship between the external argument and a state of the room. In senten e (504), there is a relationship between the external argument and the ar, namely the relation of possession is in a state of possessing the ar. The point is that no property is attributed to the ar be ause of its possession relationship with Daria. (503) Pi tures embellish the room. (504) Daria owns this In their analysis of ar. ausative eventive verbs, Folli & Harley (2005) point out that dierent phenomena are linked to They analyse ausation in eventive verbs. eat, onsumption verbs, su h as animate individual is the external argument, no implied. This is reported in (507), where the eat Daria v and show that when an ausation and no result are head is o upied by the verb and no SC is present. (505) John ate the apple. (506) Mario mangiò la mela. Mario ate the apple vP (507) v DP John (Folli & Harley 2005, ex. 30) ′ v DP ate the apple ����� ��������� �������� ��� �������� 197 When an inanimate subje t is involved, a result subevent is mandatory, sin e the subje t does not have du es teli ity and (508) (509) ontrol on the event. The result event intro- ausative semanti s. a. *The sea ate the seaside. b. The sea ate the seaside away. a. *Il mare ha mangiato la spiaggia. The sea ate the bea h. b. Il mare si è mangiato la spiaggia. The sea ate . self the bea h. Il mare ha mangiato via la spiaggia. The sea ate the bea h away. vP (510) (Folli & Harley 2005:36) v DP ′ Il mare v si ( aus) SC DP Vadjective la spiaggia mangiato Folli & Harley (2005) do not re ur to Voi e head for the introdu tion of external arguments, whi h are introdu ed by v head. Note that roles of the external argument in (507) and (510) are dierent: the former is an Agent, the latter is a Causer. We proposed that external arguments are introdu ed by a Voi e head, thus dierent subje ts roles are introdu ed by dierent Voi e heads. These are determined by the eventuality of V oice In the v auser relates to a dynami ase of stati ausation, v. In the ase of for e-dynami s, v and produ es a Causer external argument. V oicesour e relates to a stative (predi ative) and gives a Sour e external argument. In other words, whenever a rP is present in the derivation, meaning is generated. Regular statives do not obje t (512), thus they do not have (511) La sorella ama Maria. The sister loves Mary. ausative ontain rP on the internal ausative meaning. ������� �� ������������ ������������� ����� 198 vP (512) v DP ′ v XP ama Maria La sorella In the following example, a rP is present, so The Sour e individual introdu ed by Voi e is ausative semanti s arises. onsequently per eived as re- sponsible for the internal state of the obje t. (513) I quadri abbellis ono la stanza. Pi tures embellish the rooom (514) Voi eP DP ... i quadri rP DP r la stanza √ a- P bella 7.11 Predi ate of personal taste DPVs present a further element that must be dis ussed: a pragmati judge parameter. Predi ates of personal taste predi ate of questions of opinion and not of matters of fa ts (Larson 2005). In this se tion we will see the role of judge parameter in relation to verbal aspe t. The two types of DPVs dier in the allowed referen e of the judge parameter. In eventive DPVs, the judge parameter, annot relativize parts of the senten e that are spelled out as being dierent from rP. In stative DPVs, the judge parameter (Voi e, v an relativize all parts of the l-syntax of the verb and rP). This is due to the fa t that stative reated by abdu tion whi h is per se ausatives DPVs are a matter of opinion, making possible to relativize the result, the fa t that there has been ausation or the fa t that the responsible for the Theme's state is the Sour e. This statement is ����� ��������� �� �������� ����� 199 supported by results of a disagreement test (Stephenson 2007), ondu ted in Se tion 7.11.1. In formal semanti s literature, adje tives of personal taste are asso iated with the presen e of a judge parameter that relativizes truth-values to a spe i individual. (515) The ake is tasty. (516) The ar is beautiful. The truth value of senten es above dividual, tasty hanges in relation to a spe i alled the judge. Therefore, we and for whom the ar is beautiful. an question for whom In other words, the ake in- the ake is an be tasty for Mary, but quite disgusting for Daria. Dierent hypotheses have been proposed to determine whi h individual relate to an internal state or experien e, the question arises as to whose internal state or experien e is being reported in any parti ular ase  (Stephenson 2007: 490). is the judge. Predi ate of personal taste  7.11.1 Disagreement test In order to determine to whi h individual the judge of senten e like (515) refers, it is possible to resort to the disagreement test (Stephenson 2007: 492). Disagreement test onsists of questions between two individuals who dis- agree about the state of a given obje t. If the state of the given obje t is expressed by an adje tive of personal taste, a possible disagreement between the two individuals does not generate a ontradi tion, su h as in (517). Oth- erwise, if the state is expressed by another type of adje tive, disagreement generates (517) (518) ontradi tion, su h as in (518). A: Mary's B: Yes, it's C: Oh no, it's not A: Mary's B: Yes, it's red. C: # Oh no, it's not red at all. This same test �� ful' ar is ool. ool. ool at all. ar is red. bella, `beautima hina, ` ar', but it an be applied to Italian, for adje tives like . In senten e (519), the adje tive predi ates over �� Note that not all adje tives are able to introdu e a judge parameter. ������� �� ������������ ������������� ����� 200 is di ult to attribute this judgment to a pre ise subje t. Is the ar beautiful for me, for everybody, only for some people? A ording to Stephenson (2007: 492), in example (520) a predi ate of personal taste pronoun ed by subje t B, by a person C without generating a not mean that the true for C for the (519) (520) A. ontradi tion. an be denied What B is saying does ar is beautiful only in his/her opinion, and the same is ontrary statement. Questa ma This the ar is beautiful, hina è bella. ar is beautiful. Com'è questa ma hina ? How's this ar? B. È bella! It's beautiful! C. Oh no, non è bella per niente! Oh no, it is not beautiful at all! Whenever expli it referen e is made about the person who expresses her taste, by means of for in English and of on someone and disagreement generates (521) A. The B. # Oh, no, it is ugly! per in Italian, the parameter is set ontradi tion. ar is beautiful for John. In the next se tion we will use disagreement tests on DPVs of surfa e, in order to see whether their base root is a predi ate of personal taste and to see dierent possibilities of referen e of the judge parameter in relation to the verbal aspe t. 7.11.2 Judge Parameter (eventive reading of DPVs) In this subse tion I present some eviden e whi h shows that in the ase of eventive stru tures, the pragmati al possibilities of referen e of the judge parameter are restri ted to the state of the internal obje t, as shown by (522) whi h reports a disagreement test on the result part. (522) A. Cosa fa Giovanna? What does Giovanna do? B. Abbellis e la stanza. She's embellishing the room. ����� ��������� �� �������� ����� C. 201 Oh no, non la abbellis e per niente, quei quadri sono disgustosi! Oh no, she doesn't embellish it at all! That frames are really ugly. Person C does not disagree on the fa t that Giovanna is performing an a tion on the room, rather she's arguing about the result of her doing. The following representations sket h the reason of the non ontradi tory nature of the disagreement in (522). (523) B is saying [G CAUSE [the room BE beautiful for j℄℄, where j is the judge who evaluates (524) C is saying [G. CAUSE [the room BE not beautiful for j℄℄, where j is the judge who evaluates, whi h is dierent from j for B If we try to disagree about the fa t that the event of pla e, we will see that judge parameter (525) A. ontradi tion is generated. abbellire has taken In eventive DPVs, the annot relativize the eventive part. Cosa fa Giovanna? What does Giovanna do? B. Abbellis e la stanza. She embellishes the room C. ???Oh no, non fa niente!/Oh, no, lava i piatti! Oh no, she doesn't do anything!/Oh, no, she washes the dishes. In (525), the disagreement between B and C is about what does, thus about the nature of the Giovanna ausative event. Therefore it derives in a ontradi tion. (526) Giovanna abbellis e la stanza. Jeanna makes the room beautiful. a. Giovanna fa Jenna a ts b. per rendere to ause . la stanza bella (per judge). the room beautiful (for judge) 7.11.3 Judge Parameter (stative reading of DPVs) A dierent pi ture emerges with stative reading of DPVs of surfa e. We will see that the judge parameter an relativize the meaning of all the l-synta ti ������� �� ������������ ������������� ����� 202 layers. In disagreement tests no disagreement about the nature of ontradi tion is generated even in ase of ausation. Using disagreement test in (553), we see that no ontradi tion arises from the negation by C of the statement by B about the the state of the Theme ( beautiful table ). (527) A. What are these owers doing on the table? B. They are embellishing it. C. Oh no, they are not embellishing it at all. In (528), we an see that disagreement about the responsible for the table state does not lead to a ontradi tion. The judge parameter an relativize the relationship between the Sour e and the state of the Theme. (528) A. Why is the table this way? B. Be ause of the owers. C. Oh no, not at all, it is this way be ause of the light! Results of disagreement test for DPVs of surfa e means of paraphrase (529). It is useful to is an be illustrated by larify that an individual's opinion ontained not only for the denition of table state, but also for the de- nition of the individual and ausative event per eived as responsible for this state. (529) The owers embellish the table. tiful. = In the opinion of the speaker → The owers make the table beau- the table is beautiful and in the opin- ion's of the same speaker the main fa t responsible for this is the owers on the table. I suggest that this is possible be ause DPVs of surfa e involve a stati ausation, where the speaker is responsible for establishing a tween the subje t and the obje t. abdu tion, establishes a tween two obje ts. ausal link be- In other words, the speaker, through ausal link otherwise not present in the world be- Consequently, ausation is matter of speaker's opinion. This allows the judge parameter to refer to every part of DPVs of surfa e. In other words, the non- ontradi tion in disagreement test on the of stative DPVs of surfa e is due to abdu tion, whi h matter of the speaker. (530) I quadri abbellis ono la stanza. The pi tures embellish the room. ausal part onsists in a personal ����� ����������� a. 203 I quadri sono (per judge) The pi tures are (for judge) b. per rendere to make (for judge) . la stanza bella the room beautiful (for judge) 7.12 In this type Con lusions hapter, I have analyzed Italian parasyntheti abbellire, `to embellish', ingrandire, deadje tival verbs of the `to enlarge' and instupidire, `make someone stupid'. Morphosynta ti eviden e shows that the base is not adje tival, rather it involves a non- ategorized root. The prex is responsible for the proje tion of a non-eventive relational stru ture whi h involves the internal obje t as a subje t and sele ts the base root. The presen e or the non-eventive proje tion is responsible for the ausal meaning. We divided DPVs into three lasses, a ording to the semanti s of the base root: (i) psy hologi al; (ii) of form; (iii) of surfa e. The rst group has not been treated in this work. The latter two present dierent properties when the subje t is inanimate: DPVs of surfa e are stative and DPVs of form are eventive. These aspe tual hara teristi s have been put forth based on thanks to four tests: interpretation under modal; interpretation under già ; ontribution; dierent adjun ts. temporal narrative eventive DPVs are In order to a approa h to al.), Both stative and ausal. ount for stative ausative verbs, we adopted a for e-dynami s ausation (Copley & Harley 2015; Copley & Wol 2014; inter introdu ing some new tools. Having demonstrated the existen e of presen e of hange, I argued for stati ausation independently from the ausation. Stati ausation arises in the presen e of a rP and relates the existen e of the external argument to the state of the dire t obje t. We have introdu ed a virtual for e, abdu tion, whi h is brought into the system by the speaker. responsible for establishing the ausal stati Furthermore, we showed that, tionship link between subje t and obje t. ontrary to energeti en e of Sour e is mandatory in stati alled Abdu tion is ausation and no ausation, the presausal stati rela- an be re overed by the sole presen e of individual denoted by the Theme. We gave a ount for dierent l-syntax of eventive DPVs, of stative DPVs and of usual statives. 204 ������� �� ������������ ������������� ����� We have shown, by means of disagreement tests, that DPV an be rela- tivized by a judge parameter whi h is made available by the base root. Judge parameter (meaning: on the type of in someone's opinion ) ausation. In the the result part. In the of l-syntax. parameter is presents dieren es depending ase of eventive verbs it ase of stative verbs it an relativize only an relativize all dierent parts We propose that in DPVs of surfa e the behavior of the judge aused by the fa t that stati ausation is reated by abdu tion. Abdu tion is a virtual for e introdu ed by the speaker who believes that there is a ausal link. in her opinion Chapter 8 Stativity an be automati ally dete ted 8.1 Introdu tion This hapter is the out ome of a ollaboration in a proje t ondu ted by Dr. Copley (CNRS-SFL, Fran e) and Dr. Wol (Emory University, Georgia, US). I have ollaborated only in one part of the proje t whi h possible automati identi ation of stative verbs in a on erns a orpus. The aim of the wider proje t is to identify the temporal orientation of senten es from stru tural riteria dened a priori, whi h an be applied by an arti ial intelligen e. The part of the proje t whi h onstitutes this stativity diagnosti s to be implemented in automati hapter aims to identify natural language pro- essing. In this general framework, the identi ation of stative verbs is fundamental. We have seen that stative verbs involve dierent temporal onstraints ( hapter 6). For example, in a present tense senten e, eventive verbs re eive a habitual reading, while stative verbs easily re eive a parti ular reading in whi h they refer to a present ongoing situation. (531) (532) a. Mary breaks a glass (# now/on e a week). b. Mary is breaking a glass (now/on e a week). a. Mary owns a mus le b. ??Mary is owing a mus le ar (now). ar (now). There are many ways to provide the system with a tive verbs. The simplest strategy lassi ation of sta- onsists in providing a list of stative verbs. 205 206 ������� �� ��������� ��� �� ������������� �������� Although easily reated, this strategy presents dierent issues: a list is in- omplete; stative verbs Another strategy whi h an be for ed to an eventive reading by stru tures. onsists in the identi ation of some stru tural an dis riminate stative stru tures. This riteria hapter adopts this se ond way and des ribes it in details. It presents the pro edure we designed in order to automati ally identify statives, on the one hand; and produ e a gradient for stativity of English verbs, on the other hand. The rst goal has been rea hed by means of the denition of synta ti rules whi h an be interpreted by a parser. The se ond goal has been rea hed by the interpolation of data obtained by ma hines and human data. 8.2 Pra ti al appli ations Chapter 6 des ribes some of the interpretative dieren es that stative and eventive verbs generate, and underlines the importan e of using them as possible eventuality diagnosti s. All the des ribed diagnosti s are useful if used by human beings. In this se tion we will see how to ine t them in order to get them understood by a ma hine. Human beings are able to at h the dieren e between two readings gen- erated by the same stru ture. Su h as (533) and (534), where the previous implies a deonti reading and the latter a prevalent epistemi (533) John must go to the shop. (534) John must own a bi y le. Dierent readings an have other semanti onsequen es, for example (533) and (534) are subje t to two dierent future The opportunity to ma hine and it reading. onstraints. at h dierent readings is not given for free to a annot be ignored if we want to arrive at a good automati language interpretation. For this reason, stativity, the identi ation of whi h an appear to be unne essary, is in reality very important in relation to temporal interpretation of a senten e. The rst step for the automati denition of a set of semanti identi ation of stativity and synta ti onsists in the stru tures whi h behave dier- ently in the presen e of stative and eventive verbs. It is worth re alling that syntax an for e stative verbs into an eventive reading. should not expe t that rules identify a Consequently, we losed set of stative verbs. Corpus analysis will pi k up all those utteran es in whi h the stru ture generates a ���� ��������� ������������ 207 stative reading. We should rather expe t a sort of ranking of stativity, from the most likely eventive verbs to the most likely stative verbs. English disposes of several synta ti and semanti dis riminate between stativity and eventivity. rules to be pro essed by a They onstru tions whi h an be translated in omputer. We pro eed with the denition of synta ti stativity/eventivity rules, and their translation into parser (Tregex) rules. Then, we apply Tregex rules to a orpus. Ea h Tregex rule looks for a onstru tion and onsequently produ es a list of verbs with the number of utteran es in whi h a single verb was found in that onstru tion. A synta ti annotated orpus was used (Thorstad & Wol 2016) and it was explored by means of Tregex. The next subse tion reports the steps followed to get to the verb ranking. 8.2.1 Clues and notated orpus rules We have seen that stativity is dened and dete ted negatively. For this reason, we mainly employed eventivity diagnosti s in order to produ e a gradient from the most eventive to the most stative verb. Synta ti lues for eventivity are the possibility of appear with: progres- sive (535) and imperative (536), agent oriented adverbs su h as deliberately (537), una usative stru tures (538). involuntarily, Stative verbs annot ap- pear in these stru tures. (535) (536) (537) (538) a. Mary was ki king Abel. b. ??Mary was hating Abel. a. Don't eat that sandwi h! b. *Don't love that dog! a. Mary deliberately ki ked Abel. b. *Mary deliberately hated Abel. a. From the explosion the glass melted. b. *From the explosion John loved ( int. John undergone a hange from not loving to loving) These stru tures must be translated in a algorithm whi h by a omputer. an be pro essed Consequently, not all are useful for the present aim, in parti ular those whi h are based on a semanti interpretation. Sin e English 208 ������� �� ��������� ��� �� ������������� �������� does not have spe i � morphologi al means for imperative , it annot be employed here. On the other hand, English progressive is expressed by spe i logi al means: verb pro essable by a be + gerund, whi h morpho- an be translated into an algorithm omputer. It is worth noting that the global utteran e rate for a verb in a progressive form is not informative per se, and must be related to the global utteran e rate of the same verb in present and past tense. Table 8.1 (page 209), reports algorithms whi h retrieve progressive forms. The obtained results must be interpolated with the global amount of utteran es of that verb in the orpus. Table 8.2 (page 210), reports some of the senten es pi ked out in a pus of English tweets for dierent or- ountries. This shows dierent stru tures identied by ea h rule. Rules 1, 2 and 3 pi k verbs in a non-progressive form. Rule 1 sele ts for all forms of a verb in the simple present (non-third and third person) and past tense, present and past parti iple. Rule 2 sele ts verb in -ing form (its gerund or present parti iple) with or without the presen e of the verb whi h is the formal means to express the progressive. be, Rule 3 sele ts only verbs in simple present (non-third and third person). These three rules are required to normalize the utteran es of progressive forms, in order to get rid of a frequen y ee t. Rules 4 and 5 are spe i for sele ting progressive forms. Rule 4 sele ts all verbal phrases in whi h verb be and a parti iple o ur. Rule 5 sele ts verbal phrases embedded in higher senten es, this allows the program to pi k up progressive forms embedded in a bigger senten e. Another is the (539) (540) riteria that an be easily translated in a ma hine-friendly rule ausative-anti ausative alternation. a. Sandra broke the window. b. The window broke. a. Sandra loved that window. b. *The window loved. The fa t that the English anti ausative spe i onstru tion does not present morphologi al traits makes impossible to look dire tly for it. � The most prominent synta ti We hara teristi is the la k of subje t. This ould be per eived as a su ient lue, sin e English is a non pro-drop language. However, the orpus of English tweets is a non- ontrolled language whi h often la ks otherwise mandatory grammati al subje ts. ���� Tregex rule VP (V B|V BD|V BG|V BN |V BP |V BZ = verb) VP < (V BG = verb) ������������ ��������� Rule name Verb: base form Verb: gerund/Pr.Part. Verb: simple Progressive 1 Progressive 2 ROOT < (S < (V P < (V B|V BP |V BZ = verb)! < (V P < V BG))) V P < (V B|V BD|V BG|V BN |V BP |V BZ < be|am|′ m|is|′ s|are|′ re|was|were|been) < (V P < V BG = verb) ROOT < (S < (V P < (V B|V BG|V BP |V BZ < be|am|′ m|is|′ s|are|′ re) < (V P < V BG = verb))) Table 8.1: Rules involved in the sear h of progressive. 902 Ref. 1 2 3 4 5 �������� ������������� �� ��� ��������� �� �������012 Ref 1 2 3 4 5 Senten e type I am at route 66 Getting this liquor ready for next week Be happy for these moments, is your life Mika is going to be on BBC tonight Is now disappointed be ause he realises tomorrow we are going to Mostar Table 8.2: Senten es types pi ked out by rules of table 8.1. ���� ��������� ������������ must forms. 211 ompare the utteran es of a same verb in transitive and intransitive Presumably, verbs whose transitive/intransitive rate is near 1 are verbs whi h an parti ipate in the anti ausative onstru tion, making them very likely eventive. Table 8.3 (page 212), reports the rules employed to pi k verbs with a high probability of parti ipate in the anti ausative onstru tion. Table 8.4 (page 213), reports examples of senten es pi ked up by these rules. Rule 1 identies intransitive verbs, verbs without an embedded NP in their stru ture. Rule 2 pi ks those in whi h there is an embedded NP. Rule 3 identies redu ed senten es without a onjugated verb but with an embedded NP. Rule 4 pi ks embedded de larative senten es (that are introdu ed by a transitive de larative verb). Again, for ea h lexi al verb, the rate between utteran es extrapolated by rule 1 and the other rules denes the verb pla ement in the gradient. In order to nd verb frequen y, a sear h for VP has been pi ks all verbs in the ondu ted. It orpus, with no matter to tense. Sin e stative verbs in present simple refer to present ongoing situations, in addition to the usual habitual reading, we expe t that they o ur more frequently than eventive verbs. For this reason, a sear h for verbs in present tense has been ondu ted. Its output is a gradient from more probable stative verbs to more probable eventive verbs. In other words, the more frequently a verb appear in present simple, the more it is probable that the verb is stative. With the denition of dierent riteria, several stativity/eventivity gra- dients are produ ed. Namely, one for ea h rule, with the impli ation that a same verb an o upy dierent positions in dierent gradients. ple, in the ranking obtained with progressive rules, the verb For exam- obtain an be at rank 100 and in the ranking obtained with the anti ausative rules it be ranked 2000. Whi h ranking is the most meaningful? an An interpolation between gradients produ ed by dierent rules must be performed. Our ultimate aim is to obtain a single gradient whi h verbs present in the analyzed whi h an orpus. We ontains all lexi al an rea h it by means of an equation ontain the weight to be assigned to ea h rule. it is ne essary to identify whi h of all In other words, riteria is more performing in the identi ation of eventive/stative verbs. We need an independent measure of stativity/eventivity. This is obtained by the olle tion of human data by means of a semanti The semanti interpretation task. interpretation task is built on some English verbs and asks to English speakers their judgment about the stativity of verbs (please refer to the following se tion for exa t pro edure and instru tions). YES/NO value about stativity for ea h verb. Results are a Consequently, we obtain the �������� ������������� �� ��� ��������� �� �������212 Ref. 1 2 3 4 Rule name Tregex Rule Intransitive ROOT < (S < (V P < (V B|V BD|V BG|V BN |V BP |V BZ = verb))! < N P ) Transitive 1 ROOT < (S < (V P < (V B|V BD|V BG|V BN |V BP |V BZ = verb)) < N P ) Transitive 2 ROOT < (S < (V P < (V B|V BD|V BG|V BN |V BP |V BZ = verb) < (S < N P ))) Transitive 3 ROOT < (S < (V P < (V B|V BD|V BG|V BN |V BP |V BZ = verb) < (SBAR < W HN P ))) Table 8.3: Rules involved in the sear h of anti ausatives. ���� ������������ ��������� Senten e Be my home just for the day Yes, I'm eating all-bran at 4:00 Having my oee in the old while wat hing the sun limbing up Waiting what we will hear on a press next weekend Table 8.4: Senten es pi ked up by rules of Table 8.3. 312 Ref. 1 2 3 4 214 ������� �� ��������� ��� �� ������������� �������� independent measure that an be employed to weigh Tregex rules. 8.2.2 Semanti interpretation task We need independent measure in order to ompare results of Tregex rules sear hes. We provided it by means of a semanti 48 verbs were sele ted from utteran e an experiment. Of these 48 verbs, 24 were as likely eventive (the interpretation task. orpus list and employed to build hosen as likely stative and 24 omplete list is given in table 8.5 (page 216), along with senten es in whi h they were employed). The experiment onsists in a semanti interpretation task under a modal. Informants were asked to judge whether a senten e, verb must, represented a ommand (deonti about a matter of fa t (epistemi interpretation). The experiment is divided into two parts: tionnaire; (ii) the linguisti The so io-linguisti ontaining the modal interpretation) or an assumption (i) the so io-linguisti ques- part. questionnaire registers age, sex and residen e of par- ti ipants with an usual format. The linguisti part is omposed of 48 senten es: 24 senten es with an eventive verb and 24 senten es with a stative verb. Subje ts of both groups were equally divided into animate and inanimate nominals, i.e. that 24 senten es ontain an animate subje t and 24 an inanimate. Senten es Sin e generi must at the present tense. All senten es subje t + verbal omplex + dire t obje t and omplements. ontain the modal involve the s hema obje ts inuen e eventuality, we used quantized obje ts. Senten es below are examples of dierent (541) onditions of experimental items. ontains an animate subje t and a stative verb; (542) imate subje t and a stative verb; (543) eventive verb; (544) ontains an inan- ontains an animate subje t and an ontains an inanimate subje t and an eventive verb. (541) This hild must belong to Mary. (542) His answer must reveal his stupidity. (543) Sandra must plan her maternity leave. (544) The ouple must hange their wedding date. The experiment was uploaded on Ibexfarm, whi h was the a tual administering platform. Parti ipants were re ruited by means of Amazon Me hanial Turk (hen eforth MTurk). They were enrolled thanks to MTurk and that were redire ted to Ibexfarm in order to omplete the experiment. pants were paid 1,25 US dollars ea h at the Parti i- ompletion of the experiment. ���� �������������������� �������� ���������� 215 Parti ipants were asked to judge all 48 senten es, whi h were presented in random order (determined by IbexFarm). 25 (15 female) Ameri an English native speakers Mean age is 35,84 years (min. 24; max 69). ompleted the task. They were all residents in the US territory at the moment of the task. Results onrm the predi tions. From I01 to I24 verbs are stative (major- ity of answer assumption) and from I25 to I48 they are eventive (majority of answer  ommand). 8.3 Stativity/eventivity gradient produ tion Data obtained by Tregex rules and data obtained by our experiment be an now ompared. The goal is the denition of the most powerful Tregex rules ( alled also variables) by the attribution of dierent weights. This yields the denition of an equation whi h ombines the weight of the most powerful Tregex rules in order to obtain 100% a We ondu ted a logisti ura y. regression between values obtained by human beings as dependent variable ( alled: group 0 for statives and 1 for eventives) and values obtained by orpus sear h. Results show that we obtain 100% a ura y in verbal aspe t denition with three variables: VP, ration Progressive1 over VP and sum of intransitives minus sum of transitives. VP rule is important to get rid of frequen y ee t. over VP shows a very high positive tives), being a Ratio Progressive1 orrelation with group variable 1 (even- urate at 96% alone, and a ounting for a frequen y ee t. The dieren e between sum of intransitive rules and sum of transitive rules represents the ee t of transitivity. Results are reported in the regression equation in gure 8.3 (page 217). Results are interesting from two perspe tives. First, they provide a mathemati al tool that an be employed in NLP whenever the verbal aktionsart is at stake. Se ond, they are eviden e of the stativity/eventivity distin tion. Statisti s shows that some diagnosti s are better than others to dete t this aspe tual ambiguity. 8.4 This Con lusions hapter is part of a wider proje t by Dr. Copley (CNRS-SFL, Fran e) and Dr. Wol (Emory University, Georgia, US), whi h proposes an automati 216 ������� �� ��������� ��� �� ������������� �������� Ref. I01 I02 I03 I04 I05 I06 I07 I08 I09 I10 I11 I12 I13 I14 I15 I16 I17 I18 I19 I20 I21 I22 I23 I24 I25 I26 I27 I28 I29 I30 I31 I32 I33 I34 I35 I36 I37 I38 I39 I40 I41 I42 I43 I44 I45 I46 I47 I48 Verb Matter Belong Reveal Love Hinge Foster Bewilder Enthrall Buttress Regret Hate Cherish Know Need Crave Dislike Envy Deserve Dismay Detest Despise Own Believe Disappoint In rease Start Produ e Kill Change Tea h Provide Fall Keep Go Work Play Run Be ome Use Make Plan Move Leave Wait Break Write Fight Study Senten e The dis iplinary ommission de ision must matter to Sandra This hild must belong to Mary His answer must reveal his stupidity John must love this swimming pool This mathemati al problem's solution must hinge on this variable The ir umstan es must foster this type of rime The latest news from New York must bewilder the readers The magi ian must enthrall Robin This pillar must buttress the athedral's nave Sandra must regret John's leaving John must hate his neighbour Sandra must herish her po ketwat h Mary must know this answer John must need a ar Sandra must rave that phone Mary must dislike this ake John must envy his brother John must deserve that treatment John must dismay his parents Sandra must detest that ou h Mary must despise his behavior Sandra must own that pla e John must believe in the ghost Sandra must disappoint her brother Sandra must in rease her in ome John must start this poem Those workmen must produ e 2000 shirts That man must kill the hi ken The ouple must hange their wedding date The tea her must tea h the new song That magnate must provide 2000 gallons of water John must fall in that dit h Sandra must keep this door open Mary must go to the ower shop Sandra must work on Julia's birthday party John must play in the hampionship Mary must run the 2016 New York Marathon Sandra must be ome a s ientist John must use a pen Mary must make a milkshake Sandra must plan her maternity leave John must move to Los Angeles Mary must leave a message Sandra must wait for her sister John must bake twelve up akes Mary must write her PhD dissertation Sandra must ght those superstitions John must study four hapters Table 8.5: Senten es used in the experiment. ���� ����������� 217 ��������������������� ��������� ����� �������� ������ ����� ���������� � � ������� � �� � ����� � � �� ����� ������������������ ����� ������������������������ ������������������������� ������ � � ���� ���� �� ���� ������ ������������������������ ���� ���� ���� � ���� ����� ������������� �������� ���������� ���� � ���� � ����������� ���� ���� ���� � ���� ����� �������� ������� �������� ���� � ���� ���� ��������������������������������������������������������������������������������������� Figure 8.1: Regression Table for Tregex Rules. 218 ������� �� ��������� ��� �� ������������� �������� temporal interpretation of senten es. The issue of the identi ation of verbal eventuality plays a big role in temporal interpretation, sin e stative and eventive verbs are interpreted dierently in I des ribed synta ti orpus sear hes. ertain tenses. rules that target dierent eventualities in automati They produ e dierent gradients of stativity/eventivity. Furthermore, the pro edure of data interpolation is des ribed whi h is du ted between rankings and data obtained by means of a semanti pretation task of verbs under modal must, on- inter- ondu ted on 25 English native speakers. Results of a logisti regression are reported and produ e a orrelation equation that denes the most powerful variables in the identi ation of verbal aspe t. Chapter 9 Con lusion The present dissertation analyzes the synta ti deadje tival parasyntheti stru ture perspe tive, but for dierent reasons. verbs behavior of denominal and verbs. Both groups are interesting from argument Denominal parasyhtneti hallenge theories of argument stru ture in pseudo-resultative rations, while deadje tival parasyntheti ongu- verbs play an important role in the elu idation of the nature of aspe tual ambiguities and ausative statives. For this reason the dissertation is divided in two parts that share the ommon interest in argument stru ture. An in-depth ree tion about data in olle tion methodologies is presented hapter 1. I argue in favor of the use of stri ter experimental proto ols in the generative framework, parti ularly in ase of subtle interpretation judge- ments, in order to (i) avoid impre ise results and (ii) to improve s ienti ex hanges with other elds in methodology. I report: (i) for the ognitive s ien es whi h only employ stru tured ases in whi h the methodology was responsible olle tion of false results (Langendoen et al. 1970; Wasow & Arnold 2005; Gibson & Fedorenko 2013) and (ii) experimental proto ols that are employed in the present work or that would be useful in the syntax/semanti generative resear h (Ionin 2012; Gordon & Chafetz 1986; Bard, Robertson & Sora e 1996). However, I emphasize the importan e of introspe tion when a linguisti fa t has to be delineated, in the rst steps of investigation. After presentation of argument stru ture theories ( hapter 1) (Hale & Keyser 1993 and ; Ram hand 2008; Borer 2005), I adopt a general tionist approa h in whi h the lexi on is not onstru - ompletely emptied of synta ti information, and dierent readings of a same lexi al verbs are attributed to dierent synta ti stru tures. Furthermore, I adopted a for e-dynami approa h (Copley & Harley 2015) to ausation (se tion 7.8.1) whi h treats ausation as the result of appli ation of dynami I present the main lass of parasyntheti 219 for es. verbs in Italian in hapter 3. ������� �� ���������� 220 I report three theories on derivational steps responsible for parasynthesis (Darmester 1890; Ia obini 2004; S alise 1990; Crobin 1987), pointing out that none of them is able to explain the position of the prex whi h does not respe t the mirror prin iple. I demonstrate the root nature of the stem of parasyntheti verbs, and the semanti ausal ontribution of prexes. In parti ular, I propose that prexes are in the head position of a non-eventive relational proje tion aller rP (A edo-Matellan 2006), that sele ts the root and has in spe ier position the dire t obje t. The presen e of the rP made a verb ausative (Hoekstra 1998; S häfer 2008). The rst part of the dissertation fo uses on denominal parasyntheti whi h verbs an be paraphrased as (make) X be ome(s) an N , where N is the base and X is the Theme ( alled BN). In resultative hapter 4, I introdu e the pseudo- onstru tion (Levinson 2007 and .), whi h is adje tive that modies the impli it entity of the verb. dieren es between impli it and expli it ated in the omposed of an I expose stru tural reation verbs: the individual re- ourse of the event does not belong to the argument stru ture of the former, but does so in the ase of the latter. I argue for the impli it reation nature of BN verbs on the basis of three verbs (Clark & Clark 1979); (ii) they imply the riteria: (i) they are goal reation of a shadow argu- ment (Geuder 2000); (iii) they require the mandatory presen e of an ae ted obje t. glish In order to investigate whether Italian BNs behave like their En- ounterparts in the pseudo-resultative semanti onstru tion (PR), I interpretation task with 106 Italian native speakers. di ate that: (i) PR is grammati al in Italian in ondu t a Results in- ontext with expli it dire t obje t (545), sin e it re eives 85% of answers; and (ii) PR is the only possible interpretation in (545) ontext with a pronominal dire t obje t (546). Daria impilò i libri alti. Daria piled books high. (546) Dopo aver letto i libri, Daria li impilò alti. After having read the books, Daria piled them high. Sin e Italian expli itly marks gender on adje tives, grammati ality of Italian PR and results of the experiment onrm Levinson's (2007) analysis of pseudo-resultatives, in parti ular her treatment of adje tival agreement. In PR onstru tion, the adje tive agrees with the dire t obje t, even though it is not the modied entity, be ause the impli it entity, being a non- ategorized root, has to annot he k the φ features of the adje tive. Consequently, the adje tive he k them with the dire t obje t, the rst A magnitude estimation task was dieren e in a - ommanding DP. ondu ted in order to investigate the eptability between adje tival vs. adverbial modi ation in 221 pseudo-resultative onguration. Results show that adverbs are preferred to adje tives. Sin e BNs are resultative verbs and have at least two proje tions that an be modied by the adverb, I propose that adverbs s opes, narrow and wide. an have two The previous modies the resultative part, the latter the eventive part. (547) Daria ammu hia i vestiti disordinatamente. Daria sta ks the lothes untidily. a. As a result of the a tion of Daria, the lothes are in a untidy sta k. b. The a tion of Daria is untidy. With respe t to se ondary predi ates, Italian behaves in a slight dierent manner than other Roman e languages. Italian being a verb-frame language (Talmy 1991, 2000), we would expe t the absen e of strong resultatives in this language. However, a ording to Folli (2001), Italian val resultatives with a tivity verbs under ertain an form adje ti- ir umstan es, su h as the redupli ation of the adje tive (548). (548) Daria ha 3sg.pr. Daria have- martellato il hammer- det- part. sg.m. metallo piatto metal at- sg.m. piatto. at- sg.m. Daria hammered the metal at. We have seen that Italian an also form pseudo-resultatives. In order to investigate whether this is a pe uliar behavior in the Roman e panorama, I ondu t a semanti interpretation task involving PR onstru tion in Fren h with 44 Fren h native speakers (549). I dis over that PR is not generally a eptability improves signi antly for eptable in Fren h. three verbs ( However, its a empiler, `pile'; tresser, `braid'; tran her, `sli e'). I argue that this depends on the higher phonologi al transparen y. The three verbs entertain a dire t phonologi al base ( empiler, orresponden e with the noun built on the same root `to pile', pile, `a pile'), this phonologi al transparen y allows speakers to per eive the link between the impli it entity and the adje tive. (549) ? Claude a empilé les livres hauts. Claude piled the books high. In the se ond part of the dissertation, I analyze the behavior of another lass of parasyntheti verbs formed from adje tives with ausative semanti s ������� �� ���������� 222 (DPVs), and sometimes with a double aspe tual reading. I divide the further, depending on the lexi al semanti s of the root. are thus presented: lass Three sub lasses DPVs of form, DPVs of surfa e, psy hologi al DPVs. Chapter 7 studies the behavior of the rst two groups, leaving psy hologi al DPVs aside for future resear hes. In order to dene whi h DPVs alternate between an eventive and a stative reading, I onsider dierent stativity diagnosti s proposed in the literature. Chapter 6 reports and analyses synta ti and semanti stativity tests applied to Italian: ungrammati ality with progressive and imperative; to the narrative time, deonti /epistemi and future/present not the stativity onstraint. per se reading with modal I show that synta ti ontribution dovere, `must' tests seem to dete t but some related phenomena, thus these tests do not ount as reliable tools for dening stativity (Squartini 1990, 1998; Levin 2007; Bertinetto 2000). Con erning semanti tests, I show that: (i) stative verbs under modal entertain two possible interpretations, deonti while eventives only one, deonti generate a present and epistemi , (Giorgi & Pianesi 1997); (ii) stative verbs onstraint, while eventive verbs a future 2003; Condoravdi 2002); (iii) stative verbs do not onstraint (Katz ontribute to narrative time progress (Katz 2003; Dry 1983). In support of (ii), I ondu t a semanti interpretation test with 188 Italian native speakers; its results dierent interpretation of statives and eventives under modal and onrm that, whenever instru tions are useful in order to refute a parti ular analysis. semanti onrm the dovere, `must', lear, naive speakers an be In addition to synta ti and stativity diagnosti s, stative verbs are shown to dier from eventives in their pro essing orrelates. I report results of a self-pa ed reading test ondu ted by Gennari & Poeppel (2003), whi h pointed out that stative verbs are read signi antly slower than eventive verbs. Chapter 7 resorts to semanti diagnosti s in the analysis of aspe tual ambiguities of DPVs. I demonstrate that DPVs of form are always eventive, (550a) and (550b), while DPVs of surfa e an be eventive or stative. The dieren e is pointed out by the (in)anima y of the subje t, when DPVs of surfa e re eive an eventive reading the subje t is animate (551a), while a stative reading is available when the subje t is inanimate (551b). (550) a. Daria ha allargato il bu o nel muro. Daria enlarged the window. b. La mua ha allargato il bu o nel muro. The mold enlarged the hole in the wall. (551) a. Daria ha abbellito la amera nuova. Daria embellished the new bedroom. 223 b. Il divano ha abbellito la amera nuova. The sofa embellished the new bedroom. Next, I show that, even though they often are not systemati ally is involved; when o-generated: when a ausation is present, this reason, I adopt a for e-dynami 2015), whi h of an a hange, sin e o-o hange ausation an be produ ed or not. For approa h to ausation (Copley & Harley ausation even in the absen e ausation is generated by an energeti sub-event linked to an event argument (se tion 7.8.1). is involved in ausation and hange is produ ed, hange ount for the presen e of ur, for e and it is not a Energeti ausation ausative eventive verbs, sin e it is generated by the pres- en e of an energeti for e. In the ase of stative verbs, no energeti for e is produ ed in the situation. For this reason, I update the approa h by the introdu tion of abdu tion. Abdu tion is a virtual for e whi h is generated by the speaker's opinion. In other words, ausation in stative DPVs of sur- fa e (551b) is produ ed by the speaker who established a link between the existen e of the external argument and the state of the internal obje t: presen e of the sofa is related to the bedroom to be beautiful. the Consequently, ontrary to dynami ausation, the absen e of Sour e in the situation implies the impossibility of ausal link between it and the dire t obje t's state: if the sofa is not present in the situation, is beautiful be ause of something. the speaker The importan e of the speaker for stati by a pragmati annot state that the room ausation is further supported judge parameter (Larson 2005). It is introdu ed by the base root, and relativize statements on someone's opinion (the judge's opinion). Disagreement tests (se tion 7.11.1) show that the judge parameter an rela- tivize dierent parts of DPVs depending on their eventuality. (552) A. Cosa fa Giovanna? What does Giovanna do? B. Abbellis e la stanza. She's embellishing the room. C. Oh no, non la abbellis e per niente, quei quadri sono disgustosi! Oh no, she doesn't embellish it at all! That frames are really ugly. Person C does not disagree on the fa t that Giovanna is performing an a tion on the room, rather she disagrees about the result of her doing. When stative, it an relativize also the ausal part of the senten e. Stati ausation is generated by abdu tion, introdu ed by the speaker who is responsible for establishing the ausal link. In other words, stati by denition relative to a personal opinion. ausation is ������� �� ���������� 224 (553) A. What are these owers doing on the table? B. They are embellishing it. C. Oh no, they are not embellishing at all. I argue that the ausal meaning of both stative (559) and eventive DPVs (561) is stru turally determined by the presen e of a rP whi h involves the internal obje t (Hoekstra 1988; S häfer 2008; Folli & Harley 2005). Dieren e between stative and eventive readings is generated by a dierent avor of Eventive DPVs involve a dynami v. fun tional head, and are formed by dier- vbe ent avors of the same fun tional proje tions. The verbalizing head is and the proje tion in whi h merges the external argument is V oice ome auser . Sta- tive DPVs involve a stative fun tional head (predi ative head). The derivation of stative DPVs involves the verbalizing fun tional head whi h is a pred- vrelation) i ative head ( (557), and a fun tional head a ording to the responsible for the introdu tion of the right external argument, in the V oiceholder ) of stative DPVs it is Sour e ( V oicesour (555) V oiceholder = λs.holder(x, s)p(suc(s)) (556) �vbe (557) �vrelation� = λp λs. p(s) (558) ome � is ase (555). (554) e v = λf.source(x, f )p(f in(f )) = λp λf. p(f in(f )) Il divano anneris e la stanza. Sofa bla kens the room. V oicesour e Phsti (559) V oice′sour e he, sti DP il divano Voi esour ehst,he,stii v Prelationhsti v relationhst,sti rPst DP la stanza r a- √ nera 225 (560) Giovanni anneris e la stanza. John bla kened the room. V oice (561) auser Pheti V oice DP auserhe,f ti ' Giovanni Voi e v Phf ti auserhf t,he,f tii v be omehst,f ti rPhsti DP la stanza r a- √ nera In ase of stative ausatives, the result part does not determine a of the obje t, but a state whi h is not dependent on any event. hange This is expressed to the presen e of the predi ative head whi h does not introdu e hange an be derived. In this sense, derivations proposed for statives and eventives for e in the system; without energeti for e, no ausatives are similar and the dieren e resides in the avor of v. Some questions remain open, parti ularly: whether all are generated by the stru ture proposed in this work; onrmation of avor of little v ; the nature of ausative statives the morphologi al ausative psy h verbs and the origin of its pe uliarity. Stative/eventive division plays a ti s. Sin e it has important riti al role outside theoreti al linguis- onsequen es both on interpretation and gram- mati ality, it plays a big role in automati Chapter 8 aims to the automati glish verbs by means of synta ti friendly rules, and to the language pro essing. determination of the eventuality of Enrules that an be translated in parser- reation of a stativity/eventivity gradient of En- glish verbs. Not all stativity diagnosti s involve linguisti dete ted in automati means that an be orpora sear h. I was interested in determining the stativity diagnosti s useful in automati sear h whi h present some morphologi al means. In parti ular, I trans- lated into Tregex rules: progressive form; simple present; ausative/in hoative ������� �� ���������� 226 alternation. I ondu ted a sear h of these Tregex rules into orpus and a list of verbs frequen y for ea h rule is obtained. In order to reate a single gradient of English verbs, ordered from the most probable stative to the most probable eventive, I had to determine the weigh of ea h rule, its power in the sele tion of stative verbs. the frequen ies of verbs obtained by the value. This vale was obtained by means of a semanti de ision task with English native speakers about the deonti /epistemi verbs under modal must. ondu ted interpretation of Parti ipants were asked to judge whether 48 En- glish senten es (24 stative verbs, 24 eventive verbs) a I normalize orpus sear h with an independent ould be interpreted as ommand or an assumption. Results were used to determine the power of ea h rule in the determination of stativity. In parti ular, a logisti was regression ondu ted. An equation with 100% a ura y an be produ ed that evaluates and determines the probability of a verb to be stative. Su h an equation is most useful in all resear hes that investigate phenomena linked to verbal eventuality. The dissertation investigates dierent aspe ts of argument stru ture from dierent perspe tives (lexi al, synta ti , semanti and IT). Its main butions are: ree tion on methodologies about data of Italian pseudo-resultative denition of stative mati ontri- olle tion; investigation onstru tion; ree tion on stativity diagnosti s; ausation; reation of a synta ti gradient for the auto- determination of verb eventuality. Dierent questions remain open for future resear h and I look forward to answering them. Appendix A This appendix is about Part I. A.1 List of denominal parasyntheti verbs 227 Stru ture tr, pron intr tr tr intr, pron intr tr tr re tr tr tr tr tr tr tr pron intr tr intr pron tr re tr intr pron tr tr tr tr re tr tr tr re Translation to to to to to to to to to to to to to to to to to to to to to to to to to to to to to to amp make a noose roll in form of a one roll self in form of a one set into a heap make, to form a ouple make self into a ouple unify in a single organism tie sheaf make or to redu e smth in frayed anvas olle t into bundles olle t into bundles olle t into a wooden bundles olle t smth as in a ball of wool dispose self in a fetal position tangle up tangle self up (gurative) pla e in line pla e self in line pi k up in big quantity, to reate a mass reate a mass of people roll up a hank pile up in haoti form pile up in haoti form put together smth to reate a pair put together self to reate a pair redu e in a spheri al form join pie es of fabri s redu e in slavery redu e self in slavery 822 a ampare a appiare a arto iare a arto iarsi a atastare a oppiare a oppiarsi a orpare a ovonare aaldellare aardellare aastellare aas inare aggomitolare aggomitolarsi aggrovigliare aggrovigliarsi allineare allinearsi ammassare ammassarsi ammatassare ammonti hiare ammu hiare appaiare appaiarsi appallottolare appezzare asservire asservirsi (Continues on the next page) �� �������� Verb tr re tr pron intr re tr pron intr tr and intr tr tr tr tr tr tr tr tr tr tr tr e intr tr tr tr tr tr Translation to redu e smone under other's people will to redu e self under other's people will to take a tool's indi ator to zero to run out annihilate to olle t in bales to barbarize self to onvert into stone, to be ome harder to olle t, to dispose into a pile to be ome sour (gurative or not) to burn untill ashes to olle t, to organize in olumn to skewer to braid to ut into sli es to redu e in lumps to redu e smth rusty to redu e in rumbles to redu e in s raps to redu e in s raps for lothes, to ut the ollar o to redu e in thin layers to redu e in pie es to redu e in small pie es ���� Stru ture ����� ������������� ��������� �� ���� assoggettare assoggettarsi azzerare azzerarsi azzerarsi imballare imbambarirsi impietrire impilare ina etire in enerire in olonnare inlzare intre iare aettare aggrumare arrugginire sbri iolare sbrindellare sbranare s ollare sfaldare spezzare spezzettare 922 Verb �������� �� 230 A.2 Semanti interpretation task ITA EXPERIMENTAL ITEMS. Condition 1. A1 SENT: An he se non è una parru hiera, Maria intre ia i apelli stretti. Even though she's not a hair-dresser, Mary braid hair tight. QUEST: A partire dai apelli, Maria rea una tre ia stretta. From hair, Mary reates a tight braid. QUEST: A partire dai apelli stretti, Maria rea una tre ia. From tight hair, Mary reates a braid. B1 SENT: Quando prepara il salame di io olata, Maria sbri iola i bis otti ni. When she prepares the ake, Mary rumbles bus uits thin. QUEST: A partire dai bis otti, Maria rea delle bri iole ni. From bis uits, Mary reates thin rumbles. QUEST: A partire dai bis otti ni, Maria rea delle bri iole. From thin bis uits, Mary reates rumbles. C1 SENT: Per preparare i panini, Maria aetta il salame sottile. In order to prepare sandwi hes, Mary sli es salami thin. QUEST: A partire dal salame, Maria rea delle fette sottili. From salami, Mary reates thin sli es. QUEST: A partire dal salame sottile, Maria rea delle fette. From thin salami, Mary reates sli es. D1 SENT: Quando gio ano, i bambini in olonnano i lego storti. When they play, hildren olumn building-blo ks rooked. QUEST: A partire dai lego, i bambini reano delle olonne storte. From building-blo ks, hildren reate rooked olumns. QUEST: A partire dai lego storti, i bambini reano delle olonne. From rooked building-blo ks, hildren reate olumns. E1 SENT: Se hanno bisogno di posto, i bibliote ari ammu hiano i libri alti. If they need more spa e, librarians sta k books high. QUEST: A partire dai libri, i bibliote ari reano dei mu hi alti. From books, librarians reate high sta ks. QUEST: A partire dai libri alti, i bibliote ari reano dei mu hi. From high books, librarians reate sta ks. F1 SENT: Se non erano esperte nella latura, le donne aggomitolavano il otone las o. If they were not ning experts, women winded loose otton. QUEST: A partire dal lo di otone, le donne reavano dei gomitioli las hi. From otton string, women reated loose balls of wool. QUEST: A partire dal lo di otone las o, le donne reavano dei gomitoli. From a loose otton string, women reated balls of wool. G1 SENT: Quando non 'erano le ma hine, i bos aioli a atastavano la legna s ombinata. When ars didn't exists, lumberja ks dumped mixed-up wood. QUEST: A partire dalla legna, i bos aioli reavano delle ataste s ombinate. From wood, lumberja ks reated mixed-up heaps. QUEST: A partire dalla legna s ombinata, i bos aioli reavano ataste. From mixed-up wood, lumberja ks reated heaps. ���� �������� �������������� ���� ��� 231 H1 SENT: Alle fontane, le lavandaie impilavano i vestiti onfusi. At fountains, washerwomen piled mixed-up lothes. QUEST: A partire dai vestiti, le lavandaie reavano delle pile onfuse. From lothes, washerwomen reated mixed-up piles. QUEST: A partire dai vestiti onfusi, le lavandaie revano delle pile. From mixed-up lothes, washerwomen reated piles. I1 SENT: Alla festa dell'altro giorno, i bambini hanno spezzettato la io olata sottile. At the party of the other day, hildren broke-in-pie es the thin (bar of ) ho olate. QUEST: A partire dalla io olata, i bambini hanno reato dei pezzetti sottili. From ho olate, hildren reated thin pie es. QUEST: A partire dalla io olata sottile, i bambini hanno reato dei pezzetti. From thin (bar of ) ho olate, hildren reated pie es. L1 SENT: Nel numero 50 di Topolino, Paperon de' Paperoni ha ammonti hiato il denaro informe. In the n◦ 50 of Mi key Mouse magazine, S rooge M Du k piled up the shapeless ash. QUEST: A partire dal denaro, Paperon de' Paperoni ha reato dei monti informi. QUEST: A partire dal denaro informe, Paperon deP
aperoni ha reato dei monti. From ash, S rooge M Du k reated shapeless piles. From shapeless ash, S rooge M Du k reated piles. M1 SENT: Ieri, dopo averlo ra olto, il ontadino ha imballato il eno rotondo. Yesterday, after having harvested it, the farmer pa ked the hay round. QUEST: A partire dal eno, il ontadino ha reato delle balle rotonde. From hay, the farmer reated round pa ks. QUEST: A partire dal eno rotondo, il ontadino ha reato delle balle. From round hay, the farmer reated pa ks. EXPERIMENTAL ITEMS. Condition 2 For experimental questions, refer to Condition 1. In fa t, question senten es were maintained equal for ondition 1 and ondition 2. a2 An he se non è una parru hiera, quando to a i apelli, Maria li intre ia stretti. Even though she's not a hair-dresser, when she tou hes hair, Mary braids them thight. B2 Quando prepara il salame di io olata on i bis otti, Maria li sbri iola ni. When she prepares the akes with bis uits, Mary rumbles them thin. C2 Per preparare i panini, Maria prende il salame e lo aetta sottile. To prepare sandwi hes, Mary takes the salami and she sli es it thin. D2 Quando gio ano on i lego, i bambini li in olonnano storti. When they play with building-blo ks, hildren olumn them rooked. E2 Se hanno bisogno di posto, i bibliote ari spostano i libri e li ammu hiano alti. If they need more spa e, librarians move books and they piled them high. F2 Se non erano esperte nella latura del otone, le donne lo aggomitolavano las o. If they were not experts in ning the otton, women winded it loose. �������� �� 232 G2 Quando non 'erano le ma hine per la legna, i bos aioli la a nata. atastavano s ombi- When ma hine for wood didn't exists, lumberja ks dumped it mixed up. H2 Alle fontane, dopo aver lavato i vestiti, le lavandaie li impilavano onfusi. At fountains, after having washed lothes, washerwomen piled them mixed-up. I2 Alla festa dell'altro giorno, gio ando on la io sottile. olata, i bambini la hanno spezzettata At the party of the other day, playing with ho olate, hildren broke it in thin pie es. L2 Nel numero 50 di Topolino, sistemando il suo denaro, Paperon de' Paperoni lo ha ammonti hiato informe. In the n◦ 50 of Mi key Mouse magazine, arranging his ash, S rooge M Du k piled it up shapeless. M2 Ieri, il ontadino ha ra olto il eno e lo ha imballato rotondo. Yesterday, the farmer harvested the hay and he pa ked it round. FILLERS 1. SENT: Dopo i tornado, le persone abbandonano i villaggi distrutti. After tornados, people leave destroyed villages. QUEST: Le persone sono distrutte. People are destroyed. QUEST: I villaggi sono distrutti. Villages are destroyed. 2. SENT: Durante la guerra, i soldati inter ettano le omuni azioni ifrate. During the war, soldiers inter ept oded ommuni ations. QUEST: Le inter ettazioni sono ifrate. Inter eptions are oded. QUEST: Le omuni azioni sono ifrate. Communi ations are oded. 3. SENT: In al une ulture anti he, gli sposi addobbavano la asa nuova. In some an ient ultures, ouples adorned the new house. QUEST: L'addobbo era nuovo. The de oration was new. QUEST: La asa era nuova. The house was new. 4. SENT: Durante le riunioni diplomati he, i api di stato intrattengono dei dis orsi u iali. During diplomati meetings, presidents make o ial speakings. QUEST: L'intrattenimento è u iale. The making is o ial. QUEST: I dis orsi sono u iali. The speakings are o ial. 5. SENT: Durante la sua arriera, Giovanni ha strigliato i avalli rossi. During his areer, Jon urried red horses. QUEST: La strigliata era rossa. The urry was red. ���� �������� �������������� ���� ��� QUEST: I avalli erano rossi. Horses were red. 6. SENT: Come te ni a di difesa, i guerrieri smussavano le lan e aguzze. As defen e te hnique, warriors rounded sharp lan es. QUEST: La smussatura era aguzza. The round-making was sharp. QUEST: Le lan e erano aguzze. Lan es were sharp. 7. SENT: Durante il suo traslo o, Giovanni ha ingombrato il garage nuovo. During his moving, Jon en umbered the new garage. QUEST: L'ingombro era nuovo. The obstru tion was new. QUEST: Il garage era nuovo. The garage was new. 8. SENT: Essendo un esperto di profumo, Giovanni l'ha spruzzato buono. Being a perfume expert, Jon sprayed it good. QUEST: Lo spruzzo era buono. The spray was good. QUEST: Il profumo era buono. The perfume was good. 9. SENT: Giovanni era un sarto per spose, le abbigliava sempre eleganti. Jon was a brides tailor, he dresses them always elegant. QUEST: L'abbigliamento era elegante. The dress was elegant. QUEST: Le spose erano eleganti. Brides were elegant. 10. SENT: Giovanni ostruiva orologi, li asseblava minuti. Jon built lo ks, he assembled them tiny. QUEST: L'assemblaggio era minuto. The assembly was tiny. QUEST: Gli orologi erano minuti. Clo ks were tiny. 11. SENT: Gli assassini u idono le persone, le seppellis ono vive. Assassins kill people, they bury them alive. QUEST: La sepoltura è viva. The burial is alive. QUEST: Le persone sono vive. People are alive. 12. SENT: I onsiglieri preparano i vestiti della regina, li s elgono sontuosi. Queen's ounselors prepare queen's dresses, they hoose them sumptuous. QUEST: La s elta è sontuosa. The hoi e is sumptuous. QUEST: I vestiti sono sontuosi. Dresses are sumptuous. 233 �������� �� 234 13. SENT: Giovanni non ha fortuna on le donne, le ha in ontrate solo brutte. Jon does not have a han e with women, he met them ugly. QUEST: Gli in ontri erano brutti. Meetings were ugly. QUEST: Le donne erano brutte. Women were ugly. A.3 Magnitude estimation task EXPERIMENTAL ITEMS 1. a. Dopo aver lavato i apelli, Maria li intre ia stretti. b. Dopo aver lavato i apelli, Maria li intre ia strettamente. After having washed her hair, Mary braid them tight. After having washed her hair, Mary braid them tightly. 2. a. Prima di mangiare i bis otti, Mario li sbri io ia ni. Before eating bis uits, Mario rumbles them faint. b. Quando Mario usa i bis otti se hi, li sbri io ia nemente. Before eating bis uits, Mario rumbles them faintly. 3. a. Quando Mario mangia il salame, lo aetta sottile. When Mario eats the salami, he sli es it thin. b. Quando Mario mangia il salame, lo aetta sottilmente. When Mario eats the salami, he sli es it thinly. 4. a. Quando Mario spazza la polvere, la ammassa onfusa. When Mario sweeps the dust, he sta ks it rooked. b. Quando Mario spazza la polvere, la ammassa onfusamente. When Mario sweeps the dust, he sta ks it rookedly. 5. a. Quando Mario si toglie i vestiti, li impila onfusi. When Mario takes his lothes o, he piles them mixed-up. b. Quando Mario si toglie i vestiti, li impila onfusamente. When Mario takes his lothes o, he piles them mixed-up-ly. 6. a. Prima di mangiare la io olata, Mario la spezzetta sottile. b. Prima di mangiare la io olata, Mario la spezzetta sottilmente. Before eating ho olate, Mario breaks it thin. Before eating ho olate, Mario breaks it thinly. 7. a. Quando Mario ha molte ban onote, le ammonti hia verti ali. When Mario has many banknotes, he piles them verti al. b. Quando Mario ha molte ban onote, le ammonti hia verti almente. When Mario has many banknotes, he piles them verti ally. 8. a. Quando Mario lavora la lana, la aggomitola molle. When Mario knits the wool, he winds it loose. ���� ��������� ���������� ���� 235 b. Quando Mario lavora la lana, la aggomitola mollemente. When Mario knits the wool, he winds it loosely. 9. a. Quando Mario sposta i do umenti, li ammu hia aoti i. When Mario moves the do uments, he piles them haoti . b. Quando Mario sposta i do umenti, li ammu hia aoti amente. When Mario moves the do uments, he piles them haoti ally. 10. a. Dopo aver tagliato l'erba, Mario la imballa stretta. After having ut the grass, Mario pa ks it tight. b. Dopo aver tagliato l'erba, Mario la imballa strettamente. After having ut the grass, Mario pa ks it tightly. 11. a. Dopo aver riempito i barattoli, Mario li allinea verti ali. After having ll the ans, Mario lines them up verti al. b. Dopo aver riempito i barattoli, Mario li allinea verti almente. After having ll the ans, Mario lines them up verti ally. 12. a. Dopo aver tagliato i rami, Mario li spezza strani. After having ut bran hes, Mario brakes them strange. b. Dopo aver tagliato i rami, Mario li spezza stranamente. After having ut bran hes, Mario brakes them strangely. FILLERS 13. Dopo aver ra olto i ori, Mario li annusa gialli. After having pi ked up owers, Mario smells them yellow. 14. Quando prenota un tavolo, Mario lo grande hiede. When he books a table, Mario asks big it. 15. Quando mangia esoti o, Mario prende la inese zuppa. When he eats exoti , Mario takes the soup hinese. 16. Dopo aver osservato la frutta, Mario la molle ompra. After having observed fruits, Mario buys soft it. 17. Quando Mario si toglie i vestiti, li lava spor hi. When Mario takes lothes o, he washes them dirty. 18. Quando legge un libro, Mario lo tradu e pola o. When he reads a book, Mario translates it Polish. 19. Quando lan ia un sasso, Mario lo rompe grande. When he throws a stone, Mario brakes it big. 20. Dopo aver omprato le s arpe, Mario le indossa strette. After having bought the shoes, Mario wears them tight. 21. Dopo aver visitato il grande museo, Mario lo fotografa. After having visited the big museum, Mario photographs it. 22. Dopo aver esaminato un problema onfuso, Mario lo risolve. After having examined the mixed up problem, Mario solves it. �������� �� 236 A.4 Semanti interpretation FR EXPERIMENTAL ITEMS 1. SENT: Quand Marie rangeait la maison, elle amassait ses haussures haotiques. When Mary organized her house, she sta ked her shoes haoti . QUEST: À partir des haussures, Marie faisait des amas haotiques. From her shoes, Mary made haoti sta ks. QUEST: À partir des haussures haotiques, Marie faisait des amas. From her haoti shoes, Mary made sta ks. 2. SENT: Quand Zeus soue sur les nuages, il les amon elle énormes. When Zeus blow on louds, he sta ks them big. QUEST: À partir des nuages, Zeus fait des mon eaux énormes. From louds, Zeus makes big sta ks. QUEST: À partir des nuages énormes, Zeus rée des mon eaux. From big louds, Zeus makes staa ks. 3. SENT: Marie est oieuse, elle tresse les heveux serrés. Mary is a hair-dresser, she braids hair tight. QUEST: À partir de heveux, Marie fait une tresse serrée. From hair, she makes tight braids. QUEST: À partir de heveux serrés, Marie fait une tresse. From tight hair, she makes braids. 4. SENT: Quand Jean rentre à la maison, il entasse ses aaires désordonnées. When Jon omes home, he piled his belongings messy. QUEST: À partir de ses aaires, Jean fait des tas désordonnés. From his belongings, Jon makes messy piles. QUEST: À partir de ses aaires désordonnées, Jean rée des tas. From his messy belongings, Jon makes piles. 5. SENT: Quand elle était petite, Marie émiettait les bis uits ns. When she was a hild, Mary rumbled bis uits faint. QUEST: À partir des bis uits, Marie faisait des miettes nes. From bis uits, Mary made faint rumbles. QUEST: À partir des bis uits ns, Marie faisait des miettes. From faint bis uits, Mary made rumbles. 6. SENT: Quand Marie avait trop de livres, elle les empilait hauts. When Mary has too many books, she piled them high. QUEST: À partir des livres, Marie faisait des piles hautes. From books, Mary made high piles. QUEST: À partir des livres hauts, Marie faisait des piles. From high books, Mary made piles. 7. SENT: Pour la préparation des sandwi hs, Marie tran he le salami n. To prepare sandwi hes, Mary sli es the salami thin. QUEST: À partir du salami, Marie fait des tran hes nes. From salami, Mary makes thin sli es. QUEST: À partir du salami n, Marie fait des tran hes. From thin salami, Mary makes sli es. ���� �������� �������������� �� 237 8. SENT: Après la lature, les hommes pelotonnaient la laine épaisse. After having form a string, men made thi k balls of wool. QUEST: À partir de la laine, les hommes faisaient des pelotes épaisses. From wool, men did thi k balls. QUEST: À partir de la laine épaisse, les hommes faisaient des pelotes. From thi k wool, men made balls. 9. SENT: Quand elle était petite, avant de manger les bis uits Marie les émiettait ns. When she was a hild, before eating bis uits, Mary rumbled them thin. QUEST: À partir des bis uits, Marie faisait des miettes nes. From bis uits, Mary made thin rumbles. QUEST: À partir des bis uits ns, Marie faisait des miettes. From thin bis uits, Mary made rumbles. 10. SENT: Quand Jean essaye de ranger ses aaires, il les entasse désordonnées. When Jon tries to organize his belongings, he sta ks them messy. QUEST: À partir de ses aaires, Jean fait des tas désordonnés. From his belongings, Jon makes messy sta ks. QUEST: À partir de ses aaires désordonnées, Jean rée des tas. From his messy belongings, Jon makes sta ks. 11. SENT: Quand Marie oie les heveux, elle les tresse serrés. When Mary dresses hair, she braids them tight. QUEST: À partir des heveux, Marie fait une tresse serrée. From hair, Mary makes tight braids. QUEST: À partir des heveux serrés, Marie fait une tresse. From tight hair, Mary makes a braid. 12. SENT: Pour la préparation des sandwi hs, Marie a hète le salami et elle le tran he n. To prepare sandwi hes, Mary buys salami and she sli es it thin. QUEST: À partir du salami, Marie fait des tran hes nes. From salami, Mary makes thin sli es. QUEST: À partir du salami n, Marie fait des tran hes. From thin salami, Mary makes sli es. FILLERS 13. SENT: Quand il était en olère, Zeus envoyait le brouillard blan . When Zeus was angry, he sent white fog. QUEST: Pendant la olère, Zeus était blan . During his rage, Zeus was white. QUEST: Pendant la olère de Zeus, le brouillard était blan . During his rage, the fog was white. 14. SENT: Pendant l'é ole, les enfants é rivent sur le papier épais. During s hool time, hildren write on thi k paper. QUEST: Les enfants sont épais. Children are thi k. QUEST: Le papier est épais. Paper is thi k. �������� �� 238 15. SENT: Après la guerre, les personnes ont abandonné les villages dévastés. After the war, people leave destroyed villages. QUEST: Après la guerre, les personnes étaient dévastées. After the war, people were destroyed. QUEST: Après la guerre, le villages étaient dévastés. After the war, villages were destroyed. 16. SENT: Les onseillers préparent les vêtements pour la reine, ils les somptueux. hoisissent Queen's ounselors prepare queen's dresses, they hoose them sumptuous. QUEST: À ause du hoix, les vêtements sont somptueux. For the hoi e, dresses were sumptuous. QUEST: À ause du hoix, les onseillers sont somptueux. For the hoi e, ounselors were sumptuous. 17. SENT: Aux temps des rois, les gens buvaient l'eau marron. In the monar hy, people drank brown water. QUEST: Dans le passé, les gens étaient marrons. In the past, people were brown. QUEST: Dans le passé, l'eau était marron. In the past, water was brown. 18. SENT: Pendant la guerre, les soldats inter eptaient les ommuni ations odées. During the war, soldiers inter ept oded ommuni ations. QUEST: Pendant la guerre, les soldats étaient odés. Communi ations were oded. QUEST: Pendant la guerre, les ommuni ations étaient odées. Inter eptions were oded. 19. SENT: Les assassins tuent les personnes, ertains les enterrent vivantes. Assassins kill people, some of them bury them alive. QUEST: Les assassins sont vivants. Killers are alive. QUEST: Les personnes sont vivantes. People are alive. 20. SENT: Quand Jean allait à la plage, il lisait les romans longs. When Jon went at the seaside, he read long roman es. QUEST: Jean était long. Jon was long. QUEST: Les romans étaient longs. Roman es were long. Appendix B This appendix is about Part II. B.1 List of deadje tival parasyntheti verbs 239 Stru ture tr, tr tr tr, tr, tr, tr, tr, tr, tr, tr tr, tr, tr, tr tr, tr tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr tr, tr, pron intr, re pron intr intr, pron intr intr, pron intr intr, pron intr, re pron intr pron intr intr re pron tr, pron intr pron intr pron intr pron intr pron intr re intr intr, pron intr re pron intr pron intr pron intr pron intr, re re pron intr Translation to embellish to embellish to alm to abbreviate, to shorten demean to make ugly to blind to verify to alm to dull to lighten, to larify to train to sweeten to lower to make fertile, proli to wit to wit to ool to make kind, to make deli ate to repair to enlarge to broaden, to extend to lighten, to simplify to loosen to heer to distan e to extend, to lengthen to impoverish to modernize to soak 042 abbellire abbellare abbonire abbreviare abbrutire abbruttire a e are a ertare a hetare a iu hire a larare addestrare addol ire adimare aertilire aos iare aos ire areddare aggentilire aggiustare aggrandire allargare alleggerire allentare allietare allontanare allungare ammiserire ammodernare ammollare (Continues on the next page) �� �������� Verb tr, pron intr tr, pron intr tr, intr, pron intr tr, intr, pron intr tr, intr, pron intr tr, intr, pron intr tr, pron intr tr, re, pron intr tr tr, pron intr tr, intr, pron intr tr, pron intr tr tr, intr, pron intr tr, pron intr tr, pron intr tr, pron tr, pron intr tr, intr, pron intr tr, pron intr tr, intr, pron intr tr, intr, pron intr tr, pron intr tr, intr tr, pron intr tr, re tr, pron intr tr pron intr tr, intr, pron intr tr, pron intr tr, intr, pron intr tr, intr, pron intr Translation to to to to to to to to to to to to to to to to to to to to to to to to to to to to to to to soften soften make a id silen e bla ken bla ken add weight to atten make smaller deepen make or be ome ri h url make smone hoarse redden round make or be ome red hot atter make smth rough alm deafen deafen make or be ome sad make or be ome sad bring about, to ome true near, to get lose depress, to sadden revive lame make or be ome blue make or be ome less ivilized degenerate ���� Stru ture ����� ������������� ������������ �� ���� ammollire ammorbidire ammos iare ammutire annerare annerire appesantire appiattire appi olire approfondire arri hire arri iare arro hire arrossare arrotondare arroventare arruanare arruvidire asserenare assordare assordire attristare attristire avverare avvi inare avvilire avvivare azzoppare azzurare imbarbarire imbastardire (Continues on the next page) 142 Verb Stru ture tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, intr, pron intr intr intr, pron intr, pron intr, pron intr, pron intr, pron intr, pron intr, pron intr, pron intr, pron intr, pron intr, pron intr, pron pron intr intr, pron intr, pron intr, pron intr, pron intr, pron intr, pron pron intr pron intr intr, pron pron intr pron intr intr, pron pron intr intr, pron intr, pron intr intr intr intr intr intr intr intr intr intr intr intr intr intr intr intr intr intr intr intr intr intr intr Translation to to to to to to to to to to to to to to to to to to to to to to to to to to to to to to to embellish, to adorn whiten whiten make or be ome blond make or be ome a ras al make or be ome a bourgeois make or be ome a ras al make or be ome medio re darken make or be ome a beast make or be ome ugly make or be ome silly make or be ome miserable make or be ome stupid make or be ome miserable impregnate soften make or be ame pedanti make or be ome smaller make or be ome smaller make or be ome lazy impoverish get pra ti e enhan e rot exa erbate exa erbate embitter, to go sour embitter embitter dry up 242 imbellire imbian are imbian hire imbiondire imbirbonire imborghesire imbri onire imbro hire imbrunire imbrutire imbruttire immelensire immes hinire immin hionire immiserire immollare immorbidire impedantire impi iolire impi olire impigrire impoverire imprati hire impreziosire imputridire ina erbare ina erbire ina idire ina utire inagrire inaridire (Continues on the next page) �� �������� Verb tr, intr, pron tr, pron intr tr, pron intr intr, tr tr, intr, pron tr, intr, pron tr, pron intr tr, intr, pron tr, intr, pron tr, intr, pron tr, intr, pron tr, pron intr tr, intr, pron tr, intr, pron tr, intr, pron tr, intr, pron tr, intr, pron tr, intr, pron tr, intr tr, intr, pron tr, pron intr tr, intr, pron tr, intr, pron tr, intr, pron tr, pron intr tr, intr tr, intr, pron tr, intr, pron tr, intr, pron tr, intr, pron tr, pron intr intr intr intr intr intr intr intr intr intr intr intr intr intr intr intr intr intr intr intr intr intr Translation to to to to to to to to to to to to to to to to to to to to to to to to to to to to to to to embitter embitter make or be ome blue go white make or be ome wi ked make or be ome stupid make or be ome more ivilized make or be ome stupid make or be ome fero ious exa erbate loud intrigue bend weaken domesti ate sweeten harden make or be ome stupid make or be ome fetid weaken, to exhaust abate, to weaken make or be ome feeble tighten, to intensify thi ker soak weaken ool ool make the meat be ome high strengthen, to invigorate make or be ome lumsy ���� Stru ture ����� ������������� ������������ �� ���� inasprare inasprire inazzurrare in anutire in attivire in itrullire in ivilire in retinire in rudelire in rudire in upire in uriosire in urvire indebolire indo ilire indol ire indurire inebetire infetidire ina hire inevolire ino hire inttire infoltire infradi iare infralire infreddare infrigidire infrollire ingagliardire ingaglioare (Continues on the next page) 342 Verb Stru ture tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, intr, pron intr intr, pron intr intr, pron intr intr, pron intr intr intr, pron intr intr, pron intr intr, pron intr intr, pron intr intr, pron intr intr, pron intr intr, pron intr intr, pron intr pron intr intr intr, pron intr intr intr, pron intr re pron tr, re intr, pron intr intr, pron intr intr, pron intr intr, pron intr intr, pron intr intr, pron intr intr intr, pron intr intr, pron intr intr, pron intr pron tr, pron intr Translation to to to to to to to to to to to to to to to to to to to to to to to to to to to to to to to make or be ome lumsy make or be ome jealous make or be ome gentle yellowish make or be ome young make or be ome lumsy make or be ome greedy make or be ome slender enlarge, to in rease fatten make or be ome gray fatten make or be ome stupid annoy, to irritate make or be ome stupid grow wild, to make or be ome unso iable make or be ome stri t deafen dirty, to soil dirty, to soil make or be ome arrogant soften, to move warm frighten make or be ome stupid roil roil numb, to make sluggish sadden swollen dampen 442 ingagliore ingelosire ingentilire ingiallire ingiovanire ingore ingolosire ingra ilire ingrandire ingrassare ingrigire ingrossare ingrullire innervosire inottusire inselvati hire inseverire insordire insozzare insudi iare insuperbire intenerire intiepidire intimidire intontire intorbidare intorbidire intorpidire intristire inturgidire inumidire (Continues on the next page) �� �������� Verb tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, tr, intr, pron intr pron intr intr, pron intr pron intr intr, pron intr intr, pron intr pron intr pron intr pron intr intr, pron intr pron intr intr, pron intr pron intr intr, pron intr intr, pron intr pron tr intr, pron intr intr, pron intr intr, pron intr pron tr, intr, pron intr pron intr intr, pron intr pron intr pron intr intr, pron intr pron intr intr, pron intr pron intr pron tr pron intr pron intr Translation to to to to to to to to to to to to to to to to to to to to to to to to to to to to to to to age make or be ome true green make or be ome vermillion make or be ome oward make or be ome un ivilized immerse strengthen stien roughen make or be ome infertile make or be ome stupid lower whiten whiten parboil warm lighten lighten darken strengthen, to invigorate unburden rough- ut rough- ut broaden, to wider enlarge slim down halve bare make or be ome lazy ool down a little ���� Stru ture ����� ������������� ������������ �� ���� inve hiare inverare inverdire invermigliare inviglia hire inzoti hire inzuppare irrobustire irrigidire irruvidire isterilire istupidire sbassare sbian are sbian hire sbollentare s aldare s hiarare s hiarire s urire sgagliardire sgravare sgrezzare sgrossare slargare slungare smagrire smezzare snudare spigrire stiepidire (Continues on the next page) 542 Verb Stru ture tr, re tr, pron intr, re Translation to renew to devalue 642 sve hiare svilire �� �������� Verb ���� ���� �� ������������ ������������� ����� Interpretation under modal ITA STATIVES WITHOUT CAUSATIVE MEANING 1. Carla deve adorare il suo nuovo ollega. Carla must adore his new olleague. 2. Giulio deve amare il gelato al io olato. Giulio must love ho olate i e- ream. 3. Maria deve ammirare la nuova trasmissione televisiva. Maria must admire the new tv show. 4. Questo bell'anello deve appartenere a Maria. This beautiful ring must belong to Mary. 5. Giulio deve apprezzare le anzoni di Battisti. Giulio must appre iate Battisti's songs. 6. Carla deve onos ere il ontenuto del testamento di Maria. Carla must know Mary's will ontent. 7. Maria deve redere alle bugie di suo marito. Maria must believe in her husband's lies. 8. Giulio deve desiderare quelle s arpe in vetrina. Giulio must desire that shoes in the shop window. 9. Maria deve detestare quel divano marrone. Maria must detest that brown ou h. 10. Carla deve invidiare Maria. Carla must envy Maria. 11. La presenza del sole deve man are a Giulio. Giulio must miss the presen e of the sun. 12. Sandro deve odiare il aè ma hiato. Sandro must hate latte. 13. Sandro deve possedere quella ma hina sportiva rossa. Sandro must possess that red sport ar. 14. Sandro deve temere il ane del suo vi ino di asa. Sandro must fear his neighbour's dog. STATIVES WITH CAUSATIVE MEANING 15. I brutti sogni devono angos iare il bambino di Maria. Nightmares must anguish Mary's baby. 16. Questa tisana deve agitare Carla. This infusion must agitate Carla. 17. Il on erto deve annoiare Sandro. The on ert must annoy Sandro. 247 �������� �� 248 18. Lo spetta olo del mago deve divertire Giulio. The magi ian's show must amuse Giulio. 19. La giostra del par o deve impaurire Maria. The park arousel must s ary Mary. 20. Maria deve infastidire Carla. Maria must annoy Carla. 21. Le bolli ine sulla pelle di Carla devono inquietare Giulio. Blisters on Carla's skin must unsettle Giulio. 22. Il libro sulla storia d'Italia deve interessare Maria. The book about Italian history must interest Mary. 23. La puntura del alabrone deve intimorire Giulio. The hornet sting must s are Giulio. 24. L'assenza del presidente deve stupire gli impiegati. The prin ipal's absen e must astonish the o e workers. 25. Maria deve preo upare sua mamma. Mary must worry her mother. 26. La musi a ubana deve rallegrare la festa. Cuban musi must heer up the party. 27. L'aumento del prezzo dei bus deve s o iare molti utenti. Bus ti ket pri e in rease must bother many users. 28. Il olore di apelli di Sandro deve stupire Carla. Sandro hair olor must astonish Carla. EVENTIVES 29. Giulio deve agitare bene lo s iroppo. Giulio must shake the sirup properly. 30. Maria deve porre delle ondizioni pre ise. Mary must di tate pre ise onditions. 31. Il essibile deve spezzare la atena della bi i letta. The angle grinder must brake the bi y le's hain. 32. L'aumento delle tasse del 2017 deve azzerare le dierenze so iali. The 2017 tax in rease must reset so ial dieren es. 33. Sandro deve s iogliere del burro. Sandro must melt the butter. 34. Maria deve diventare una dottoressa. Mary must be ome a do tor. 35. Giulio deve guadagnare il suo primo stipendio. Giulio must earn his rst salary. 36. Carla deve vendi are la morte di suo fratello. Carla must avenge her brother's death. ���� ���� �� ������������ ������������� ����� 37. La erimonia di apertura deve intrattenere gli spettatori oreani. The opening eremony must entertain Korean spe tators. 38. Giulio deve lavorare alla sua tesi. Giulio must work on his dissertation. 39. Carla deve attare la sua asa in ampagna per un mese. Carla must rent her ountryside house for a month. 40. Maria deve pesare il pros iutto. Mary must weight the ham. 41. L'azienda deve importare 8 ontainer di pezzi di ri ambio. The so iety must import 8 ontainers of spare parts. 42. Sandro deve sostituire la sua ve hia automobile. Sandro must hange his old ar. 43. Maria deve votare il nuovo delegato sinda ale. Mary must vote the new labor union delegate. 44. Carla deve riferire la notizia a Giulio. Carla must refer the news to Giulio. 45. Giulio deve rubare mille euro dalla assaforte di suo papà. Giulio must steal 1000 euro from his father's safe. 46. La pro edura dis iplinare deve de lassare Sandro. The dis iplinary pro edure must downgrade Sandro. 47. La medi ina deve guarire Sandro. The ure must ure Sandro. 48. La manovra nanziaria deve azzerare il debito pubbli o. The nan ial law must reset the national debt. 49. La disinfestazione deve eliminare metà delle zanzare. The extermination must eliminate half of the mosquitos. 50. La legge deve abolire la s hiavitù. The bill must abrogate slavery. 51. Il ris aldamento autonomo deve rimpiazzare quello entralizzato. The independent heating system must repla e the entral one. 52. La ristrutturazione deve allontanare i due muri portanti. The renovation must distan e the two load-bearing walls. 53. Il dibattito televisivo di stasera deve ontrapporre gli avversari. The tv debate of tonight must ontrast the rivals. 54. La nuova giunta omunale deve distruggere il ve hio entro ommer iale. The new muni ipal oun il must eliminate the old mall. 55. Sandro deve avvelenare tutti i topi he abitano nel suo granaio. Sandro must poison all mi e living in his garden. 56. La ala di antidoto per il veleno deve risvegliare Giulio. The poison antidote phial must wake Giulio up. 249 250 Interpretation under modal ENGL �������� �� STATIVES 1. The dis iplinary ommission de ision must matter to Sandra. 2. This hild must belong to Mary. 3. His answer must reveal his stupidity. 4. John must love this swimming pool. 5. This mathemati al problem's solution must hinge on this variable. 6. The ir umstan es must foster this type of rime. 7. The latest news from New York must bewilder the readers. 8. The magi ian must enthrall Robin. 9. This pillar must buttress the athedral's nave. 10. Sandra must regret John's leaving. 11. John must hate his neighbour. 12. Sandra must herish her po ketwat h. 13. Mary must know this answer. 14. John must need a ar. 15. Sandra must rave that phone. 16. Mary must dislike this ake. 17. John must envy his brother. 18. John must deserve that treatment. 19. John must dismay his parents. 20. Sandra must detest that ou h. 21. Mary must despise his behaviour. 22. Sandra must own that pla e. 23. John must believe in the ghost. 24. Sandra must disappoint her brother. EVENTIVES 25. Sandra must in rease her in ome. 26. John must start this poem. 27. those workmen must produ e 2000 shirts. 28. That man must kill the hi ken. 29. The ouple must hange their wedding date. ���� ���� �� ������������ ������������� ����� 30. The tea her must tea h the new song. 31. That magnate must provide 2000 gallons of water. 32. John must fall in that dit h. 33. Sandra must keep this door open. 34. Mary must go to the ower shop. 35. Sandra must work on Julia's birthday party. 36. John must play in the hampionship. 37. Mary must run the 2016 New York Marathon. 38. Sandra must be ome a s ientist. 39. John must use a pen. 40. Mary must make a milkshake. 41. Sandra must plan her maternity leave. 42. John must move to Los Angeles. 43. Mary must leave a message. 44. Sandra must wait for her sister. 45. John must bake twelve up akes. 46. Mary must write her PhD dissertation. 47. Sandra must ght those superstitions. 48. John must study four hapters. 251 252 �������� �� Bibliography A edo-Matellan, V. 2006. Prexes in Latin and Roman e and the satellite-/verb-framed dis- tin tion. lingbuzz/000295. A edo-Matellan, V. 2012. a Adje tival resultatives ount. In ross-linguisti ally: Pro eedings of ConSOLE 17, Pp. a morphophonologi al 125. Adger, D. 2004. Core syntax: a minimalist approa h, Core linguisti s, reprinted edi- tion. Oxford: Oxford Univ. Press. OCLC: 178836363. Alexiadou, A., E. Anagnostopoulou, and F. S haefer 2006. The properties of anti ausatives pretation, rosslinguisti ally. Phases of inter- 91:187211. Alexiadou, A. and F. S haefer 2006. Instrument subje ts are agents or ausers. In Pro eedings of WCCFL, volume 25. Arad, M. 2003. Lo ality onstraints on the interpretation of roots: The ase of he- Natural Language & Linguisti Theory, 21(4):737 brew denominal verbs. 778. Baker, M. C. 1988. In orporation: a Theory of Grammati al Fun tion Changing. Chi ago: University Chi ago Press. Bard, E. G., D. Robertson, and A. Sora e 1996. Magnitude estimation of linguisti a eptability. Language, 72(1):32 68. Barwise, J. and J. Perry 1981. Situations and Attitudes. Journal of Philosophy, 253 78(11):668691. ������������ 254 Belletti, A. and L. Rizzi 1988. Psy h-verbs and theta-theory. ory, Natural Language & Linguisti The- 6:291352. Beretta, M. 1993. Morfologia. In ed., Pp. Introduzione all'italiano ontemporaneo, A. Sobrero, 193245. Roma: Laterza. Bertinetto, P. M. 1991. Il Verbo. In Pp. Grande grammati a italiana di onsultazione, volume 2, 13162. Bologna: Il Mulino. Bertinetto, P. M. 2000. The progressive in Roman e, as ompared with English. and Aspe t in the Languages of Europe, O. Dahl, ed., Pp. In Tense 559604. De Gruyter: [publisher unknown℄. Borer, H. 2005. Stru turing Sense Volume 2: The Normal Course of Events. Oxford University Press. Brennan, J. and L. Pylkkänen 2010. Pro essing psy h verbs: dierent types of semanti Behavioural and MEG measures of two omplexity. Language and Cognitive Pro esses, 25(6):777807. Chomsky, N. 1965. Synta ti stru tures, volume 4 of Janua Linguarum. Mouton: [pub- lisher unknown℄. Chomsky, N. 1986. Barriers, volume 13. MIT press. Chomsky, N. 1993. Le tures on government and binding: the Pisa le tures, in generative grammar, 7th ed edition. Berlin ; New York: Studies Mouton de Gruyter. Chomsky, N. 2000. New horizons in the study of language and mind. land℄ ; New York: Cambridge University Press. Chomsky, N. 2006. Language and mind. Cambridge University Press. Cambridge [Eng- ������������ 255 Chomsky, N. and A. Kasher 1975. Ree tions on Language. In tion and dis ourse. Pragmati s: Communi ation, intera - SUNY Press. Clark, E. V. and H. H. Clark 1979a. When nouns surfa e as verbs. Language, 55(4):767811. Language, Pp. Clark, E. V. and H. H. Clark 1979b. When nouns surfa e as verbs. 767811. Clifton, C., G. Franselow, and L. Frazier 2006. Amnestying Superiority Violations: Pro essing Multiple Questions. Linguisti Inquiry, 37:5168. Condoravdi, C. 2002. Temporal Interpretation of Modals. Modals for the present and for the past. In The Constru tion of Meaning, L. Casillas, eds., Pp. D. Beaver, S. Kaufman, and 5988. CSLI Publi ations, Center for the Study of Language and Information. Conroy, A., E. Takahashi, J. Lidz, and C. Phillips 2009. Equal Treatment for All Ante edents: How Children Su Prin iple B. Linguisti Inquiry, eed with 40(3):446486. Copley, B. and H. Harley 2015. A for e-theoreti Philosophy, framework for event stru ture. Linguisti s and 38(2):103158. Copley, B. and F. Martin 2014. Causation in grammati al stru tures, number 52 in Oxford studies in theoreti al linguisti s, rst edition edition. Oxford, United Kingdom: Oxford University Press. Copley, B. and P. Wol 2014. Theories of In ausation should inform linguisti Causation in grammati al stru tures, theory and vi e versa. B. Copley and F. Martin, eds., number 52 in Oxford Studies in Theoreti al Linguisti s. Oxford: Oxford University Press. Corbin, D. 1987. Morphologie dérivationelle et stru tura tion du lexique. Niemeyer. Tübingen: ������������ 256 Cowart, W. 1997. Experimental syntax: Applying obje tive methods to senten e judgments. CA, Sage: Thousand Oaks. Cro o Galèas, G. and C. Ia obini 1993. Lo sviluppo del tipo verbale parasinteti o in latino: i pressi ad , in , ex. Quaderni Patavini di Linguisti a, 12:3168. Croft, W. 1998. Event stru ture in argument linking. ments: Lexi al and Synta ti Constraints, Pp. 2163. In The Proje tion of Argu- M. Butt and W. Geuder, eds., Stanford, Calif: CSLI Publi ations, Center for the Study of Language and Information. Croft, W. 2012. Verbs: Aspe t and Causal Stru ture. Oxford England ; New York: OUP Oxford. Cruse, David 1970. A note on English ausatives. Linguisti Inquiry, 3:520528. Cupples, L. 2002. The stru tural hara teristi s experien er-verb senten es. and on-line omprehension of Language and Cognitive Pro esses, 17(2):125 162. Dabrowska, E. 2010. Naive v. expert intuitions: An empiri al study of a ments. The Linguisti Review, eptability judg- 27(1):123. D'Agostino, M. 2007. So iolinguisti a dell'Italia ontemporanea, Itinerari. Linguisti a. Bologna: Il mulino. De Mauro, T. 1972. Storia linguisti a dell'Italia unita. Bari: Laterza. Devoto, G., G. C. Oli, L. Serianni, and M. Trifone 2013. Il Devoto-Oli: vo abolario della lingua italiana 2014. Firenze: Le Monnier. OCLC: 879975356. Douven, I. 2011. Abdu tion. In ed. The Stanford En y lopedia of Philosophy, E. N. Zalta, ������������ 257 Dowty, D. 1991. Themati proto-roles and argument sele tion. Language, 67:547 619. Dowty, D. R. Word meaning and Montague grammar: the semanti s of verbs and times in generative semanti s and in Montague's PTQ, number v. 7 in 1979. Synthese language library. Dordre ht ; Boston: D. Reidel Pub. Co. Doyle, G. and R. Levy Pro eedings of the 34th Annual Meeting of the Berkeley Linguisti s So iety. 2008. Environment Prototypi ality in Synta ti Alternation. In Dry, H. 1983. The movement of narrative time. Journal of Literary Semanti s, 12:1953. Ernst, T. 2014. The syntax of adverbs. In The Routledge book of Syntax, Y. Sato, and D. Siddiqi, eds., Pp. A. Carnie, 108130. New York: Routledge. Ertes hik-Shir, N. and T. R. Rapoport 2005. The syntax of aspe t: deriving themati and aspe tual interpretation, number 10 in Oxford studies in theoreti al linguisti s. Oxford ; New York: Oxford University Press. Evans, J., J. L. Barston, and P. Pollard 1983. On the oni t between logi Memory and Cognition, and belief in syllogisti reasoning. 11:295306. Featherston, S. 2007. Magnitude estimation and what it wh- onstraints in German. Lingua, an do for your syntax: Some 115(11):15251550. Fedorenko, E., E. Gibson, and D. Rohde 2006. The nature of working memory Eviden e against domain-spe i Memory and Language, apa ity in senten e omprehension: working memory resour es. Journal of 54(4):541553. Ferreira, F. 2005. Psy holinguisti s, formal grammars, and Linguisti Review, 22:365380. ognitive s ien e. The ������������ 258 Fillmore, C. 1965. Indire t Obje t Constru tions in English and the Ordering of Transformations. Nederlands: Mouton. Fodor, J. 1970. Three reasons for not deriving `kill' from ` ause to die'. Inquiry, Linguisti 1:429438. Fodor, J., A. Bever, T. G. Garrett, and F. M The Psy hology of Language: An Introdu tion to Psy holinguisti s and Generative Grammar. M graw-Hill. 1974. Folli, R. Constru ting teli ity in English and Italian. 2001. PhD, Oxford, Oxford. Folli, R. and H. Harley 2005. Flavors of v. In Aspe tual inquiries, Pp. 95120. Springer Nether- lands. Fábregas, A. and R. Marín 2015. Deriving individual-level and stage-level psy h verbs in Spanish. Linguisti Review, The 32(2). Gennari, S. and D. Poeppel 2003. Pro essing omplexity. Cognition, omprehension pro esses: The ase of rela- orrelates of lexi al semanti 89(1):B27B41. Gennari, S. P. and M. C. Ma Donald 2009. Linking produ tion and tive lauses. Cognition, 111(1):123. Geuder, W. 2000. Oriented Adverbs. Issues in the Lexi al Semanti s of Event Adverbs. PhD, Tübingen. Gibson, E., T. Desmet, D. Grodner, D. Watson, and K. Ko 2005. Reading relative lauses in English. Gibson, E. and E. Fedorenko 2013. The need for quantitative methods in syntax and semanti s resear h. Language and Cognitive Pro esses, 28(1-2):88124. Gibson, E., S. T. Piantadosi, and E. Fedorenko 2013. Quantitative methods in syntax/semanti s resear h: A response ������������ to Sprouse and 259 Almeida (2013). Language and Cognitive Pro esses, 28(3):229240. Giorgi, A. and F. Pianesi 1997. in Tense and aspe t: from semanti s to morphosyntax, Oxford studies omparative syntax. New York: Oxford University Press. Goldhahn, D., T. E kart, and U. Quastho 2010. Building large monolingual di tionaries at the leipzig orpora olle - Pro eedings of the 8th International Language Ressour es and Evaluation (LREC'12). Cornell University. tion: From 100 to 200 languages. In Gordon, P. 1998. The Truth-Value Judgment Task. In syntax, Methods for assessing hildren's D. M Daniel, C. M Kee, and H. Smith Cairns, eds. Cambridge: MIT Press. Gordon, P. and J. Chafetz 1986. Lexi al Learning and Generalization in the Passive A quisition. Gordon, P. C., R. Hendri k, and M. Johnson 2004. Ee ts of noun phrase type on senten e Memory and Language, omplexity. Journal of 51(1):97114. Grimshaw, J. B. 1990. Argument stru ture, number 18 in Linguisti inquiry monographs. Cambridge, Mass: MIT Press. Grossmann, M. 2004. La formazione delle parole in Italiano. Tübingen: Niemeyer. Hale, K. L. and S. Keyser 1993. On argument stru ture and the lexi al expression of synta ti rela- In The View from Building 20: Essays in Linguisti s in Honor of Sylvain Bromberger. Cambridge: MIT Press. tions. Hale, K. L. and S. J. Keyser 2002. guisti Prolegomenon to a theory of argument stru ture, number 39 in Lin- inquiry monographs. Cambridge, Mass: MIT Press. Halliday, M. A. K. 1967. Notes on transitivity and theme in English, part 1. Linguisti s, 3:3781. Journal of ������������ 260 Harley, H. 1995. Subje ts, events and li ensing. PhD, MIT, Cambridge, MA. Harley, H. 2005. How do verbs get their names? denominal verbs, manner in orporation and the ontology of verb roots in english. In The Syntax of Aspe t, N. Ertes hik-Shir and Tova Rapoport, eds., Pp. 4264. Oxford: [publisher unknown℄. Harley, H. 2014. On the identity of roots. Theoreti al Linguisti s, 40(3-4):225276. Hartshorne, J. K., T. J. O'Donnell, Y. Sudo, M. Uruwashi, J. Snedeker, and J. Hartshorne 2010. Linking meaning to language: Verbs of psy hologi al state and the Pro eedings of the Thirty-Se ond Annual Conferen e of the Cognitive S ien e So iety. linking problem. In Hay, J., Kennedy, Christopher, and Levin, Beth 1999. In S alar Stru ture Underlies Pro eedings of SALT 9, Pp. Teli ity Matthews, in "Degree A hievements". Tania and D. Strolovit h, eds., 127144, Itha a, NY. Cornell University. Heider, H. and K. Netter 2013. Representation and derivation in the theory of grammar. New York: Springer. Higginbotham, J. 1997. Lo ation and ausation. Ho kett, C. F. 1947. Problems of morphemi analysis. Language, 24:414441. Hume, D. 1993. An Enquiry Con erning Human Understanding: with Hume's Abstra t of A Treatise of Human Nature and A Letter from a Gentleman to His Friend in Edinburgh, 2 edition edition. Indianapolis: Ha kett Publishing Company, In . Hume, D. 2012. A Treatise of Human Nature. Platform. CreateSpa e Independent Publishing ������������ 261 Hung, H. 1988. The stru ture of derived verbs and nouns in Malagasy: a synta ti a ount. Ia obini, C. 2004. La parasintesi. In mann, ed., Pp. La formazione delle parole in italiano, M. Gross- 165188. Tübingen: Niemeyer. Ia obini, C. 2010. Formazione delle parole. Ia obini, C. and G. Cro 1993. o Galèas Parasintesi e doppio stadio derivativo nella formazione verbale del latino. Ar hivio glottologi o italiano, 78:167199. Ionin, T. Resear h Methodologies in Se ond Language A quisition. A pra ti al guide, A. Ma key and S. Gass, 2012. Formal Theory-Based Methodologies. In eds. Malden, Oxford: Wiley-Bla kwell. Ja kendo, R. 1972. Semanti Interpretation in Generative Grammar. Cambridge, Mass., MIT Press. Ja kendo, R. 1983. Semanti s and Cognition. Cambridge: MIT Press. Ja kendo, R. 1990. Semanti Stru tures. Cambridge: MIT Press. Ja kendo, R. 1991. Parts and boundaries. Cognition, 41:945. Johnson, K. 1991. Obje t positions. Natural Language & Linguisti Theory, 9:577636. Katz, G. 2003. On the stativity of the English perfe t. In Perfe t Explorations, A. Alexiadou, M. Rathert, and A. von Ste how, eds., Pp. 205234. Walter de Gruyter. Kayne, R. 1983. Chains, Categories External to S and Fren h Complex Inversion. Natural Language & Linguisti Theory, 1:107139. ������������ 262 Kayne, R. 1984. Conne tedness and Binary Bran hing. Dordre ht: Foris. Keller, F. 1999. Review of The Empiri al Base of Linguisti s: Grammati ality Judgments and Linguisti Methodology, Carson T. S h&#xf ;tze. Logi , Language, and Information, Journal of 8(1):114121. Kennedy, C. and L. M Nally 2005. S ale stru ture and the semanti Language, typology of gradable predi ates. 81:345381. Kra ht, M. 2002. On the semanti s of lo atives. Linguisti s and Philosophy, 25:157 232. Kratzer, A. 1989. An investigation of the lumps of thought. Linguisti s and Philosophy, 12(5):607653. Kratzer, A. 1996. Severing the external argument from its verb. and the lexi on, J. Roory k and L. Zaring, eds., Pp. Kluwer A ademi In Phrase stru ture 109137. Dordre ht: Publishers. Krifka, M. 1992. Themati Relations as Links between Nominal Referen e and Tem- poral Constitution. Pp. 2953. Kyparsky, P. 1982. Word-formation and the Lexi on. Kansas. Lawren e. Lako, G. 1966. Stative adje tives and verbs in English. Langendoen, D. T., N. Kalish-Landon, and J. Dore 1973. Dative questions: a study in the relation of a mati ality of an English senten e type. Cognition, eptability to gram- 2:451477. Larson, R. 1988. On the Double Obje t Constru tion. Larson, R. 1995. Olga is a beautiful dan er. Linguisti Inquiry, 19:335391. ������������ 263 Lasersohn, P. 2005. Context Dependen e, Disagreement, and Predi ates of Personal Linguisti s and Philosophy, Taste*. 28(6):643686. Levelt, W. 1972. Some psy hologi al aspe ts of linguisti data. Linguistis he Beri hte, 17:1830. Levin, B. 2007. The Lexi al Semanti s of Verbs II: Aspe tual Approa hes to Lexi al Semanti Representation. Levinson, L. 2007. The Roots of Verbs. PhD, New York University, New York. Levinson, L. 2010. Arguments for pseudo-resultative predi ates. Linguisti Theory, Natural Language & 28(1):135182. Lewis, D. 1973. Causation. Journal of Philosophy, 70:556567. Marantz, A. 1984a. On the nature of grammati al relations. Marantz, A. 1984b. On the nature of grammati al relations. graphs Cambridge, Mass., Linguisti Inquiry Mono- (10):1339. Marantz, A. 1997. No es ape from syntax: priva y of your own lexi on. in Linguisti s, Don't try morphologi al analysis in the University of Pennsylvania Working Papers 4(2). Marantz, A. 2000. Words. Marantz, A. 2013. Verbal argument stru ture: Events and parti ipants. Lingua, 130:152168. Marantz, Ale 2005. Generative linguisti s within the The Linguisti Review, 22:429455. ognitive neuros ien e of language. ������������ 264 Marazzini, C. 2002. La lingua italiana: prolo stori o, Strumenti Linguisti a, 3. ed edi- tion. Bologna: Il Mulino. OCLC: 249292686. Martin, F. and L. Tovena 2012. How deadje tival verbs entail their adje tival base. Mateu, J. 2000. Why Can't We Wipe the Slate Clean? proa h to Resultative Constru tions. ti s, A Lexi al-Synta ti Ap- Catalan Working Papers in Linguis- 8:7195. Mittwo h, A. 2014. The purported Present Perfe t Puzzle. of Nouns and Verbs. In Meaning and Grammar Düsseldorf: Düsseldorf University Press. Nagata, H. 1987. Long-Term Ee ts of Repetition on Judgments of Grammati ality. Per eptual and Motor Skills, 5(1):265299. Napoli, D. J. 1992. Se ondary resultative predi ates in Italian. Journal of Linguisti s, 28(01):53. Nilsen, D. The Instrumental Case in English: Synta ti and Semanti Considerations. The Hague: Mouton. 1973. Noh, B. 2003. Themati i ations. Stru ture and Syntax: Revisiting English Depi tive Pred- Language Resear h, 39(1):2137. Numberg, G. 1995. Transfers of Meaning. Journal of Semanti s, 12:109132. Panagiotidis, E. P. 2011. Categorial features and ategorizers. The Linguisti Review, Pellegrini, G. B. 1977. Carta dei dialetti d'Italia. Pisa: Pa ini. Pesetsky, D. 1989. Language-parti ular pro esses and the earliness prin iple. 28(3). ������������ 265 Pesetsky, D. M. 1996. Zero syntax: Experien ers and as ades. MIT press. Podesva, R. and D. Sharma 2013. Resear h methods in linguisti s. Cambridge ; New York: Cambridge University Press. Pottier, B. Systématique des éléments de relation, 1962. klin ksie k edition. Paris: [publisher unknown℄. Pustejovsky, J. 1991. The syntax of event stru ture. Cognition, 41:4781. Pylkkänen, L. Events as Grammati al Obje ts. The Converging Perspe tives of Lexi al Semanti s and Syntax, C. Tenny and 2000. On stativity and ausation. In J. Pustejovsky, eds. Stanford, Calif: CSLI. Ram hand, G. 2008. Verb meaning and the lexi on: a rst-phase syntax, number 116 in Cambridge studies in linguisti s. Cambridge, UK ; New York: Cambridge University Press. Ram hand, G. 2015. Event stru ture and verbal de omposition. Ram hand, G. C. 1998. De onstru ting the lexi on. In and Compositional Fa tors, Pp. The Proje tion of Arguments: Lexi al 6596. Standford: CSLI Publi ations, Center for the Study of Language and Information. Rana, C. 2011. Comprehensive Nepali. Speaking and writing. Kathmandu: [pub- lisher unknown℄. Rappaport Hovav, M. and B. Levin 1998. Building Verb Meanings. In and Compositional Fa tors, Pp. The Proje tion of Arguments: Lexi al 97134. Standford: CSLI Publi ations, Center for the Study of Language and Information. Reinheimer-Ripeanu, S. 1972. Sux zéro? Revue Roumaine de Linguistique, 17:261269. ������������ 266 Reinheimer-Ripeanu, S. Les dérivés parasynthétiques dans les langues romanes. 1974. The Hague/Paris: Mouton. Renzi, L. 2010. Grande grammati a italiana di onsultazione. 2: I sintagmi verbale, aggettivale, avverbiale. La subordinazione, Strumenti Linguisti a e riti a letteraria, nuova ed edition. Bologna: Il Mulino. Ritter, E. and S. T. Rosen . Delimiting Events in Syntax. In and W. Geuder, eds. The Proje tion of Arguments, M. Butt Stanford, Calif: CSLI Publi ations, Center for the Study of Language and Information. Rohlfs, G. 1972. Studi e ri er he su lingua e dialetti d'Italia. Firenze: Sansoni. Rothmayr, A. 2006. The stru ture of stative verbs. PhD, University of Wien. S alise, S. 1984. Morfologia lessi ale. Clesp. S alise, S. 1990. Morfologia e lessi o. Bologna: Il Mulino. S haefer, F. 2008a. Middles as voi ed anti ausatives. In ume 37, P. PROCEEDINGS-NELS, vol- 183. S haefer, F. The syntax of (anti-) ausatives: external arguments in hange-ofstate ontexts, number v. 126 in Linguistik aktuell = Linguisti s today. 2008b. Amsterdam ; Philadelphia: John Benjamins. S hlesinger, I. 1989. Instruments as Agents: On the Nature of Semanti nal of Linguisti s, Relations. Jour- 25:189210. S hutze, C. and J. Sprouse 2013. Judgment data. In and D. Sharma, eds., Pp. University Press. Resear h Methods in Linguisti s, 2750. Cambridge ; New York: R. Podesva Cambridge ������������ 267 S hutze, C. T. 1996. The empiri al base of linguisti s. Chi ago: University of Chi ago Press. Serrano-Dolader, D. 2015. Parasynthesis in Roman e. In Word-Formation, an International Handbook of the Languages of Europe, P. O. Mueller, I. Ohnheiser, S. Olsen, and F. Rainer, eds., volume 1, Pp. 524536. Gruyter Mouton. Shibatani, M. 1976. The Grammar of Causative Constru tions. New York: A ademi Press. Shohamy, E. 1996. Competen e and performan e in language testing. In Performan e and Competen e in se ond language a quistition, G. Brown, K. Malmkjaer, and J. Williams, eds., Pp. 138151. Cambridge ; New York: Cambridge University Press. Simone, R. 2010. En i lopedia dell'italiano, ed. spe iale per la libreria edition. Roma: Ist. della En i lopedia Italiana. Smith, C. 1970. Jespersens `move and hange' lass and ausative verbs in English. Linguisti and Literary Studies: In Honor of Ar hibald A. Hill, 2:101109. Soroli, E. and M. Hi kmann 2011. Language and spatial representations in Fren h and in English: Spa e in Language: Pro eedings of the eviden e from eye-movements. In Pisa International Conferen e, Pp. 581597. Editri e Testi S ienti i. Sprouse, J. and D. Almeida 2010. A quantitative defense of linguisti methodology. Sprouse, J. and D. Almeida 2013. The empiri al status of data in syntax: Fedorenko. Language and Cognitive Pro esses, A reply to Gibson and 28(3):222228. Squartini, M. 1990. Contributo per la aratterizzazione aspettuale delle perifrasi ital- iane andare + gerundio, stare + gerundio, venire + gerundio. Uno studio dia roni o. STUDI E SAGGI LINGUISTICI, 30:117212. ������������ 268 Squartini, M. Verbal Periphrases in Roman e. Aspe t, A tionality, and Grammati alization, mouton de gruyter edition. Berlin/New York: [publisher 1998. unknown℄. Stephenson, T. 2007. taste. Judge dependen e, epistemi Linguisti s and Philosophy, modals, and predi ates of personal 30(4):487525. Stevens, S. S. 1956. The dire t estimation of sensory magnitudes-loudness. Journal of Psy hology, Ameri an 69:125. Talmy, L. 1976. Semanti tions, ausative types. In M. Shibatani, ed., Pp. The Grammar of Causative Constru - 43116. New York: A ademi Press. Talmy, L. Pro eedings of the Georgetown University Round Table on Languages and Linguisti s, 1985a. For e dynami s as a generalization over Pp. ausative. In 6785, Washington, DC. George Town University Press. Talmy, L. Papers from the Parasession on Causatives and Agent ivity at the Twenty-First Regional Meeting of the Chi ago Linguisti So iety, Pp. 293337, Chi ago. Chi ago 1985b. For e dynami s in language and thought. Linguisti In So iety. Talmy, L. 1988. For e dynami s in language and ognition. Cognitive S ien e, 12:49 100. Talmy, L. 1991. Paths to realization: A typology of event integration. papers in linguisti s, Bualo working 91:147187. Talmy, L. 2000. Toward a Cognitive Semanti s. MIT Press. Google-Books-ID: g7IoanNUNksC. Tenny, C. 1987. Aspe t and ae tedness. of Massa husetts at Amherst. Pro eedings of NELS 18. GLSA, University ������������ 269 Tenny, C. 1992. The Aspe tual Interfa e Hypothesis. In Lexi al Matters, I. Sag and A. Szabols i, eds. Standford: Center for the Study of Language and Information. Tenny, C. and J. Pustejovsky Events as grammati al obje ts: the onverging perspe tives of lexi al semanti s and syntax, number no. 100 in CSLI le ture notes. Stanford, 2000. Calif: CSLI Publi ations, Center for the Study of Language and Information. Thorstad, R. and P. Wol 2015. What Causal Illusions Might Tell us about the Identi ation of Causes. Van Voorst, J. 1992. The aspe tual semanti s of psy hologi al verbs. Philosophy, Linguisti s and 15:6592. Verkuyl, H. J. 1972. On the ompositional nature of the aspe ts. Dordre ht: D. Reidel ase of synta ti diusion. Publishing Company. Walla e, W. 1981. Obje t marking in the history of Nepali. A Studies in the linguisti s ien e, 11(2):107128. Washio, R. 1997. Resultatives, ompositionality and language variation. East Asian Linguisti s, Journal of 6:149. Wasow, T. and J. Arnold 2005. Intuitions in linguisti argumentation. Lingua, 115(11):14811496. Wol, P. 2003. Dire t events. ausation in the linguisti Cognition, oding and individuation of ausal 88(1):148. Wol, P. 2007. Representing eral, 136(1):82111. ausation. Journal of Experimental Psy hology: Gen-