DRT’s treatment of inference and presupposition as a source of semantic enrichment

279 DRT’s ea men o in e ence and p esupposi ion as a sou ce o seman ic en ichmen F ancisco J. Salgue o Lamilla Uni e sidad de Se illa [email p o ec ed] 1 In e en ial seman ic en ichmen in discou se in e p e a ion The Discou se Rep esen a ion Theo y (DRT) (Kamp & Reyle 1993) p o ides an expla- na ion o he seman ic en ichmen p ocesses ha a e p esen in discou se, by means o he s uc u e o he successi e ep esen a ions associa ed o a conc e e sequence o u e ances. In he case o p onominal anapho a o inde ini es, in e p e a ion is jus possible in ela ion wi h hose elemen s p esen in discou se o in associa ed da abases. Each Discou se Rep e- sen a ion S uc u e (DRS) can be seen as a da abase wi hin which ce ain in o ma ion abou he wo ld is codi ied. So, when he speake codi ies a message, we can hink on a selec ion o lexical elemen s sui able o ep esen ing he e e ed indi iduals and he p ope ies hey ha e, he ela ions hey hold in a conc e e desc ip ion o he wo ld and e en he p ope ies and ela ions be ween his desc ip ion o he wo ld and o he ones. All hese elemen s a e combined using ce ain ules; so he hea e ob ains om he speake a se ies o linguis ic da a o which he mus apply his g amma ical, logical and encyclopaedic knowledge in o de o be able o decodi ying he comple e in o ma ion. In a e y gene al way, he ollowing scheme could be an illus a ion o discou se in e p e a ion p ocess: Rules o g amma Rules o logic Knowledge o he wo ld Discou se in e p e a ion Figu e 1 G amma , logic and encyclopaedic knowledge appea in his scheme as sepa a e mod- ules wi h a unique and inal ela ion: he in e p e a ion o he speake u e ances. Ne e he- less, he e a e enough easons o hink ha he es ablished ela ions among hese modules a e close . Wi hou discussing ques ions abou syn ac ic o mo phological iconici y and hose logical aspec s ha a e mo e e iden in na u al language g amma , we can conside 280 ha he in e p e a ion o any linguis ic u e ance es s on he applica ion o all hese knowl- edge modules o he in e media e ep esen a ions o discou se ha a e necessa y o e alu- a ing he speake u e ances. Conside he scheme in Figu e 2, mo e complex han he o - me one: Rules o G amma (Linguis ic Knowledge) Rules o Logic Meaning pos ula es Rules o lexical in o ma ion Si ua ion & Con ex (Knowledge o he wo ld) Anapho ic a iable ins an ia ion ules Discou se (P e ious in o ma ion) U e ance In e p e a ion Figu e 2 As i can be seen, he ins an ia ion o anapho ic a iables is conside ed as an in e medi- a e module o he in e p e a ion o u e ances. The e o e, he in e p e a ion o anapho a (p onouns, inde ini e noun ph ases, de ini e desc ip ions, ela i e clauses, spa ial o empo- al deic ic e ms, ense and aspec o e bs, e c.) is he common elemen o he con luence o he o he modules. Hence he impo ance o anapho ic in e ence in DRT and ano he dynamic ames o discou se in e p e a ion. In he in e p e a ion o discou se, i is a e o ha e a unique da abase o es ablishing i s global meaning. Se e al ela ed da abases a e usually needed o ob ain a model. The gi en ela ions among ep esen a ions o necessa y da abases o he e alua ion o discou se u - e ances a e a away om hose in classical model heo y. Concei ed in his manne as linguis ic da abases, he di e en ep esen a ions o discou se a e simila o possible wo lds. So hey can be seen as momen s, gi ing a empo al in e p e a ion o discou se ep- esen a ions; o hey can be seen as successi e s a es o he speake /hea e ’s mind, and hen we ha e ha discou se ep esen a ions a e simila o K ipke’s in o ma ion s a es. Wha e e he case could be, he ep esen a ions o discou se a e in e p e ed as pa ial models o eal- i y, and he e o e hey can be ea ed as Hin ikka se s (Salgue o 1994). This pe spec i e allows us dealing wi h discou se in e p e a ion as a so o logical cal- culus, a se ies o in e ences on he successi e ep esen a ions o discou se ha ake he 281 hea e o he meaning o i s cons i uen u e ances. Such in e ences depend on he ules o he g amma ical and logical modules, and on he hea e ’s knowledge o he wo ld. Bu he e is e en ano he decisi e ac o ha in e enes in hese in e ences: p esupposi ions. 2 P esupposi ions and DRT The no ion o p esupposi ion has been discussed since An iqui y, wi hou ha ing eached a consensus abou i s linguis ic o logical na u e up o he day. We can see i as a seman ic ela ion be ween sen ences and de ine i saying ha a sen ence B p esupposes ano he sen ence A i A implies B o i he u h alue o A is a condi ion o he u h alue o B. I is also usual o unde s and he no ion o p esupposi ion p agma ically as a se o a i udes and no ices, being used by language use s in hei speech ac s. In his case, we e e o speake /hea e ’s p esupposi ions ins ead o sen ence p esupposi ions. Bu he e is no opposi ion be ween hese wo pe spec i es because we can ea p esupposi ion as a one mo e pa o he in e p e a ion o he u e ances ha cons i u e discou se in a conc e e ep- esen a ion o i s meaning, and hen bo h pe spec i es a e sa is ied: he seman ic and he p agma ic one. This signi ies ha we will no ind a well de ined no ion o p esupposi ion in he spe- cialised li e a u e, bu a se o linguis ic da a ha mus be explained in a heo y abou dis- cou se in e p e a ion (Le inson 1983). Typical examples o hese linguis ic da a a e sen- ences like: (1) Ha e you s opped hi ing you wi e? (2) The ac ual king o F ance is no bald In (1) we ha e ha he e bal aspec and he possessi e you p esuppose ha he hea e has a wi e and ha he hi s he . In (2) we a e in on o a classical exis ence p esupposi ion in oduced by he de ini e de e mine he. Bo h sen ences a e classical examples o p esup- posi ions, bu hey a e no he only ones. We can classi y he linguis ic elemen s ha can be he sou ce o p esupposi ions in discou se, ollowing (Bea e 1997): 1. De ini e noun ph ases, in oduced by he de ini e a icle, demons a i es o posses- si es, p ope names o e en ela i e clauses. 2. Quan i ied noun ph ases, in oduced by a quan i ica ional de e mine . 3. Fac i e e bs o e bal nouns and noun ph ases ela ed o hem, such as ejec , know, he ac ha , he acquain ance o , e c. In his case we can dis inguish be- ween cogni i e ac i es, ela ed o he knowledge o he ac s, and emo i e ac- i es, ela ed o emo ional a i udes owa ds he ac s. 4. Cle s (i -, wh- o pseudo-cle s), like I was no John who called you up, p esup- posing someone called you up. 5. Wh-ques ions ha p esuppose he exis ence o an en i y sui able o an answe , like Wha colou is he sna k?, p esupposing he exis ence o a “sna k”. 6. Coun e ac uals, p esupposing he alsi y o he an eceden . 7. Lexical p esupposi ions. Fo ins ance, dog p esupposes animal. 8. Ac ion e bs, like s op doing some hing, aspec ual o empo al modi ie s, like be- o e o a e in ce ain sen ences ( . g .: A e ha ing you b eak as , go ou o school, p esupposing you a e ha ing b eak as ). 282 9. I e a i e exp essions, like again, o he p e ix e-, p esupposing some kind o epe- i ion. 10. Some discou se connec i es, like al hough, bu , because. Casuis y is wide and we do no p e end o ha e a lis wi h e e y sou ce o p esupposi- ion in discou se. Bu we can see now ha p esupposi ions in he sen ences (1) and (2) depend on e bal aspec and ense, s udied, o ins ance, in (Kamp & Reyle 1993:483 .), and on he in e p e a ion o noun ph ases and he quan i ica ional scope o he de e mine s. This las one is he mos usual sou ce o exis en ial p esupposi ion in discou se and we will analyse i in his pape . 2.1 Accommoda ion o p esupposi ions in DRSs I is e iden , as we ha e men ioned be o e, ha p esupposi ions play an impo an ole in discou se in e p e a ion. They a e indeed a pa in he seman ic and logical s uc u e o he discou se u e ances and, he e o e, hey mus appea in any ep esen a ion o ha s uc u e. So we need a me hod o making p esupposi ions explici in Discou se Rep e- sen a ion S uc u es (DRSs), since hey a e he way in which DRT ep esen s discou se in e p e a ion. We a e adop ing an de Sand ’s me hod, consis ing in he explici ex ension o discou se con ex in o de o include in he co esponden DRS all he p esupposi ions in e ening in he in e p e a ion o he u e ances ( an de Sand 1992, 1993). This me hod is called accommoda ion. Accommoda ion consis s in he addi ion o discou se e e en s and condi ions o a DRS. This addi ion is made in any o he accessible ep esen a ions ha con ain he discou se elemen ha igge s he p esupposi ion. When a igge elemen causes a p esupposi ion wi hou an an eceden , accommoda ion will consis in he ans e ence o discou se ma ke s and condi ions o he p esupposi ion om he ep esen a ion o he igge o an accessible DRS. The e o e, we can dis inguish among local accommoda ion (when he addi ion is made in he same DRS o he igge ), global accommoda ion (when he addi ion is made in he main DRS ha includes all he o he ep esen a ions), and in e media e accommoda ion (when he addi ion is made in ano he DRS, di e en om he main one and accessible om he DRS whe e he igge elemen appea s) ( an de Sand 1992). T igge s a e seen as anapho ic elemen s in discou se in e p e a ion because hese ele- men s canno be in e p e ed wi h independence o he s uc u al ela ion es ablished be- ween he posi ion in which hey a e ep esen ed in a DRS and he place whe e he ep e- sen a ions o hei an eceden s a e: an accessible DRS, like in he case o p onominal anapho a’s an eceden s. Le us see an example aken om (Bea e 1997): (3) Ma y doesn’ ealise ha somebody is escaping. The co esponden DRS (DRS1) includes he p esupposi ion somebody is escaping, ha will be ep esen ed wi hin a double line box. This p esupposi ion ma ks he exis ence in he discou se uni e se o an indi idual z and e i ies he condi ion is_escaping(z). The ma ke w can be ins an ia ed as z because he p esupposi ion is accessible om i s DRS. We can compa e (DRS1) wi h he in e p e a ion o he sen ence (4), ep esen ed by (DRS2): (4) F ed is escaping bu Ma y doesn’ ealise somebody is escaping. 283 (DRS1) y Ma y(y) z is_escaping(z) w ealises(y, is_escaping(w) ) (DRS2) x,y F ed(x) is_escaping(x) Ma y(y) z is_escaping(z) w ealises(y, is_escaping(w) ) The main DRS in (DRS2) is accessible om he nega ed DRS, so he ma ke z can be ins an ia ed by means o he ma ke x. The e o e, he uni e se o he p esupposi ion can be 284 eached om he global uni e se o discou se, as he condi ion o he p esupposi ion is accessible like a global condi ion. This means ha he p esupposi ion has an an eceden and, he e o e, as i does no play ano he ole in (DRS2), i could be elimina ed sal a in- e p e a ione. When we elimina e a p esupposi ion by means o i s esolu ion in o an an e- ceden as in (DRS2), o when we accommoda e i in o a DRS accessible om he igge ’s ep esen a ion, we say ha he p esupposi ion is cancelled. Bu , how does he accommoda- ion me hod wo k in ela ion wi h he p esupposi ions in a DRS? Conside he ollowing sen ence: (5) I Ma y chose he She y hen she ealises i is a good wine. In he ollowing DRSs, we will simpli y he discou se ma ke s using indi idual con- s an s m and s, deno ing Ma y and She y espec i ely, and he bidimensional s uc u e will be ep esen ed linea ly o sa ing space in he page. Then, in (DRS3) he signs [ and ] a e single line box limi s and he signs { and } a e double line box limi s. An emp y box [ ] means ha he e is no indi idual ma ke o ha ep esen a ion. (DRS3) [ [m,s] [ [ ] chose(m,s)] [ [ ] { [ ] good_wine(s) } ealises(m, [ [ ] good_wine(s) ] ) ] ] In o de o ha e a DRS whe e no p esupposi ion appea s wi hou an an eceden , we mus mo e he DRS ep esen ing he p esupposi ion owa ds any o he si es ha a e acces- sible om he igge ’s DRS, ha is o say: om he ep esen a ion o he sen ence Ma y ealises ha he She y is a good wine. This mo emen p oduces h ee di e en DRSs depending on he accommoda ion ype: Global accommoda ion: The She y is a good wine and i Ma y chose i , she ealises i is a good wine. (DRS4) [ [m,s] good_wine(s) [ [ ] chose(m,s)] [ [ ] ealises(m, [ [ ] good_wine(s) ] ) ] ] In e media e accommoda ion: I he She y is a good wine and Ma y chose i hen she ealises i is a good wine (DRS5) [ [m,s] [ [ ] good_wine(s) chose(m,s)] [ [ ] ealises(m, [ [ ] good_wine(s) ] ) ] ] Local accommoda ion: I Ma y chose he She y hen i is a good wine and she ealises i is a good wine. (DRS6) [ [m,s] [ [ ] chose(m,s)] [ [ ] good_wine(s) ealises(m, [ [ ] good_wine(s) ] ) ] ] Each one o hese h ee DRSs supposes a di e en in e p e a ion o he exis ing ela ion be ween he sen ence (5) and he p esupposed sen ence The She y is a good wine. Now we can ask wha kind o accommoda ion is he bes one o he in e p e a ion o an u e ance like (5) and y o es ablish a p io i y. Fi s , we mus y o ins an ia e he p esupposi ion ma ke s in he ac ual accommoda ion. Tha is wha is called esolu ion. Bu esolu ion is possible only when we ha e accessible ma ke s o he ins an ia ion, om he DRS con- aining he p esupposi ion igge up o he main DRS. When we canno ind ma ke s o he esolu ion hen we mus appeal o accommoda ion, bu now going in he e e se way: 285 om he main DRS up o he igge ’s DRS. This means he p io i y is esolu ion and hen, when esolu ion is no possible, global accommoda ion, in e media e accommoda ion and inally local accommoda ion. This p ocess has ce ain cons ain s ha can be ound in ( an de Sand 1992) o in (Bea e 1997). 2.2 Exis en ial p esupposi ions and DRSs We can now eco e he classical example by Be and Russell o illus a e he ea - men o exis ence p esupposi ions in DRT: (2) The ac ual king o F ance is no bald The p oblem o P edica e Logic is whe he his sen ence is ue o alse because he e is no en i y in he domain o discou se ha could be conside ed as a good ins an ia ion o he indi idual a iable in o de o sa is y he p edica e o_be_ he_king_o _ ance. Ne e heless, Russell said ha he logical o m o his sen ence p esupposes he exis ence o such an indi idual: [2] x(king_o _ ance(x) bald(x)) I is e iden ha i [2] is ue hen we can in e [2’], ha is o say: The e is a king o F ance; bu i [2] is alse hen we can in e [2”], ha is o say: All he kings o F ance a e bald. [2’] x(king_o _ ance(x)) [2”] x(king_o _ ance(x) bald(x)), To a oid his kind o undesi able p esupposi ions, Russell p oposed his heo y o de i- ni e desc ip ions, in oducing a logical ope a o as a desc ip o . In his heo y, wo possi- ble scopes we e pos ula ed o nega ion: a es ic ed scope [2*] (co esponding o he global accommoda ion o he p esupposi ion) and a gene al scope [2+] (co esponding o he local accommoda ion): [2*] x[king_o _ ance(x)]( bald(x)) [2+] ( x[king_o _ ance(x)](bald(x))) (DRS7) Global accommoda ion:(DRS8) Local accommoda ion: x king_o _ ance(x) x king_o _ ance(x) bald(x) bald(x) 286 In o dina y discou se, o a oid he undesi able e ec p oduced by he exis en ial p e- supposi ion in (2), we can en ich ou u e ance wi h an explana ion in he ollowing way: (6) The ac ual king o F ance is no bald. The e is no ac ual king o F ance. The co esponding DRS o (6) mus include, howe e , he exis en ial p esupposi ion and deal wi h i in he adequa e manne , esol ing i s discou se ma ke s o accommoda ing i in he co esponding si e inside he DRS: (DRS9) [ [ ] [ [ ] bald(x) { [x] king_o _ ance(x) } ] [ [y] king_o _ ance(y) ] ] The wo possible accommoda ions o he p esupposi ion in (DRS9), a e lea ing he possibili y o esolu ion o he ma ke x aside, a e he ollowing ones: (DRS10) [ [x] king_o _ ance(x) [ [ ] bald(x) ] [ [y] king_o _ ance(y) ] ] (DRS11) [ [ ] [ [x] bald(x) king_o _ ance(x) ] [ [y] king_o _ ance(y) ] ] (DRS10) ep esen s he global accommoda ion o he p esupposi ion and (DRS11) ep- esen s he local accommoda ion. Bu , be ween bo h op ions, we mus p e e he local ac- commoda ion (DRS11), because he global accommoda ion (DRS10) con ains a con adic- ion since we can esol e he ma ke y ins an ia ing i in he ma ke x ha appea s in an accessible ep esen a ion. Then, he exis en ial p esupposi ion is cancelled in (DRS11). The addi ion o an de Sand ’s accommoda ion heo y o DRT o deal wi h p esuppo- si ions gi es us ce ain ad an ages o e o he al e na i e heo ies. I includes, o ins ance, a esolu ion mechanism o anapho a we canno ind in he heo y o de ini e desc ip ions. Mo eo e , i o e s a c i e ion o dis inguishing be ween di e en scopes o he p esuppo- si ion and choosing he mos adequa e one in each case o cancel i . I also o e s he possi- bili y o ea ing in a di e en way he di e en p esupposi ions appea ing in discou se, depending on he ela ions wi h hei igge elemen s and hei si es inside he global DRS ha ep esen s hei meaning. Re e ences BEAVER, D. (1997). “P esupposi ion”, in BENTHEM, J. an & MEULEN, A. e (eds.) (1997), Handbook o Logic and Language, No h-Holland, Ams e dam, pp.: 939-1008. KAMP, H. & REYLE, U. (1993). F om Discou se o Logic, Kluwe , Do d ech . LEVINSON, S. (1983). P agma ics. Camb idge Uni e si y P ess, Camb idge. SALGUERO LAMILLAR, F. J. (1994). “Anapho ic ins an ia ion p oblems in an in e en ial model o u e ance ep esen a ion”, in MARTÍN VIDE, C. (ed.) (1994), Cu en issues in ma hema ical linguis ics, No h-Holland, Ams e dam, pp.: 39-48. SANDT, R. an de (1992). “P esupposi ion p ojec ion as anapho a esolu ion”. Jou nal o Seman ics, 9: 333-377. SANDT, R. an de (1993). “Resolu ion and accomoda ion”, in KAMP, H. (ed.) (1993), P esupposi ion, DYANA-2. Dynamic In e p e a ion o Na u al Language. ESPRIT Ba- sic esea ch p ojec 6852. Deli e able R2.2.A, Pa II, pp.: 233-242.