SELECTED CHESS COMPOSITIONS - GEORGE GR ATZER - 2 GEORGE GR ATZER
12 Pages
English

SELECTED CHESS COMPOSITIONS - GEORGE GR ATZER - 2 GEORGE GR ATZER

-

Downloading requires you to have access to the YouScribe library
Learn all about the services we offer

Description

  • expression écrite
SELECTED CHESS COMPOSITIONS GEORGE GRATZER 1
  • magyar sakkelet
  • chess masters at the chess club
  • chess compositions george gratzer
  • chess magazine
  • chess compositions
  • youngster on the right side of the picture
  • white wins

Subjects

Informations

Published by
Reads 27
Language English

AconstructionalapproachtoEnglishverbalgerunds
RobertMalouf
StanfordUniversity
Englishverbalgerundshavelongbeenofinteresttosyntacticians. Verbal
gerundphrasesdisplayamixofnominalandverbalpropertieswhichprovide
achallengetoanysyntacticframeworkthatassumesastrictversionofX-bar
theory. Variousapproacheshavebeenproposedtogetaroundtheseprob-
lems, buttheyallinvolveabandoningafundamentallydesirabletheoretic
assumptionoradoptingahighlyabstractstructureforwhichindependent
motivationisdi cultto nd,orboth.Anidealanalysisofverbalgerundsin
Englishwouldbeabletoaccountfortheirmixedverbalandnominalproper-
tieswithouttheadditionofotherwiseunmotivatedmechanisms.Inthispaper,
IwillproposeananalysisbasedonrecentworkinConstructionGrammarand
Head-drivenPhraseStructureGrammarthattreatsverbalgerundsasahy-
bridcategorythatinheritssomepropertiesofnounsandsomepropertiesof
verbs.
1 Propertiesofverbalgerunds
Thestrongestevidenceforthenominalnatureofverbalgerundscomesfrom
theexternaldistributionofverbalgerundphrases(VGerPs).VGerPsappear
incontextswhereotherwiseonlynounphrasescanoccur. Forone,clauses,
unlikeNPs,aregenerallyprohibitedfromoccurringsentenceinternally,as
shownin(1).
(1) a.*IbelievethatPattookaleaveofabsencebothersyou.
b.*WhydoesthatPattookaleaveofabsencebotheryou?
However,VGerPsaresubjecttonosuchconstraint:
(2) a. IbelievethatPat’s/Pattakingaleaveofabsencebothersyou.
b. WhydoesPat’s/Pattakingaleaveofabsencebotheryou?
Thisisapointaboutwhichtherehasbeensomedisagreementintheliterature.
Reuland(1983),forinstance,claimsthataccusativesubjectVGerPscannot
appearclauseinternally.However,considerthefollowingexamples:
(3) a.*DidthatPatgotarrestedbotheryou?
b.*DidforPattogetarrestedbotheryou?
c.*Didtogetarrestedbotheryou?
d.?DidPatgettingarrestedbotheryou?e. Didgettingarrestedbotheryou?
f. DidPat’sgettingarrestedbotheryou?
g. DidPat’sarrestbotheryou?
While(3d)maybesomewhatawkward,thereisacleardi erenceinaccept-
abilitybetween(3a{c)ontheonehandand(3d{g)ontheother.ThereforeI
thinkitisreasonabletoconcludethatwithrespecttotheprohibitionagainst
sentence-internalclausalarguments,VGerPsbehavelikeNPsandnotlikeSs.
Onethingworthobservinghereisthatverbalgerundphrasesdonothave
thefulldistributionofNPs. Inparticular,asweseein(4),verbalgerunds
cannotbepossessivespeci ers.
(4) a. Pat’sleaveofabsence’sbotheringyousurprisesme.
b.*Pat’s/Pattakingaleaveofabsence’sbotheringyousurprisesme.
But,asZwickyandPullum(1996)observe,onlyarestrictedsubclassofwhat
areotherwiseclearlyNPscanshowupaspossessives. So,(4)suggeststhat
verbalgerunds,likeoftheothercasestheydescribe,fallintoa\functionally
restricted"subclassofnounsthatcannotheadpossessivephrases.
Ontheothersideofthings,therearecontextswhichadmitverbalgerunds
butnotregularNPs. J rgensen(1981)andQuirketal. (1985:1230)discuss
aclassofpredicativeadjectiveswhichselectforanexpletivesubjectanda
verbalgerundcomplement,asin(5).
(5) There’snouse(you/your)tellinghimanything.
Thefactthatthecomplement’ssubjectcanappearinthepossessiveshows
thatthecomplementreallyisaverbalgerundphraseandthatthisisnota
caseofsubject-to-objectraising. Examplessuchasthisprovidesuggestive
evidencethatverbalgerundsformasubcategoryofnoun.
WhiletheexternalsyntaxofverbalgerundsismuchlikethatofNPs,their
internalstructureismorelikethatofVPs. Forone,VGerstakeaccusative
NPcomplements,whilethenominalgerundin(6b)canonlytakeaPPcom-
plement:
(6) a. (Pat’s/Pat)loudlycalling(*of)therollstartedeachday.
b. Theloudcalling*(of)therollstartedeachday.
Anotherverbpropertyofverbalgerundsisthatverbalgerundstakeadverbial
modi ers.Incontrast,commonnounstakeadjectivalmodi ers:
(7) a. Pat nanced(me/my)carefullyrestoringthepainting.
b. Thecareful/*carefullyrestorationofthepaintingtooksixmonths.Similarly,verbalgerunds,unlikenouns,canbenegatedwiththeparticlenot:
(8) a. Pat’snothavingbathedforaweekdisturbedtheotherdiners.
b.*Thenotprocessingoftheelectionresultscreatedascandal.
Thesefactshavebeenusedtomotivatetheclaimthatverbalgerunds
mustbeverbsatsomelevel. However, noneofthebehaviorexhibitedin
(6){(8)isuniquetoverbs. Someoftheverb-likepropertiesofgerunds,such
aslicensingadverbialmodi ers,arealsosharedbydeterminers,prepositions,
andadjectives:
(9) a. Sandyisawakenedearlyalmosteverymorning.
b. Sandylivesdirectlybeneathadancestudio.
c. Sandy’sapartmenthasaninsu cientlythickceiling.
Similarly,notcanbeusedinsomecircumstancestonegateadverbs,adjectives,
PPs,anddeterminers:
(10) a. Notsurprisingly,thedefendanttooktheFifth.
b. TheconferencewillbeheldinSaarbruc¨ken,notfarfromtheFrench
border.
c. NotmanypeoplewhohavegoneoverNiagaraFallslivetotellabout
it.
Thesefactsaboutmodi cationandnegationdonotshowthatverbalgerunds
areverbs.Whattheyshowisthatverbalgerunds,unlikecommonnouns,are
partofalargerclassofexpressionswhichincludesverbs.
Thecomplementationfactsalsodonotconstituteastrongargumentthat
verbalgerundsmustbeverbs. Likeverbsandverbalgerunds,prepositions
alsocantakeNPcomplements.Ontheotherhand,someverbsonlytakePP
complements:
(11) Thestrikeextended*twoweeks/throughthesummer.
Thefactthatsomeverbalgerundstakeaccusativeobjectsisthereforenotes-
peciallystriking.Whatisimportantisthataverbalgerund,unlikeanominal
gerund,takesthesamecomplementsastheverbfromwhichitisderived:
(12) a. Chriscasuallyputtheroastintheoven.
b. Chris’s/Chriscasuallyputtingtheroastintheovenappalledthe
visitingvegetarians.
c. Chris’scasualputtingoftheroastintheovenappalledthevisiting
vegetarians.So,whatwecansayisthataVGerPheadedbythe-ingformofaverbhas
thesameinternalsyntaxasaVPheadedbya niteformofthatsameverb.
Tosummarize, VGerPshavefourbasicpropertiesthatneedtobeac-
countedfor.Thesearegivenin(13).
(13) a. Averbalgerundtakesthesamecomplements astheverb from
whichitisderived.
b. Verbalgerundsaremodi edbyadverbsandnotbyadjectives.
c. Theentireverbalgerundphrasehastheexternaldistributionofan
NP.
d. Thesubjectofthegerundisoptionaland,ifpresent,canbeeither
agenitiveoranaccusativeNP.
Thepropertiesin(13)aresharedbyaccusativesubject(acc-ing),genitive
subject(poss-ing),andsubjectless(pro-ing)verbalgerundphrasesandare
notsharedbyanyotherEnglishconstructions. Thethreetypesofverbal
gerundsseemtobesubtypesofasinglecommonconstructiontype,andany
analysisofverbalgerundsoughttobeableaccountfortheirsimilaritiesina
systematicway.
Itisimportanttonote, however, thattherearedi erencesamongthe
threetypeswhichalsomustbeaccountedfor(Reuland1983,Abney1987).
Ofcourse,themostobviousdi erenceisthede nitionalone,namelythecase
ofthesubject. Inthatrespect,poss-ingsaremorelikeNPs,whileacc-ings
aremorelikeSs.Anotherdi erencecanbefoundintheiragreementbehavior
whenconjoined:
(14) a. ThatPatcameandthatChrisleftbothers/??botherme.
b. Patcoming(sooften)andChrisleaving(sooften)bothers/??bother
me.
c. Coming(sooften)andleaving(sooften)bothers/??botherme.
d. Pat’scomingandChris’sleaving??bothers/botherme.
e. PatandChris*bothers/botherme.
Conjoinedacc-ingorpro-ingVGerPs,likeconjoinedSs,prefersingular(or
default)numberagreementontheverb.Conjoinedposs-ingVGerPs,likecon-
joinednouns,preferpluralagreement. Furthermore,thetwotypesofverbal
gerundscannotbecomfortablyconjoined:
(15) a.*Pat’scomingandChrisleavingbothers/botherme.
b.*PatcomingandChris’sleavingbothers/botherme.Thepatternsofcompatibilityin(14)and(15)follownaturallyfromtheas-
sumptionthatacc-ingandposs-ingVGerPsareofdi erentsemantictypes.
Poss-ingVGerPs,likeNPs,havenominalsemantics,withanindexspeci ed
forperson,number,andgender. Incontrast,acc-ingVGerPs,likeSs,have
propositionalsemantics.
Anotherdi erencebetweenthetwotypesofVGerPspointedoutbyAbney(1987)
isthatposs-ingbutnotacc-ingVGerPswithwhsubjectscanfrontunder
‘piedpiping’:
(16) Thisisthereporterwhose/*who(m)winningthePulitzerPrizesurprised
Sandy.
Again,thesamecontrastcanbeseenbetweenNPsandSs:
(17) a. ThisisthereporterwhosesuccesssurprisedSandy.
b.*ThisisthereporterforwhomtowinthePulitzerPrizesurprised
Sandy.
Hereagainisaninstancewhereposs-ingVGerPspatternmorelikeNPswhile
acc-ingVGerPspatternlikeSs.However,itishardtoseehowthisdi erence
canbeattributedtoadi erenceinthesemanticsofthetwotypesofgerund
phrases. Instead,whatthisevidenceshowsisthatatsomepurelysyntactic
level poss-ingVGerPshavesomethingincommonwithNPswhileacc-ing
VGerPshavesomethingincommonwithSs.
2 HPSGpreliminaries
AnidealanalysisofverbalgerundsinEnglishwouldbeabletoaccountfor
theirmixedverbal/nominalpropertieswithouttheadditionofotherwiseun-
motivatedmechanisms.RecentworkinConstructionGrammar(FillmoreandKaytoappear)
andHead-drivenPhraseStructureGrammar(PollardandSag1994)provide
thefoundationforsuchananalysis.Sag(toappear)proposesanelaboration
oftheHPSGX-bartheorytoincludehierarchicallyclassi edphrasestructure
rules. Underthisview,theinternalstructureofaphraseisdeterminedby
boththelexicalpropertiesoftheheadandbytheconstructiontypeofwhich
thephraseisaninstance.Inthissection,Iwillpresentabriefoverviewofthe
relevantfeaturesofSag’s(toappear)hierarchyofphrasetypes.
InHPSG,wordsandphrasesaretakentobetypesofsigns,\structured
complexesofphonological,syntactic,semantic,discourse,andphrasestruc-
turalinformation"(PollardandSag1994:15).Signsarerepresentedbytyped
featurestructures,andthegrammarofalanguageisrepresentedasasetof
constraintsontypesofsigns. Thesesigntypesarefurtherorganizedintoa
multiple-inheritancehierarchytoallowlinguisticgeneralizationstobepre-
ciselystated.ConsiderableworkinHPSGhasfocusedonexaminingthehierarchical
structureofthelexicon. Morerecently,Sag(toappear)hasinvestigatedap-
plyingthesamemethodsofhierarchicalclassi cationtotypesofphrasalsigns.
Asmallpartofthephrasetypehierarchyisgivenin(18).
(18) phrase
CLAUSALITY HEADEDNESS
clause non-clause non-headed headed
head-subj head-spr head-comp
head-comp-control-cx
non n-head-subj-cx head-comp-cx
n-head-subj-cx noun-spr-cx
Phrasescanbedividedintotwotypes:endocentricheadedphrasesandexocen-
tricnon-headedphrases.Sincesyntacticconstraintsarestatedasconstraints
onparticulartypesofsigns,theHPSGHeadFeaturePrinciplecanberep-
resentedasaconstraintonallsignsofthetypeheaded. Headedphrasescan
befurthersubdividedaccordingtothekindofsubcategorizationdependency
theydischarge: subject,speci er,orcomplement. Inaddition,constructions
inheritconstraintsfromthecross-cuttingclassi cationofphrasesintoeither
clausesornon-clauses.Amongotherthings,clauseshaveaconstraintonthe
typeoftheirsemanticcontent. Aclause’scontentmustbeaparameterized
stateofa airs(psoa), somethingthatroughlycorrespondstoaproposition.
Thesetwohierarchicalclassi cationsde neasetofconstraintsonphrasal
signs.Aconstructionisaphrasalsigntypethatinheritsfromboththephrase
hierarchyandtheclause hierarchy. Sinceaconstructionlicensesatypeof
complexsign,itmustincludeinformationabouthowboththeformandthe
meaningareassembledfromtheformandthemeaningofitscomponentparts.
Aconstructionmayinheritsomeaspectsofitsmeaningfromitssupertypes,
andincontrasttothestrictlyhead-drivenviewofsemanticspresentedby
PollardandSag(1994),aconstructionmayalsohaveidiosyncraticmeaning
associatedwithit.
The n-head-subj-cx andthenon n-head-subj-cx constructionscombine
asubcategorized-forsubjectwitha niteandnon- nitehead,respectively.
The niteversion, fornormalEnglishsentenceslike Theywalk requiresa
nominativesubject. Thenon- niteversion,for‘minor’sentencetypeslike
absolutives,requiresanaccusativesubject. Thenoun-poss-cx construction
combinesanounheadwithapossessivespeci ertoformaphrasewitha nom-obj(i.e.,anindexbearingunit)asthecontentvalue. Tobemoreprecise,
theconstructiontypenoun-poss-cxissubjecttothefollowingconstraint:
2 3
(19) noun-poss-cx
6 72 3
6 7
6 7CATjHEADnoun
6 74 5SYNSEMjLOCAL6 7
6 CONTnom-obj 7
6 72 36 7
6 7noun6 74 54 5SPR-DTRjSYNSEMjLOCALjCATjHEAD
CASEgen
3 Anewanalysis
Aswesawinsection1,verbalgerundsdisplayamixofnominalandverbal
propertiesthatseemspuzzlinggivenmanyassumptionsaboutsyntacticstruc-
ture. Variousapproacheshavebeenproposedtogetaroundtheseproblems.
Abney(1987)arguesforahighlyabstractphrasestructureinvolvingphono-
logicallynullheadsandsyntacticwordformation. Pullum(1991)suggests
allowingaVtoprojectanNPundercertaincircumstances,buthisanalysis
cruciallydependsonthedefaultnatureoftheGPSGHeadFeatureConven-
tion,somethingwhichhasitselfraisedseriousformalproblems(Shieber1986,
Bouma1993).Lapointe(1993)proposesamoreconservativemodi cationto
standardnotionsofendocentricitybyintroducingduallexicalcategorieslike
hNjVi,aVwhichprojectsaVPdominatedbyanNP.Wescoat(1994),onthe
otherhand,proposestopreservephrasalendocentricitybyallowingasingle
wordtoprojecttwodi erentunorderedlexicalcategoriesandthereforetwo
di erentmaximalphrases.Whiletheseanalysisdi ergreatlyintheirtechni-
caldetails,theyallassignVGerPssomevariationofthefollowingstructure:
NP(20)
NP VP
Brown’s V NP
painting hisdaughter
ThisreflectsthetraditionaldescriptionofVGerPsas‘verbalinside,nominal
outside’quiteliterallybygivingVGerPsaVPnodedominatedbyanNPnode.
However,since(20)isquiteunlikethestructuresonetypically ndsinEnglish,
eachoftheseanalysesrequiresabandoningafundamentallydesirabletheoretic
assumptionoradoptingahighlyabstractstructureforwhichindependent
motivationisdi cultto nd.
ThefactorizationofsyntacticinformationintheHPSGlexiconallowsan
analysiswhichrequiresnosuchmove. WordsinHPSGselectforargumentsofaparticularcategory.Therefore,categorialinformationprojectedfromthe
lexicalheaddeterminestheexternaldistributionofaphrase. Selectionalin-
formation,fromalexicalhead’svalencefeatures,determineswhatkindsof
otherphrasescanoccurinconstructionwiththathead. Constructionalin-
formation,representedasconstraintsonparticularconstructions,controlsthe
combinationofsyntacticunits.Withineachofthesethreedomains,VGerPs
showfairlyconsistentbehavior. Whatisunusualaboutverbalgerundsis
theircombinationofnoun-likecategorialpropertieswithverb-likeselectional
properties.
WithinHPSG,thecategorialpropertiesofverbalgerundsaredetermined
bytheirlexicallyspeci ed headvalue.Likeallotherlinguisticobjects,types
of headvaluescanbearrangedintoamultipleinheritancetypehierarchy
expressinggeneralizationsacrosscategories. ThedistributionofVGerPscan
beaccountedforbythe(partial)hierarchyofheadvaluesin(21).
head(21)
noun verbal
p-noun c-noun gerund verb adjective
Sincegerund isasubtypeofnoun,aphraseprojectedbyagerundwillbe
abletooccuranywhereanNPisselectedfor.Thus,VGerPswillhavetheex-
ternaldistributionofNPs.Adverbsmodifyobjectsofcategoryverbal,which
includeverbs,adjectives,andverbalgerunds,amongotherthings. Sincead-
jectivesonlymodifyc(ommon)-nouns,VGerPswillcontainadverbialrather
thanadjectivalmodi ers. Since verbisadistinctsubclassofverbaldisjoint
fromgerund,VGerPswillnothavethedistributionoftrueVPs. Thiscross-
classi cationdirectlyreflectsthetraditionalviewofgerundsasintermedi-
atebetweennounsandverbs. Byformalizingthisintuitiveviewasacross-
classi cationof headvalues,wecanlocalizetheidiosyncraticbehaviorof
verbalgerundstothelexicon.
Thepositionofgerundinthehierarchyof headvaluesprovidesanim-
mediateaccountofthefactsin(13b)and(13c). Theremainingtwogerund
propertiesin(13)canbeaccountedformostsimplybythelexicalrulein(22).
2 3" #(22) 2 3verb
6 7HEAD HEAD gerund
6 7VFORMprp 6 2 376 7 6 76 2 37 SUBJ h 1i6 76 7 6 7=)6 7SUBJ h 1NPi6 7 6 76 7VALENCE COMPS 26 6 77 4 4 556 6 77VALENCE COMPS 24 4 55 SPR h 1i
SPR hip
tIN
eyn
tIN
eyn
p
2 3
PHON h i
6 72 32 36 7" #6 7
6 7verb6 76 76 76 HEAD 76 76 76 76 VFORMprp 76 76 76 76 76 76 76 76 76 76 7ARG-ST h 1, 2i6 76 76 76 2 37CAT 6 76 76 76 76 76 SUBJ h 1NP i 76 736 76 6 776 76 76 6 776 7SYNSEMjLOCAL6 7COMPSh 2NP i6 VALENCE6 776 746 76 4 574 56 76 76 7SPR hi6 76 76 76 76 2 3 76 76 76 paint-rel 76 76 76 76 76 76 76 7CONT ARTISTE 36 74 54 54 5
MODEL 4
Figure1:Lexicalentryforpainting(participle).
Thisruleproducesalexicalentryforaverbalgerundfromthepresentpartici-
pleformoftheverb.Theverbalgerunddi erssyntacticallyfromtheparticiple
intwoways:itisofcategorygerundanditsubcategorizesforbothaspeci er
andasubject.Sinceaverbalgerundselectsforthesamecomplementsasthe
verbitisderivedfrom,thephraseformedbyaverbalgerundanditscom-
plementswilllooklikeaVP.And,sinceagerundselectsforbothasubject
andaspeci er,itwillbeeligibletoheadeithera non n-head-subj-cx, which
combinesaheadwithanaccusativeNPsubject,oranoun-poss-cx,which
combinesaheadwithagenitiveNPspeci er. Sincethesubjectandspeci-
erareidenti edwitheachother,noverbalgerundwillbeabletocombine
withbothasubjectandaspeci er. GenitivesubjectVGerPswillinheritall
theconstraintsthatapplytopossessiveconstructionsingeneral,forexample,
restrictionsonthespeci erNPandonpiedpiping.Thedi erencesinagree-
mentfoundbetweenverbalgerundswithaccusativesubjectsandthosewith
genitivespeci ersfollowfromthedi erencesbetweenthetwoconstructions:
thenoun-poss-cxconstructionlicensesaphrasewithnominalsemanticswhile
thenon n-head-subj-cx constructionlicensesaphrasewithpropositionalse-
mantics.
Toseehowtheseconstraintsinteracttoaccountforthesyntaxofverbal
gerunds,itwillbeusefultoconsideranexampleofeachtype. First,con-
siderthe(partial)lexicalentryforthepresentparticipleoftheverbpaint,
inFigure1. Thisentrystatesthatthereisaword(pronounced/ /)
whichisthepresentparticipleformofaverb. Itselectsfortwoarguments,
asubjectandacomplement,which llthe artisteandmodelrolesofthe
verb’smeaning. MostoftheinformationinalexicalentrylikeFigure1is
inheritedfromhigherlexicaltypes.Ingeneral,onlythephonology,semantics,2 3
noun-poss-cx
6 72 36 7HEAD 16 7
6 72 36 76 76 7SUBJ h 2i6 76 7CAT6 6 776 7COMPShi6 VAL 4 574 54 5
SPR hi
2 3 2 3" #
head-comp-cxnoun 6 74 5 2 32 CATjHEAD 6 7CASEgen HEAD 16 7
6 72 36 76 76 7SUBJ h 2i6 76 7CAT6 6 776 76 VAL 4COMPShi 574 54 5
SPR h i2
h i2 32 3Brown’s 3 CATjHEADnounHEAD 1gerund
6 72 36 76 76 7SUBJ h 2i6 76 7CAT6 76 76 76 7VAL COMPSh 3i4 54 54 5
SPR h i2
painting hisdaughter
Figure2:Brown’spaintinghisdaughter
andperhapssubcategorizationframeneedtobestipulatedforeachlexicalen-
try.FromtheentryinFigure1,thelexicalrulein(22)producesamatching
entrythatdi ersonlyintheshadedvalues.Theoutputofthelexicalruleis
ofcategorygerund,ratherthanverb,andthegerundselectsforbothasubject
andaspeci er.Allotherinformationabouttheverbsgetscarriedoverfrom
theinputtothelexicalrule.
Nowweturntotheconstructionswhichaverbalgerundiseligibleto
head. Therearetwocases,poss-ingVGerPsandacc-ingVGerPs. Firstwe
willlookatthestructureoftheVGerPBrown’spaintinghisdaughter,shown
inFigure2. Theheadofthisphrase,painting,isaverbalgerundformedby
thelexicalrulein(22).ItcombineswithitscomplementNP(marked 3)via
thehead-comp-cxconstruction. Itthencombineswithagenitivespeci erto
formanoun-poss-cxcon.
Notethattheresultingverbalgerundphrasestillhasanunsatis ed subj
requirement,whichistoken-identicaltothespeci er Brown’s. Sincethetwo
head/subjectconstructionsrequireanominativeoraccusativesubject,the
verbalgerundphrasewillnotbeabletoheadahead/subjectconstructionand