AlanLibertandChristoMoskovsky
JournalofUniversalLanguage3 March2002,97-132
OntheFormandFunctionofReflexivesin ArtificialLanguages
AlanLibertandChristoMoskovsky UniversityofNewcastle
Abstract
Thispaperexploresissuesinrelationtotheformandfunctionof reflexivepronounswhichalanguagedesignershouldtakeinto considerationintheconstructionofanartificiallanguagegrammar. Thefirstmainpartofthepaperdealswiththerathercomplexand versatilenatureofreflexivepronounsinnaturallanguages,discussing reflexivesintermsofsyntacticanddiscoursefunction,morphemic structure,lexicalspecificationforphi-features,subject-orientation,etc. Typologicalandfunctionalaspectsofreflexivepronounsinanumber ofexistingartificiallanguagesareconsideredinthenextpartofthe paper.Thefinalpartaddressesthequestionofwhatwouldbethe optimalreflexivepronounfromtheperspectiveofartificiallanguage construction.Itisproposedthatanycomponentofanartificial languagegrammar,includingthereflexivepronoun,wouldinevitably beacompromisebetweensimplicityofformandfunctionontheone hand,andclarityontheother.Itisalsosuggestedthat,asartificial languagesarebytheirverynaturenon-primarylanguages,designersof artificiallanguagesshouldtakeintoaccountessentialpropertiesof secondlanguageacquisitionanduse. 98 OntheFormandFunctionofReflexivesinArtificialLanguages
1.Introduction
Reflexivityisoneofthemoredifficultphenomenaofnatural languagestoanalyze.1 Evenafterconsiderablestudy,ithasprovenhard topreciselyformulatetheconditionsunderwhichreflexivepronouns (henceforthRPs)areused,eveninEnglish,themoststudiedlanguage. Possiblypartlybecauseofthelackofpreciseunderstandingofthisarea innaturallanguage,itappearsthatdesignersofartificiallanguages (henceforthALs)usuallydonotspecifyhowitworksintheir languages,andperhapsdonotgiveadequatethoughttothechoiceof formsandfunctionsofreflexivewords.Inthispaper,wepresenta surveyofreflexivepronounsinALs,andthenmakesomesuggestions forhowthisareaofthegrammarshouldbeconstructed.2 Toour knowledge,littleworkhasbeendoneinthisfield;Morneau(1994)has proposedguidelinesforanaphorainartificiallanguages,buthedoesnot dealspecificallywithreflexives.WebelievethatALsareoftenbest understoodinthecontextofacomparisontonaturallanguages. Therefore,beforediscussingthereflexivesofALs,wefirstdescribe
1 Weusethefollowingabbreviationsinglossesofexamples: ABS-absolutive INSTR-instrument(al) ACC-accusative M-masculine ADJ-adjective NOM-nominative ASP-aspect OBJ-object CAUS-causative PARTIC-participle CL-clitic PASS-passive COMP-complementizer POSS-possessive DETR-detransitivizer PRES-present ERG-ergative REFL-reflexive EXT-extentofaction SG-singular FOC-focus VOC-vocative INFIN-infinitive Also,"OT"indicatesthatatranslationisours,ratherthanbeingmadebythesource. Wehaveoftennotshowntheinternalstructureofwordswherethisisnotrelevant. 2 Thispaperwillnotdealwithlanguageswhichhavebeencreatedinconnectionwith someworkoffiction,suchasKlingon. AlanLibertandChristoMoskovsky 99 theirnaturallanguagecounterparts.Thispartofthepapermayalso proveusefultoALdesigners,astheycanseewhatpossibilitiesoccur innaturallanguages.
2.NaturalLanguages
Pronounsareastandardfeatureinnaturallanguagesandassuch havelongbeenanintegralpartoflinguisticdescription.Oneessential propertyofpronounsisthattheyhavenointrinsicreferenceoftheir ownand,inordertobeinterpreted,theyneedtobelinkedtoa linguisticordiscourseentityandderivetheirreferencebywayofthis link.Itis,therefore,hardlysurprisingthat,taxonomicandtypological issuesapart,therehasbeenaconsiderableamountofinterestinaspects ofinterpretabilityofpronouns.Theissuereceivedanewdimension afterLeesandKlima’s(1963)seminalpaper‘RulesforEnglish Pronominalization’whichrelatedtheinterpretationofpronounstotheir distribution.Thepaperdrewattentiontothefactthatthereisacertain complementarityinthedistributionofnon-reflexiveandreflexive pronouns:informally,reflexivescannotbetoofarawayfrom,while pronounscannotbetoocloseto,theirantecedents.Inthecaseof English,RPsnormallyrequireaclause-internalantecedent:
3 (1)a.Johni despiseshimself i.
b.*John i thinksthatMarydespiseshimselfi.
Ontheotherhand,(non-reflexive)pronounsasarule“lookfor”an antecedentoutsideoftheclausetheyarein:
(2)a.Johni thinksthatMarydespiseshim i.
3 Assignmentofidenticalindicesisastandardwaytoindicatethattwo(ormore) elementsareinacoreferentialrelation. 100 OntheFormandFunctionofReflexivesinArtificialLanguages
4 b.*John i despiseshim i.
SinceLeesandKlima’spaper,theamountofresearchdevotedto theinvestigationofanaphorainarangeoftypologicallydistinct languagescanonlybedescribedascolossaland,asaresult,knowledge inthisareahasmaderemarkableprogress.5 Studiesinvestigating anaphoracanbroadlybedividedintotwomaincategories:those assumingastructuralapproach-moreoftenthannotwithinagenerative framework,andthoseassumingapragmatics-basedapproach.Inthe former,thedistributionofreferentiallydependentelements,suchas reflexivesandpronouns,issubjecttotheoperationofBinding Conditions(Chomsky1981,1986,amongmanyothers)whichinvolve aspecificconfigurationalrelationcalledc(onstituent)-command6 (Reinhart1983)andalocalityconstraintbasedonanother configurationalrelationcalled government7 ;inthepragmatics-based approach,thedistributionofreflexivesandpronounsisseenasaresult ofthediscoursefunctiontheyperform,e.g.,todisambiguatebetween twoormorepotentialreadings(Dowty1980),andisanalysedinterms ofasetofGriceandiscourseprinciples(Levinson1991).The phenomenon,however,hasturnedouttobesocomplexthatneither
4 Thesentenceisungrammaticalonlyintheindicatedreading. 5 Anaphoraisabroadtermforreferentiallydeficientelements;thetermanaphorwill occasionallybeusedinthispapertorefertobothreflexiveandnon-reflexive pronouns. 6 Inanapproachassumingthatclauseshaveaninternalabstracthierarchical“tree” structure,c-commandisarelationbetweentwoconstituentsneitherofwhich dominatestheother;thenodeimmediatelydominatingthec-commandingconstituent alsodominatesthec-commandedconstituent.Thus,inastandardclause,thenoun phrasewhichc-commandsallothernounphrasesisthesubject;thesignificanceof thisisthat,withinsuchaclause,thereisnonounphrasethatc-commandsthe subject.Therefore,thereisnothingintheclausethatqualifiesasapotential antecedentforthesubject,anditcouldonlytakeaclause-externalantecedent. 7 Thisisarelationbetweenalexicalheadandanotherconstituentwithinthesame structuralcategory. AlanLibertandChristoMoskovsky 101 structuralnorpragmaticapproacheshavebeencompletelysuccessfulin accountingfordistributionaland/orinterpretiveaspectsofpronouns. Thereislittledoubtthatonebasicreasonforthislackofsuccessis theratherversatilenatureofRPsandthecontextstheycanbeusedin. Inwhatfollows,wediscusspropertiesofreflexiveformsinavariety oftypologicallydistinctlanguageswhich,intheviewtakenhere, shouldbetakenintoaccountintheconstructionofanartificial language.
2.1.TypologyofRPsinNaturalLanguages
Naturallanguagesemployavarietyofdevicestomarkreflexivity. WhatwearespecificallylookingatareRPsorpronoun-likewords (possibly,clitics),butitshouldbenotedthatreflexivitycanbemarked withamorphologicalaffixattachedtotheverbaswell.Therearealso languageswhichdonothavearequireddistinctreflexiveformofeither type.Forinstance,inGumbaynggir(Eades1979:312,citedinLevinson 1991:134),theobjectformsof3rdpersonpersonalpronounscanbe interpretedeitherascoreferentialornon-coreferentialwiththesubject:
(3)gula:-du bu:rwang gula:na maga-yu. 3SG-ERG painted 3SG-ABS redpaint-INST
“Hei paintedhimi/j withredpaint.”
Suchlanguagesseemtobeextremelyrare.Mostlanguagesdohave RPsandgenerallyformtheminthefollowingways(wegiveherethe classificationasinSchladt(1999:105-6)):
(A) fromthelexicalitem'body',orlexicalitemsforbodyparts,e.g., “head”(oraformthatwasdiachronicallyatsomestagealexical itemforbodypart): 8
8 Thisisbyfarthelargestgroupoflanguagesinthecorpus. 102 OntheFormandFunctionofReflexivesinArtificialLanguages
(4) Kabuverdiano Man?el fer?i se? cabec쨑a. Manuel hurt 3SG.POSS head “Manuelhurthimself.” (Estudoslingui?sticoscrioulos1967:22,citedinSchladt1999:105)
(B)fromelementswithmeaningssuchas“person”,“owner”:
(5) Paumari? Abono-ra na-noki-a-a-ha. self-OBJ CAUS-see-DETR-ASP-THEME “Heseeshimself.” (ChapmanandDerbyshire1991:178,citedinSchladt1999:105)
(C)fromalexicalitemwithameaningofreversalofdirection:
(6) Sanuma Atakusa a-no¨ kama nia sapa gun 3SG-INSTR 3SG shoot reverse.DIR
ko-pa-so-ma. return-EXT-FOC-COMP “Heshothimselfwithagun.” (Borgman1990:43,citedinSchladt1999:105)
(D)fromalexicalitemwhoseoriginalmeaninginvolved'reflection':
(7) Finnish Jussi na¨ki itse-nsa¨. Jussi(NOM) see.PAST reflection.on.water-3SG.POSS “Heseeshimself.” (Faltz1985:68,citedinSchladt1999:105) AlanLibertandChristoMoskovsky 103
(E)fromlocativeprepositions:
(8) Zande M?i-i?ím ?i tI-r? I-kill on-me “Ikillmyself.” (TuckerandBryan1966:150,citedinSchladt1999:106)
(F) fromitemswhichwereoriginallyusedasemphaticpronouns,and whichmaystillbeusedinthisway,e.g.,theEnglishhimself(see section1.3).9
Whatevertheirorigin,synchronicallyRPsareuniversallydivided intotwogeneraltypes:simplexformsconsistingofasinglemorpheme (monomorphemicRPs):
(9) Dutch Oscargedraagtzich. “Oscarbehaveshimself.”(Reuland2001:451) andcomplexforms,usuallyconsistingoftwomorphemes (polymorphemicRPs):10
(10) Dutch Oscarhaatzichself. “Oscarhateshimself.”(Reuland2001:451)
19 Thisisapparentlyarelativelysmallgroup,madeupexclusivelyofEuropean languages,accordingtoSchladt(1999). 10 AwidelyadoptedwaytorefertothesamedistinctionisintermsofSE(=simplex) andSELF(=complex)RPs(seeReinhartandReuland1991,1993). 104 OntheFormandFunctionofReflexivesinArtificialLanguages
Astheexamplesaboveindicate,Dutchhasbothasimplexanda complexRP,buttherearelanguages(e.g.,Russian,Czech)inwhich thereisonlyasimplexform:
(11) Russian
Jai chital [NP egoj stat"ju o sebe i/j ]. I read his article about RP “Ireadhisarticleaboutme/himself.” (Rappaport1986:105)
Therearealsolanguages(suchasEnglish)inwhichonlya complexRPexists.Inlanguageswhichhavebothasimplexanda complexRP,itisnotuncommonforthecomplexRPtobeaform derivedfromcombiningthesimplexformandacorrespondingpersonal pronoun:
(12) Chinese
Zhangsani renwei Lisij zhidao Wangwu k xihuan Zhangsan think Lisi know Wangwu like
ta ziji*i/*j/k. him RP “ZhangsanthinksLisiknowsWangwulikeshimself.” (ColeandSung1994:357)
ThedivisionbetweensimplexandcomplexRPs,whichcanbe tracedbacktoFaltz(1985,butseealsoPica1987,1991),isseenas veryimportantasithasbeenfoundtobeimplicationallyrelatedtoa numberofotherpropertiesofRPs,suchaspresence/absenceof phi-features(person,number,gender),beinglong-distance(e.g.,the abilitytorefertoaclause-externalantecedent),andsubject-orientation (e.g.,beingrestrictedtoreferencetoclausalsubjectsonly,andnotto objects).Thesearediscussedinsection2.3. AlanLibertandChristoMoskovsky 105
2.2.FunctionandMeaningofRPs
Traditionally,reflexivityisregardedasindicatingaverbalactivity whichisdirectedbacktotheentity(usuallytheclausalsubject)that initiatedit.Fromadifferentperspective,reflexivityinvolvesa predicate-argumentstructure(usuallyaclause,butsometimesalsoa nounphrase)inwhichtwooftheargumentshavethesamereferent:
(13) [ s Hedescribedhimselfasanhonestman]
(14) [ np Hisdescriptionofhimself]
Thus,theprincipalfunctionoftheRPistoimposeacoreferential readingonthetwoarguments(whileanon-RPinthesamecontext wouldnormallybeinterpretedasdisjointinreferencefromthesubject). ThiscanbeaccomplishedwithaRP(asinthesentenceabove),with aclitic(e.g.,intheRomancelanguages),orwithaboundreflexive morpheme(e.g.,inRussian). TheRPhasalsodevelopedanumberofrelatedfunctions.Itcanbe usedtoindicateanactivitywhich,whilenotexactlydirectedbackto theperformingentity,isseenasstrictlyassociatedwiththisentity (sometimesreferredtoas'middlevoice'):
(15) Dutch Oscargedraagtzich. “Oscarbehaveshimself.”(Reuland2001:451)
Insomelanguages(e.g.,English)itisoftenpossibletoomittheRP insuchcases:
(16)Oscarbehaved/washed/dressed(himself).
Inaddition,theRPcanalsobeusedtoindicatepassivemeaning 106 OntheFormandFunctionofReflexivesinArtificialLanguages
(asarule,withinanimatesubjects):therearelanguages(e.g.,someof theSlaviclanguages)inwhichareflexivecliticorareflexive morphemeisusedasanalternativedevicetoformthepassivevoice:
(17) Bulgarian a. Problem-at beshe reshe-n ot problem-the was resolved-PASSPARTIC from
bord-a na direktori-te. board-the of directors-the
b. Problem-at se resh-i ot
problem-the RP(CL) resolve-PAST from
bord-a na direktori-te. board-the of directors-the “Theproblemwasresolvedbytheboardofdirectors.”
Likewise,insomelanguagesRPscanbeusedinimpersonal constructions:
(18) Bulgarian Vliza se prez goljama gradina.
enterRP(CL) through big garden “You(can)enter(thehouse)throughalargegarden.”
ItisnotunusualforaRPtobeusedtoindicatereciprocalmeaning, evenifthereisadistinctreciprocalpronoun:
(19) Bulgarian Te se poglednaha.
they RP(CL) looked “Theylookedateachother.” AlanLibertandChristoMoskovsky 107
Fromadiscourseperspective,itseemsthattheprincipalfunctionof theRPistodisambiguatebetweenpotentialantecedents.Dowty (1980:32)hasproposedthatthedistributionofreflexiveand non-reflexivepronounsisregulatedbythefollowingprinciple:
(20)Neo-GriceanConversationalPrinciple Ifalanguagehastwo(equallysimple)typesofsyntactic structuresAandB,suchthatAisambiguousbetween meaningsXandYwhileBhasonlymeaningX,speakers ofthelanguageshouldreservestructureAfor communicatingmeaningY(sinceBwouldhavebeen availableforcommunicatingXunambiguouslyandwould havebeenchosenifXiswhatwasintended).11
Thisoperatestoeliminateambiguityinsentenceslike(21):
(21)a.*Shei sawheri.
b.She i sawherself i.(Dowty1980:32)
Inthetermsoftheconversationalprincipleformulatedin(20), whichwecaninformallycallavoidambiguity,theRP herself isthe unambiguoussyntacticstructureB: herself canreferonlytothesubject she and,therefore,isalwayschosenwhensuchistheintended reference.Ontheotherhand,thepronominal her canhaveanumber ofreferentsincluding she;heris,then,theambiguoussyntactic structureAwhich,inlinewith(20),isreservedforthecaseswhenthe intendedreferenceisoutsideofthesentence. Whiletheproposedconversationalprinciplecaptureswhatappears tobethebasicdiscoursefunctionofRPs,itisclearlyunabletoaccount forallaspectsofthedistributionofanaphors,becausethepatternwe seein(21),andwhichispresumablytheeffectof AVOIDAMBIGUITY,can
11 Foramuchmoreelaboratedproposal,seeLevinson(1991). 108 OntheFormandFunctionofReflexivesinArtificialLanguages alsobeobservedin(22):
(22)a.*Isawme. b.Isawmyself.
Here,thepronoun me isasunambiguousastheRP myself and, therefore,shouldnotbeaffectedbytheoperationofavoidambiguity. Theoppositeproblemismanifestedinsentenceslike(23)(notethat judgementsaboutsuchsentencesseemtovary):
(23) a. Max i sawagunnearhimself i/him i.
b. Max i likes[NP jokesabouthimselfi/him i]. (ReinhartandReuland1993:661)
Intheseexamples,thepronoun him isambiguousinthatitcanrefer notonlytothenominal Max,butalsotoaclause-externalentity.Thus, (23)representsexactlythecaseinwhichtheproposed AVOIDAMBIGUITY principleshouldapply,butitdoesn't. Infact,theneatpatternthatisobservedinsentencessuchas(21) and(22)isnotascommonasonemightthink,andthebreakdownin thecomplementaritybetweenreflexivesandpronounsseemstobequite wide-spread.Inthefirstplace,therearelanguages(e.g.,theRomance languages,butalsosomeGermaniclanguages)inwhichthereisno oppositionbetweenareflexiveandanon-reflexiveforminthe1stand 2ndpersonaltogether:thereflexiveisonlyusedinthe3rdperson, whileinthe1stand2ndpersonnon-reflexivepersonalpronounsare used:
(24) German
a. Ichi hab’ miri ein Auto gekauft. I haveme a car bought “Ihaveboughtmyselfacar.” AlanLibertandChristoMoskovsky 109
b. Eri hat sichi ein Auto gekauft. He has RP a car bought “Hehasboughthimselfacar.”
Itisinterestingtonotethatinlanguagesinwhichthereisadistinct reflexiveforminthe1stand2ndperson,itsuseisoften,butnot always,optionalandthenitcanfreelybeinterchangedwithapersonal pronoun:
(25) Bulgarian Az ne mislja za mene/sebesi. I not think about meRP “Iamnotthinkingofme/myself.”
Thesameisnottruefor3rdpersonsentences:
(26) Ivani ne misli za *nego i /sebesii. Ivan not thinks about himRP “Ivanisnotthinkingofhimself.”
Itshouldbenotedthatsentenceslike(23)and(25),inwhichthere isabreakdowninthecomplementaritybetweenreflexivesand pronouns,areequallyproblematicforstructuralaccountsofanaphora. Ithasbeensuggested(Farmer&Harnish1987;Reinhart&Reuland 1991,1993)thatthecomplementarityrelationholdsincasesinwhich theanaphorisaco-argumentoftheantecedent,asin(21)and(22), andbreaksdownincasesinwhichtheanaphorisanadjunct,asin (23a),orisembeddedinanargument,asin(23b).Asadescriptive generalization,thisseemstobefairly,butnotcompletely,accurateas itfailstocapturethefactthatcomplementarityisobservedforsome anaphorswhichareprepositionalarguments:
(27) Iammadatmyself/*me. 110 OntheFormandFunctionofReflexivesinArtificialLanguages
butnorforothers:
(28) Iam(not)thinkingofmyself/me.
Thepictureisadditionallycomplicatedbyinstancesoflogophoricity (seee.g.,Sells1987,Reinhart&Reuland1991)inwhichtheuseof theRPisarguablydeterminedby“thesubjectiveperspectiveofthe referentoftheantecedentofthereflexive”(Levinson1991:122). Considerthefollowingexamples(fromReinhart&Reuland 1991:311-2):
(29)a. ShegavebothBrendaandmyselfandadirtylook. b. Luciesaidthat(youagreedthat)apictureofherself wouldbeniceonthatwall. c. Lucieboastedthatthechairmaninvitedherhusbandand herselfforadrink. d. ShewrotethatMaxandherselfarehavingagreattime inLima.
Inthefoursentencesabove,theuseoftheRPseemshighly irregularasitisnotinthesameclauseasitsantecedentandrepresents acaseoflong-distanceanaphora(seethefollowingsection),whichis generallynotallowedinEnglish.Also,asobservedinReinhart& Reuland(1991:313),logophoricityonlyseemstoaffectRPswhichare eitheradjuncts,orareembeddedinanargument.Analogoussentences inwhichtheRPisactuallyanargumentareungrammatical:
(30)*Lucieboastedthatthechairmaninvitedherselfforadrink. (Reinhart&Reuland1991:312)
Itshouldbenotedthatlanguagesseemtovarywithregardtothe logophoricuseofRPs:empiricaldatasuggestthatsuchauseisrare ornon-existentinsomelanguages(e.g.,Bulgarian). AlanLibertandChristoMoskovsky 111
Finally,insome(butnotall)languages,theRPisalsousedasan emphaticpronoun:
(31)Thepresident himself willinspectthetroops.
Many(e.g.,König&Siemund1999)haveobservedthatthisfact isunlikelytobeacoincidence,especiallyasitisfoundinanumber oflanguages,notjustasinglelanguage.Therehavebeenproposals (e.g.,Faltz1985)thathistoricallythereflexiveformhasevolvedfrom combiningapronounandanemphaticelement.Considerthesituation inOldEnglish,wherenoRPsexistedandpersonalpronounswereused instead:
(32) þæt ic ænigra me weana ne wende. that I any me hope not expected “thatIexpectednohopeformyself.” (Beowulf932-93,citedinvanGelderen1999:191)
Theformsylfnewasoriginallyusedforemphasis:
(33)Sittan læteichinewiðmesylf-ne. remainletIhimwithmeself-ACC.M.SC “Ilethimremainwithmyself.” (Genesis438,JuniusManuscript,citedinvanGelderen 1999:192)
However,itgraduallybecameattachedtothepersonalpronounand cametomarkcoreference,not(only)emphasis.
2.3. Phi-features,Long-DistanceAnaphora,and Subject-Orientation 112 OntheFormandFunctionofReflexivesinArtificialLanguages
Thereappearstobeanimplicationaluniversalaccordingwhichif simplex(monomorphemic)RPsare un- orunderspecifiedfor phi-features(person,nember,gender),theyaredefinedaslong-distance (LD),becausetheycantakeaclause-externalantecedent,andtheyare subject-oriented,becausetheyinvariablyselectaclausaloranominal subjectastheirantecedent.ConsidertheIcelandicRP sig. Itis unspecifiedforperson,number,orgender,anditcanrefertoa'distant' antecedent,anditcannothaveanobjectasitsantecedent:
(34) Icelandic Jo?n i sagði Mar?iuj að þu? elskaðir sig i /* j. John told Maria that you loved RP “JohntoldMariathatyoulovedhim.” (ColeandSung1994:359)
Astheindexationintheexampleshows,theRP sig canonlyrefer tothe‘distant’matrixclausesubjectJon,butnottotheobjectMaria. Thesubject-orientationofsimplexRPsisevenmoreobviouswithina singleclause:
(35) Norwegian
Johni bad Maritj kikk-e bak segi/*j. John asked Mary look-INFIN behind RP
“Johni askedMarytolookbehindhimi .” (Hestvik1992:578)
Incontrast,complexRPsaregenerallyspecifiedforphi-features, theyarenotnormallyLD(e.g.,theycannottakeaclauseexternal antecedent)andtherearenorestrictionsontheclassofpotential antecedentstheycanselect.AtypicalexampleofacomplexRPisthe Englishhimself:itiscompletelyspecifiedforphi-features,and normallyitcanonlyrefertoanominalexpressionwithinthesame clause: AlanLibertandChristoMoskovsky 113
(36)a.Johni despiseshimselfi . b.*Johni thinksthatMarydespiseshimselfi .
anditisnotsubject-oriented(e.g.,itcanselecteitherthesubject ortheobjectasitsantecedent):
(37)Johni toldBillj abouthimselfi / j.
Itshouldbenotedthat,whileRPsinmostlanguagesseemtofollow thispattern,thereareneverthelessanumberofexceptionssuggesting thatitmightbemorecorrecttoseethisasastrongcrosslinguistic tendencyratherthanasatrueuniversal.Forinstance,theformsof complexRPsinanumberofGermaniclanguages(e.g.,theDutch zichself,theNorwegian segselv,theIcelandic sjalfursig)areas underspecifiedforphi-featuresastherespectivesimplexRPs(zich,seg, sig):bothtypesareonlymarkedfor(3rd)person.Ondataprovidedin Toman(1991),theCzechRPisafeaturelessmonomorphemic reflexive,butcontrarytoexpectationsitbehaveslikeacomplex,rather thansimplex,RP:itisneitherLD,norsubject-oriented:
(38)Kareli narovnal destickyj na sebei/ j. Karl stacked plates on RP “Karlstackedtheplatesonhimself/oneanother.” (Toman1991:155)
TheBulgarianRP sebesi alsobreaksawayfromthegeneral pattern:itclearlyisacomplexRPconsistingoftheboundreflexive morpheme sebe‘self’(which,similarlytotheEnglish self,isalsoa word-formativeprefix)andthereflexivepossessiveclitic si,butitis completelyvoidofphi-featuresandissubject-oriented. 114 OntheFormandFunctionofReflexivesinArtificialLanguages
(39)Ivani informira Petarj za sebesii/*j. Ivan informed Peter about RP “IvaninformedPeterabouthimself.”
2.4.SubjectRPsandRPsinsidetheSubject
StudiesinvestigatingRPsreportthatinsomelanguagestheyare barredfromoccurringinsubjectposition(whichinthefollowing examplesismarkedwithsquarebrackets):
(40)*Johnsaidthat[himself]wouldcome.
whileinothersthereisnosuchrestriction:
(41) Chinese
Zhangsani shuo [ziji i] hui kai. Zhangsan say RP will come “Zhangsansaidthathewouldcome.” (Progovac1993:759)
AsthisisanissuethatbearsonthedistributionofRPs,itwill brieflybeconsideredhere.(41)indicatesthatinprincipleRPscan occurinsubjectposition.Thequestionthatneedstobeaddressedthen iswhyRPsinEnglish(andanumberofotherlanguages)cannot.In theviewtakenhere,thereareatleasttworelevantreasonswhy himself isruledoutinsentenceslike(40):firstly, himself asanon-long-distant RPneedstotakeanantecedentinthesameclause,butifitisin subjectposition,therewouldbenothinghierarchicallyhigherthatcan actuallyfunctionasitsantecedent(seefootnote6);secondly,itisquite commonforRPstobemarkedforobjectivecase,andconsequentlyin languageswhichmarkthesubjectforthenominative,RPswouldbe barredfromoccurringinsubjectposition,becausetheyaremarkedfor adifferentcase.This,ofcourse,shouldnotbarthemfromoccurring AlanLibertandChristoMoskovsky 115 insidethesubject:
(42)a. [John’sdescriptionofhimself]surprisedus. b. [ThewayJohndescribedhimself]surprisedus.
Thiscompletesour(necessarilybrief)overviewofthedistribution anduseofRPsinnaturallanguages.Inthenextpartofourpaper,we shallreviewRPsinanumberofartificiallanguages.
3.ArtificialLanguages
OursurveyofRPsinALsissomewhatrandom,beingrestrictedto ALsaboutwhichwehaveafairamountofinformation,andindeed, notevenallofthosearediscussedhere.Adequateinformationmaynot beeasytocomeby,asprimaryworksareoftenoldandnotwidely distributed.Givensuchlimitations,wehopetogivesomeideaofwhat hasbeendonebylanguagedesignersinthisareaofthegrammar. Mostartificiallanguagesarebasedlargelyonnaturallanguages,and soitisnotsurprisingiftheirreflexivewordsresembleorareidentical toreflexivewordsinnaturallanguages.Asnotedabove,there apparentlyarenaturallanguageswithoutRPs,theirfunctionbeing fulfilledbymiddleorreflexiveverbformsorinsomeotherway.There seemtobefew,ifany,ALswhichexpressreflexivityonlythrough verbalinflection.Volapükhasaverbalaffixwhichmarksreflexivity (andwhichcanappearinmorethanonepositionrelativetopersonand numberaffixes).Inthesingularonehastheoptionofusingaseparate RPinsteadoftheverbalmarker.12 Inthe3rdperson,thisRPisidentical totheaffix,exceptthatitbearsacasemarker;inotherpersonsitis identicaltothepersonalpronouns:
12 Inthepluraltheuseofseparatepronounsisusedforareciprocalreading(only), whichiswhyreflexivitymustbesignaledbytheverbalaffix. 116 OntheFormandFunctionofReflexivesinArtificialLanguages
(44)a.Vatu¨k-ob. wash-1SG “Iwash.” b.Vatu¨k-ob-ok. wash-1SG-REFL “Iwashmyself.” c.Vatu¨k-ob ob-i. wash-1SG I-ACC “Iwashmyself.”(Sprague1888)
ThefollowingexamplefromtheALAmericanisintriguing:
(45) s?o agi maxtom facir s?o-m it has power make-INFIN it-ACC
mor-ermoso-m more-beautiful-ACC “ithasthepowertomakeitselfmorebeautiful.” (O'Connor1917:38)
Herea3rdpersonpersonalpronounisbeingusedasaRP.Itis notcleartouswhetherthiswasthegeneralpracticewhichO’Connor intendedtoprescribe:ifso,AmericanwouldbelikeGumbaynggirin not(obligatorily)distinguishingpersonalandreflexivepronouns.One question,perhapsespeciallyrelevanttoapriorilanguages,iswhether RPsareetymologicallyrelatedtoanyothertypeofword;thisis roughlyequivalenttothequestionofwhetherRPsaremonomorphemic orpolymorphemic.Toourknowledgetherearenoartificiallanguages withRPsofthetypeillustratedbyKabuverdiano,Sanuma,Finnish,or Zande. SomeaprioriALhavelexicalsystemsinwhichlexemeswith similarmeaningsorofsimilartypesshareoneormoresoundsin AlanLibertandChristoMoskovsky 117 common.Onecouldarguethatinsuchsystemsthereareno monomorphemicwords,andsoRPscouldbeseenaspolymorphemic, butnotinthewayconsideredbythosemakingthisdistinctionamong reflexives.Forexample,theOzRPis ek, ofwhichthe e- seemstobe anelementsharedbyotherpronouns,e.g., ep 'I,me'.Thepossessive formis kek. SomeALshaveRPswhicharemonomorphemicandhaveno phi-features,e.g.,Unitario's sin.However,theRPsofthemajorityof ALs,whethermonomorphemicorpolymorphemic,seemtohaveat leastonephi-feature.TheRPofVeltparl,evidentlyborrowedfrom English,ismonomorphemicandhasnophi-featureforperson,as shownbythefollowing:
(46)og filal self. I love RP “Ilovemyself.”(Couturat&Leau1903/1979:201)
Itdoesapparentlyhavethenumberphi-feature,sincetheplural formis selfy. SomeALshaveRPswhicharemonomorphemicandhave phi-featuresforperson.Therearetwowaysinwhichthiscanhappen: 1)Inmanylanguages,thereisonlyadistinctRPfor3rdperson;in theotherpersonsthepersonalpronounisalsousedasaRP.Therefore, ifthedistinctRPisused,itis3rdperson,bydefault,onemightsay. 2)Inotherlanguages,allRPsaredistinctfrompersonalpronounsin allpersons,andtherearedistinctRPsforallpersons. TheformercaseholdsinEsperanto,whose(3rdperson)RPis si, forbothsingularandplural,andallgenders;it,therefore,hasthe phi-featureforperson,butnotforgenderornumber.Thesameistrue ofInterlingua(IALA)andRomanova,bothofwhichhaveseastheir (3rdperson)RP. OtherALshavepolymorphemicRPs,allofwhichseemtoinvolve apersonalpronoun,oranassociatedpossessiveformandamorpheme 118 OntheFormandFunctionofReflexivesinArtificialLanguages meaning‘self’.TheseincludeseveralALsbasedonEnglish,orpartly onEnglish,suchasAnglo-Franca.Inthislanguage,therearenocase formsofpersonalpronouns,e.g.,thenominativeistheonlyform,and theRPsalsocontainthisform,e.g., theyselfs .Notethatpluralityis markedonbothpartsofthis,e.g.,thepersonalpronounandthe‘self’ part.13 TheRPsofOlingoconsistofpossessiveformsand -ego 'self', e.g., hiaego “himself.” ThegeneralRPofEurolengois sel,forallpersonsandnumbers. Itisthusmonomorphemicandlacksphi-features;asJones(1972:5) says,“Addtheword‘sel’ atallstages.”However,heusestheform lasel for‘herself’;thatis,forthefemininesingular3rdpersonRP,and apparentlyonlyforit,thereisapolymorphemicformwithphi-featurs:14
(47) Entretempo my sposa musto-preparar le Meanwhile my wife mustprepare the
lunch for lasel and le kinders lunch for herself and the children(ibid.:24)
LetusnowturntotheuseofRPsinartificiallanguages. Informationonthisishardertogatherthaninformationaboutforms,
13 Unitariohasthecuriousfeaturethatwhatwouldappearfromtheirformtobe reflexivepronounsareusedasnon-subjectpronouns(andnotasRPs,asfaraswe cantell):
(i) ilo se-mismo tenas. he/she/it him/her/it holds “Sheholdsit.”(Pleyer1990:27;OT)
Thereareothernon-subjectpronouns,suchas lo 'him/her/it',buttheuseofthe -mimso formsdoesnotseemtobeaquestionofemphasis,oratleastnotonlythat. Asnotedabove,theRPofthislanguageis sin. 14 Giventheexistenceof lasel,itmaybemoreaccuratetosaythatseldoeshavesome phi-features,namely[+MALE,-PLURAL],thantosayithasnofeatures. AlanLibertandChristoMoskovsky 119 becauseauthorsoftengivelessinformationaboutsyntaxthanabout morphology.Inaddition,whetherornotexplicitinformationisgiven, onemaywanttochecktextsinthelanguageforsyntacticdata,but oftenthereisnotenoughtextualmaterialinalanguagetofully determinesomepointsofsyntax.Onemightwanttobearinmindthat RPshave(orshouldhave)apragmaticfunction,thedisambiguationof possiblereadings,andasyntactic/semanticfunction,imposing co-referentialreadingsontwonounphrases. WeknowofnoinstancesofRPsinALsbeingusedinconstructions whichareclearlypassive,e.g.,whichhaveanexpressedagent,asin thenaturallanguageexample(17b).However,insomeALs,RPscan beusedinso-called‘middle’constructions.Thisisexplicitlystatedwith respecttoInterlingua(IALA):“Reflexiveconstructionsarealsousedto expresspassiveideaswhenthereisnoagentinvolved.‘Thesebooksare soldatBloomingdale’s’maybereplacedbythetranslationof‘These bookssellthemselvesatBloomingdale’s’.Notethatthiscovers constructionsofthetype,‘Thesebookssellwell.’”(GodeandBlair 1951:25).Oneoftheirexamplesis:
(48)Tal cappellos se vide frequentemente. such hats RP see frequently “Suchhatsareoftenseen”or“Oneoftenseessuchhats.” (ibid.)
Eurolengo'sRPisalsousedinthisway:
(49)Pipes fumo sel mor in le Unitado Kingdom... Pipes smoke RP more in the United Kingdom “PipesaresmokedmoreintheUnitedKingdom...” (Jones1972:46)
However,inthesamepassage,andwiththesameverb,wefinda truepassiveconstruction: 120 OntheFormandFunctionofReflexivesinArtificialLanguages
(50)so mucho as ever ist-o fum-ado. as many as ever be-PRES smoke-PASTPARTIC “asmanyaseveraresmoked”(ibid.)
Itisuncleartouswhyconstructionshavebeenchosenhereor whethertheyareinfreevariation.Toourknowledge,noALhasan impersonaluseofRPs,asBulgariandoes. Thereseemtobefew,ifany,ALsinwhichRPsarefreelyused asreciprocalpronouns. Likenaturallanguages,ALsvarywithrespecttoidentityofform ofRPsandintensivepronouns.Forexample,Esperantohasaseparate formforthelatter, mem.ThesameseemstobetrueofINTAL,in which self isfoundasanintensivepronoun(asopposedto se,the3rd personRP).Ontheotherhand,OlingousesRPsasintensives,e.g.,:
(51)Imiaego uri bonfortuna. Imyself am good.fortune “Imyselfamgoodfortune.”(Jaque1944:29)
InEurolengoandaUIaswellthereseemstobeidentityofform ofRPsandintensivepronouns.IntheBlueLanguagethesituationis notclear,atleastaspresentedinoursourceonthislanguage,Bollack (1900).Onp.20 su issaidtobeaRP,notdeclinedforperson,and onthesamepageaseriesof“emphatic”pronounsisgiven,whichdo havephi-features,e.g., eme‘myself’,and ete‘thyself’.However,on p.57isthefollowingsentenceinwhichtheemphaticpronounseems tobefunctioningasaRP:
(52)Et keni ete! thou(VOC) know ete “Knowthyself!” AlanLibertandChristoMoskovsky 121
Wecannotdeterminewhetherthisissimplyanerroronthepartof theauthor,orwhetherhemeanttoallowforsuchusageofthe “emphatic”pronouns. Pantos-Dimou-Glossaapparentlyhasanunusualtypeofidentityof form,betweenRPsandindefinitepronouns. Asfarasweknow,nolanguagedesignerexplicitlyallowsfor logophoricuseofRPs.Thismaynotbesurprising,consideringthat logophoricityhasonlycometotheattentionofeventheoretical linguistsinthelasttwodecades,andissomethingwhichnative speakersandlanguagedesignersmaynotbeconsciouslyawareof. However,italsowouldnotbesurprisingifinactualuse,RPswere sometimesusedaslogophoricpronouns. VeryfewsourcesonALsdiscusswhetherRPscanbelong-distance. ThedetailedgrammarofEsperantoWennergren(2001)isexpliciton thispoint:generallyEsperantodoesnotallowLDreflexives.Thisis shownbythefollowingpairofsentences:
(53)a.Liordon-isal laservisto. Heorder-PAST to theservant
vest-i si-n. dress-INFIN RP-ACC
“Hei orderedtheservant j todresshimself* i/j.”
b.Li ordon-is al la servisto He order-PAST to the servant
vest-i li-n. dress-INFIN he-ACC
“Hei orderedtheservant j todresshim i/*j .” (Wells1969:21)
Notethatthisiscontrarytotheimplicationaluniversalfornatural 122 OntheFormandFunctionofReflexivesinArtificialLanguages languagesmentionedabove.Giventhat si ismonomorphemicandisnot fullyspecifiedforphi-features,onewouldexpectittobelong-distance. Therearesomecircumstancesinwhichitcanbelong-distance,for exampleinaninfinitivephrasewithnoovertsubject:
(54) La reg?o send-is vok-i si-a-n the king send-PAST call-INFIN RP-ADJ-ACC
kuracisto-n. doctor-ACC “Thekingi sent(someone)tocallhisi doctor.”(PMEG;OT)
TheEsperantoRPissubject-oriented,andisexplicitlysaid(e.g.,in PMEG)tobeunabletoactasasubject.Unfortunately,detailedand clearinformationisusuallylackingindescriptionsofotherALs. ConsidertheremarkbyGodeandBlair(1951:25),whichis,toour knowledge,oneofthemostdetailedsourcesforthegrammarof Interlingua(IALA):“Reflexivepronounsareprimarilyofthetypein whichtheobjectoftheverbhappenstobelogicallyidenticalwiththe subject.”Thequestioniswhatismeantby“primarily”:doesitallow fornon-subject-orientation,orfortheRPtobesomethingotherthan averbalobject(e.g.,theobjectofapreposition),orboth? OzistheonlyALthatweknowofinwhichanRPcanactasa subject(andindeedasthesubjectofafiniteclause):
(55)ap ipOv ed ek pinfoid at. he said that RP loved her
“Hei saidthathei lovedher.”(Elam1932:26)
Elam(ibid.:16)describes ek as“referringonlytothesubjectofthe propositioninwhichitisused”.Thismeansthatitissubject-oriented, butgivenexamplesliketheoneabove,wemightnotinterpretElam’s wordstomeanitisnotlong-distance,e.g.,“proposition”maynotmean AlanLibertandChristoMoskovsky 123
“clause”. Ozisunusualforanotherreason:theRPcanrefertoanantecedent whichisnotinahierarchicallyhigherstructuralpositionanddoesnot c-commandit(ifthefollowingsentencedoesnotcontainanerror):
(56)hEv az ansAlUt iftlEplezais k-ek because the wicked not.immediately.receive RP-POSS
anpAtpyaup ek iftEgtOg adpad astlaup. just.deserts RP grow.bold on transgression “Becausethewickedi donotreceivetheiri justdeserts 15 immediately,theyi growboldintransgression.”(ibid.:36)
Here, az ansAlUt doesnotc-command ek (becauseitisina subordinateclause),andyetitfunctionsastheantecedentfortheRP. Theconditionsunderwhichthe(3rdperson)possessivereflexive andRPofHom-idyomo(sua and se respectively)aresupposedtobe usedareunusual.Cárdenas(1923:33)saysoftheformer:
Theuseofsomeofthethird-personpossessivesmaygiverise toambiguity.Inordertoavoidit,useismadeofthesubstitute possessive sua,whichgovernsthecomplementofasentenceor clausehavingtwosubjects,provided(1)thetwosubjectsareof thethirdperson;(2)theyhavethesamenumberandgender;(3) thecomplementrefersonlytothefirstofthetwosubjects.
Inotherwords,thereflexivepossessiveshouldonlybeusedto indicatecoreferencetothefirstoftwoconjoined3rdpersonsubjects whichhavethesamephi-features,asinthefollowingexample:
15 Elam(1932)doesnotgiveenoughinformationforustobeabletodividesomeof thewordshereintotheirconstituentmorphemes,butthisisnotofsignificancefor theissueathand. 124 OntheFormandFunctionofReflexivesinArtificialLanguages
(57)Mary kay Kate irey a sua domo. Mary and Kate went to sua house 16 “Mary i andKatewenttoheri house.”(ibid.)
Cárdenas(ibid.:34)saysthat“Thesamerulesapplytothereflexive se.”However,inthefollowingsentence,heuses se whentheabove conditionsdonothold,namelywhenthereisanon-conjoinedsubject:
(58)La marito adresey se a una tansport-agentio.... the husband betook RP to a transportation-agency
“Thehusband i betookhimselfi toatransportationagency....”
Asfarasweknow,nonaturallanguagehasdistinctsubject-and object-orientedRPs.Thelogic-basedALLojbanhassomethingclose tothem;however,ithasaseriesofRPswhichareusedtoindicate coreferencewithdifferentargumentsofthepredicate,e.g., vo'a for coreferencewiththefirstargument, vo'e withthesecondargument,and soon.Belowisanexampleofoneoftheseinuse:
(59)miklamale zarci vo'e. I go thestore RP
“Igotothestorei fromitselfi.” (LojbanReferenceGrammar chapter7.8)
TheRP vo'e isused,sincecoreferencewiththesecondargument of klama isintended.Wecould,therefore,callita2nd argument-orientedRP(whichisnotthesameasobject-orientation;the 2ndargumentinLojbandoesnotalwayscorrespondtothetraditional
16 Weareusingthesymbol
4.Recommendations
Weshallfirstmaketwogeneralrecommendations,e.g., recommendationsnotrestrictedtoRPs.Itsometimeshappensthat designersofartificiallanguagesdonotseemtofollowtheirownrules, e.g.,thereisoccasionallyamismatchbetweentherulethatthey proposeandthepracticethattheyfollowinthetextswhichthey provide.WehaveseenapossibleinstanceofthiswiththeBlue Language.Thiscanmakeitdifficulttodescribeoranalyzethe language,andperhapsmoreimportantly,couldcreateproblemsfor learners. Therefore,languagedesignersshouldtakecarenottomakeerrors inthedescriptionsoftheirlanguages,andinthetextstheywritein them.Also,designersshouldbeveryexplicitaboutthepropertiesof thelanguage.Forexample,withrespecttoRPs,theyshouldstate whethertheyaresubject-oriented,long-distance,andsoon. Aswithotheraspectsofconsciouslydesignedlanguages,thereisa potentialconflictbetweenclarityandsimplicity.OnecouldhaveanAL (orpartofone)whichisverysimple,butinwhichambiguityor unclaritywillfrequentlyarise,oralanguageinwhichsuchproblems veryrarelyoccur,becausethereisacomplexsystemforthatpartof thegrammar.Americanmayrepresentthegreatestsimplicityinthe domainofRPs,itdoesnotseemtohaveany,andpersonalpronouns canbeusedreflexively,eveninthe3rdperson.Thereis,however,a costofambiguityinthe3rdperson,e.g.,(45)apparentlycanalsomean
“iti hasthepowertomakeiti morebeautiful”.Thisraisesthequestion 126 OntheFormandFunctionofReflexivesinArtificialLanguages ofhowmuchambiguityisacceptableforanAL,andwhattypesof ambiguityareacceptable.Onemightfeelthatthekindofambiguity thatarisesinAmericanisunacceptable,althoughitmaybedifficultto provideaprincipledbasisforwhatisanunacceptabledegreeortype ofambiguity.However,thefactthattherearealmostnonatural languageswhichareambiguousinthiswaytoalargeextentmightbe takenassupportforthepositionthatanoptimalALshouldnotallow it,andhenceshouldhavesomeovertmeansfordisambiguation. Aspointedoutatthebeginningofthispaper,pronounsare referentiallydeficientandthereforeareintrinsicallyambiguous.Ifthe designer'sgoalistoavoidambiguityatallcosts,one(ratherextreme) waytodoitwouldbetoeliminatebothreflexiveandpersonal pronounsandsimplyrepeatnounphrases,e.g., JohnsawJohn.This wouldeliminatemuchambiguity,butveryfew,ifany,ALshavedone this.Onemightaskwhynot.Onepossiblereasonisthatitis inefficient,especiallyinthecaseoflongnounphrases.Further,itmay beaestheticallydispleasing,thoughtherolewhichsuchfactorsshould haveisnotcleartous. Whilewerecognizethatstrikingagoodbalancebetweenclarityand simplicityisessentialindesigningagrammaticalformwithaspecific function(asisthecasewithRPs),webelievethatitisalsovery importanttoconsidertheconstructionofanALfromadifferent perspective:viz.theperspectiveofthelanguagelearnerandlanguage user.Intermsoflanguageacquisition,ALsarebytheirnature non-primarylanguages;inotherwords,anALisalmostalwayslearned andusedasasecondlanguage.Thisbeingthecase,theviewtaken hereisthatthespecificpropertiesofsecondlanguageacquisitionmust betakenintoaccountintheconstructionofanAL.Thisisclearlya hugeissuewhosediscussioniswellbeyondthegoalsofthispaperand whichcertainlydeservesseparatetreatment. AsregardstheformoftheRPinALsfromtheperspectiveoftheir prospectiveusers,insecondlanguageacquisitionliteraturethereisnow substantialevidencethatintheacquisitionofthegrammaticalstructures AlanLibertandChristoMoskovsky 127 ofasecondlanguage,learnersdisplayadistinctpreferencefor form-functionrelationshipswhicharebothtransparentandisomorphic. Whatthismeansisthatsecondlanguagelearnersaremoresuccessful inacquiringgrammaticalstructureswhosefunctionissufficientlyclear, andtendtoestablishaone-to-onerelationshipbetweenagrammatical formanditsfunction.TheimplicationsfortheformofRPsinALsare obvious:theRPmusthaveadistinctform,itmustbeassimpleas possible,anditmustbeassociatedwithadistinctsinglefunction. Aswehavenoted,somenaturalandartificiallanguagesindicate reflexivitythroughverbalinflection(e.g.,themiddlevoiceformsof ClassicalGreek).Alanguagewithsuchverbformswouldnotneed reflexiveaccusativepronouns,butitmightbethoughtgoodtohave pronominalformsforreflexiveindirectobjectsand/orobjectsof prepositions.Wewouldarguethatthisisnotanefficientsystem--ifone isgoingtohaveRPs,itwouldbebettertohavethemforthewhole rangeofpossiblenon-subjectfunctions,ratherthanaddingverbforms whichwillcomplicatetheverbalparadigms. IfanALdesigneroptsforareflexiveformwhichisapronoun, fromtheperspectiveoftheALlearneranduseritwouldappearthat thatareflexiveformbasedon,orderivedfrom,thepersonalpronoun systemofthatALwouldbeeasiesttolearnanduse:insuchacase, thelearnerwillonlyneedtolearntheformofasimplereflexive morpheme(e.g., se or si)andthatitisattachedtotherespectiveform ofthepersonalpronountoformaRP(e.g., me-se or se-me). Therearemanylanguages,bothartificialandnatural,whichdonot havedistinctRPsforthefirstandsecondpersons.Thisdoesnotlead toambiguity,asitwouldinthe3rdperson;however,onecouldargue thathaving1stand2ndpersonRPsaddsalevelofclarity(desirable forthesecondlanguagelearner)andredundancy.Forexample,ifone doesnothearthesubjectofasentence,when myself istheobject,the subjectcanbereconstructedtobe I.However,thisgreaterredundancy comesatthecostofhavingaslightlymorecomplexsystemof pronounstobelearned. 128 OntheFormandFunctionofReflexivesinArtificialLanguages
AnotherquestioniswhetherRPsshouldhaveovertmarkingfor phi-features,andifso,forwhichones.Again,wefeelthathavingat leastpersonphi-featureswouldbedesirablefromthepointofviewof the2ndlanguagelearner.IfthemodelfortheformationofRPs suggestedaboveisadopted,thentheywillinheritthephi-featuresof thepronounfromwhichtheyareformed. AsregardsthefunctionofRPsinALs,wewouldrecommendthat theybeusedforasinglepurpose,viz.tomarkreflexivity.We, therefore,thinkthatanoptimalALwouldhaveseparateformsfor reciprocalandreflexivepronouns.Againthisinvolvesaquestionof whetheracertainkindofambiguityisacceptable,orwhetheritis worthaddingaformtothegrammartoeliminateit.Giventhe2nd languageorientationofALsand2ndlanguagelearners’preferencefor isomorphicform-functionrelations,wefeelitisbettertooptforclarity. Thesameremarksapplytointensivepronouns.Althoughsome languages,bothnaturalandartificial,employthesameformsforRPs andintensives,wefeelthatanoptimalALwouldhavedistinctforms. LikewisewedonotrecommendhavingthesameformsforRPsand indefinitepronouns,asPantos-Dimou-Glossadoes.Wealsosuggest avoidingunusualrulesforuseofRPs,likethoseofHom-idyomo. WenowturntosomemoresubtlequestionsofthefunctionofRPs. First,letusconsiderwhetherRPsshouldbesubject-oriented.Consider thefollowingsentence:
(60) JohntoldBillabouthimself.
IftheRPinanALequivalentofthissentencewere subject-oriented,thenthatsentencewouldnotbeambiguous,unlikethe Englishversion.Giventhatsubject-orientationofRPsisquite wide-spreadcrosslinguistically,andthatspeakersgenerallyhaveno problemsidentifyingthesubjectasasyntacticconstituent,itisunlikely thatstipulatingsubject-orientationforRPswouldsignificantlyaddto learnabilityorprocessabilityofALs.Thereisanother,possiblyvery AlanLibertandChristoMoskovsky 129 important,aspectofthesubject-orientationofRPs.Atthebeginningof thepaper,webrieflyreferredto c-command asahierarchicalstructural relationthatgenerallyholdsbetweenanaphorsandtheirantecedentsin naturallanguages.Whythereshouldbesuchacomplexconfigurational relationisaquestionthat,tothebestofourknowledge,noonehas answered(or,forthatmatter,evenattemptedtoanswer).Considering itsuniversality,however,andthatitmayberelevanttoother componentsofgrammar,c-commandislikelytobeanecessary conditionforlanguage-basedcommunicationsystems,andtherefore shouldnotbeexcludedfromthespecificationofALgrammars.Inview ofitscomplexnature,itwould,however,posehugelearnability problems.Theadvantageofhavingsubject-orientedRPs,apartform reducingpotentialambiguity,isthatbyvirtueofbeingthe hierarchicallyhighestconstituentwithintheclause,thesubject c-commandsallotherconstituents.Thus,subject-orientationentails c-commandwithouttheneedtoexplicitlylearnit. Letusconsiderwhetherlong-distanceRPsshouldbeallowedinan AL.Althoughtheyoccurinsomenaturallanguages,weconsiderthem tobeacomplicatingfactorinalanguagedesignedfor2ndlanguage learners.Thesimplestrulewouldbeonewhichrestrictedpossible antecedentsofanRPtothoseinthesameclause,whetheritisfinite ornot,ortothoseinthesameNP.Ahearer(orreader)wouldthen onlyneedtosearchinthatclause(orNP)forpossibleantecedents,and wouldnotneedtoconsiderpotentialantecedentswhichhadoccurred earlier,e.g.,inamatrixclausecontainingtheclauseinquestion.We wouldfavorsucharule,withnoexceptions(unlikeEsperanto);itcould leadtoambiguityinsomecases,butsuchambiguitywouldoftenbe obviatedbyhavingsubject-orientedRPswithphi-features,andwe believetheremainingpotentialforambiguityisoutweighedbythe simplicityresultingfromforcingRPstoselecttheirantecedentsinthe sameclauseorNP. IftherearenoLDreflexives,andifweassumethatanALwill havethesamesortsofconditionsontherelationbetweenanaphorsand 130 OntheFormandFunctionofReflexivesinArtificialLanguages theirantecedentsasnaturallanguages(e.g.,involvingc-command),then RPswillneverbeabletobesubjects. Takingalltheaboveintoaccount,inourviewtheoptimalRP wouldbeapolymorphemicone,whichborephi-features,andwhich clearlywasreflexive,whichwasdistinctinformfrombothreciprocals andintensives,andwhichwassubject-oriented,butnotlong-distance. Further,allofthesefactsshouldbeclearlyandexplicitlystatedinthe designer'spresentationofhislanguage.
References
Bollack,L.1900. AbridgedGrammaroftheBlueLanguage.("Englishversion byProfessorTischer").Ms.,Paris. Borgman,D.M.1990.Sanuma.InD.C.Derbyshire&G.K.Pullum(eds.), HandbookofAmazonianLanguages 2,15-248. Ca짫rdenas,C.1923. Hom-idyomo (secondedition).Ms.,Leipzig. Chapman,S.&D.C.Derbyshire.1991.Paumari짫.InD.C.Derbyshire&G. K.Pullum(eds.), HandbookofAmazonianLanguages 3,161-352. Chomsky,N.1981. LecturesonGovernmentandBinding.Dordrecht:Foris. Chomsky,N.1986. KnowledgeofLanguage:ItsNature,Origin,andUse. NewYork:Praeger. Chomsky,N.&H.Lasnik.1992.ThePrinciplesandParametersTheory.In J.Jacobs,A.vonStechow,W.Sternefeld,&T.Vennemann(eds.), Syntax: EinInternationalesHandbuchZeitgeno짲ssischerForschung 506-569. Cole,P.&L.Sung.1994.HeadMovementandLong-distanceReflexives. LinguisticInquiry 25,355-406. Couturat,L.&L.Leau.1903,1907/1979. HistoiredelaLangueUniverselle. Hildesheim:GeorgOlmsVerlag. Dowty,D.R.1980.CommentsonthePaperbyBach&Partee.InK.J. Kreiman&A.E.Oteda(eds.), PapersfromtheParasessiononPronouns andAnaphora 29-40. Eades,D.1979.Gumbaynggir.InR.M.W.Dixon&B.Blake(eds.), HandbookofAustralianLanguages 1,245-361. Elam,C.M.1932. TheCaseforanAPrioriLanguage. Cincinnati:TheOpen AlanLibertandChristoMoskovsky 131
SesamePress. EstudosLingui짫sticosCrioulos.1967.Lisbon:AcademiaInternacionalda CulturaPortuguesa. Faltz,L.M.1985. Reflexivization:AStudyinUniversalSyntax. NewYork: Garland. Farmer,A.&M.Harnish.1987.CommunicativeReferencewithPronouns.In M.Papi&J.Verschueren(eds.), ThePragmaticPerspective 547-565. Gode,A.&H.E.Blair.1951. Interlingua. NewYork:StormPublishers. Hestvik,A.1992.LFMovementofPronounsandAntisubjectOrientation. LinguisticInquiry 23,557-594. Jaque,R.S.1944. OneLanguage. SantaBarbara:J.F.RownyPress. Jones,L.1972. Eurolengo,theLanguageforEurope:APracticalManualfor BusinessandTourism. NewcastleuponTyne:OrielPress. König,E.&P.Siemund.1999.IntensifiersandReflexives:ATypological Perspective.InZ.Frajzyngier&T.Curl(eds.), Reflexives:Formsand Functions 41-74. Koster,J.&E.J.Reuland.(eds.).1991. Long-distanceAnaphora. Cambridge: CambridgeUniversityPress. Lees,R.&E.Klima.1963.RulesforEnglishPronominalization. Language 39,17-28. Levinson,S.C.1991.PragmaticReductionoftheBindingConditions Revisited. JournalofLinguistics 27,107-161. LojbanReferenceGrammar.AvailableatURL
Pica,P.1991.OntheInteractionbetweenAntecedent-governmentand Binding:TheCaseofLong-distanceReflexivization.InJ.Koster&E. Reuland(eds.),119-136. Pleyer,M.1990. Unitario.Bensheim:UnitarioPress. Pollard,C.&I.Sag.1992.AnaphorsinEnglishandtheScopeoftheBinding Theory. LinguisticInquiry 23,261-303. Progovac,L.1993Long-distanceReflexives:Movement-to-inflVersus RelativizedSubject. LinguisticInquiry 24,755-772. Rappaport,G.1986.OnAnaphorBindinginRussian. NaturalLanguageand LinguisticTheory 4,97-120. Reinders-Machowska,E.1991.BindinginPolish.InJ.Koster&E.Reuland (eds.), 137-150. Reinhart,T.1983. AnaphoraandSemanticInterpretation.London:Croom Helm. Reinhart,T.&E.J.Reuland.1991.AnaphorsandLogophors:AnArgument StructurePerspective.InJ.Koster&E.Reuland(eds.), 283-321. Reinhart,T.&E.J.Reuland.1993.Reflexivity. LinguisticInquiry 24, 657-720. ReulandE.J.2001.PrimitivesofBinding. LinguisticInquiry 32,439-492. Schladt,M.1999.TheTypologyandGrammaticalizationofReflexives.InZ. Frajzyngier&T.Curl(eds.), Reflexives:FormsandFunctions 103-125. Sells,P.1987.AspectsofLogophoricity. LinguisticInquiry 18,445-481. Sprague,C.E.1888.HandbookofVolapu¨k.TheOfficeCompany,NewYork. AvailableatURL
Zribi-Herz,A.1989.AnaphorBindingandNarrativePointofView:English ReflexivePronounsinSentenceandDiscourse. Language 65,695-727.