Chomsky_s_Philosophy_of_Language乔姆斯基语言哲学

Chomsky’sPhilosophyoflanguageOutlineI.Descartesvs.LockeRationalismvs.EmpiricismII.Empiricist’stheoryoflanguageIII.Chomsky’scriticismsIV.Chomsky’srationalisttheoryoflanguageV.Chomsky’sresearchmethodsI.Descartesvs.LockeDescartesLanguagedistinguisheshumansfromanimalsThereareinnateideasClearanddistinctideasIthink,thereforeIamGodexistsmathematicsGeometry…Maybealsoknowledgeoflanguage??LockeTherearenoinnateideasAllknowledgecomesfromexperienceAllknowledgearebasedonsimpleideasandcomplexideasThemindhassomeabilitiestoPerceiveCompareRememberAbstract…Stimulus+Senseorgans+mind’sabilitiesknowledgeKnowledgeoflanguagealsoobtainedinthisway??II.Empiricist’sTheoryofLanguageKnowledgeoflanguageislearnedTherearesomegenerallearningmechanismsInduction/Generalization（归纳/推衍）Analogy（类比）Harris’s“discoveryprocedures”（乔姆斯基博士导师）Otherlearningalgorithms(人工智能领域的）Probabilisticalgorithms……III.Chomsky’sCriticismsCriticismofinduction/generalizationCriticismofanalogyChomsky:‘analogy’wrong(1)Johniseasytoplease(2)JohniseagertopleaseTheyaresimilar,sothechildshouldinterpretthemanalogously.Butthechilddoesnotdothis.Sotheideaofanalogyiswrong.Chomsky:‘generalisation’wrong(1)Johnateanapple.(2)Johnate.(3)JohnistoostubborntotalktoBill.(4)Johnistoostubborntotalkto.Generalisingfrom(1)and(2):(5)Wheneveranobjectismissing,anarbitraryobjectismeant.Butifthechildapplies(5)to(4),itwillgetawronginterpretationButthechilddoesnotmakesuchamistake.Anotherexamplethemanistall—isthemantall?thebookisonthetable—isthebookonthetable?Generalisation:Toformaquestion:preposethefirst‘is’Thechildwoulddo:themanwhoistallisintheroom—isthemanwhotallisintheroom?Butthechildactuallydoes:themanwhoistallisintheroom—isthemanwhoistallintheroom?HeartofthesecriticismsProponentsofanalogyorgeneralizationmightsaythatthechilddoesnotmakethoseanalogiesorgeneralizationsButtheycannotsayexactlywhenanalogy/generalizationshouldbeorshouldnotbemadeWithoutthis,theoryofanalogy/generalizationisemptyChomsky:otherlearningtheoriesarealsowrongSimilarcriticismsIV.Chomsky’sRationalistTheoryofLanguageThechilddoesnotmakesuchmistakesItshowsthatthechildhasgrammaticalknowledgethatisnotlearnedbyanalogyorgeneralisationThisknowledgeconsistsofinnategrammaticalrulesorprinciplesThisknowledgeisUniversalGrammar(UG)（普遍语法）UGprincipleexample:Subjacency（邻接条件）Subjacency(邻接条件):Movementcannotacrossmorethanoneboundingnode,whereboundingnodesareIPandNP.移动不能超过两个界限点SpecifiedSubjectConstraint*[CP[IPJohniappears[CP[IPitislikely[IPtitowin]________________________Cannowbeexplainedas:Johnmovedacrosstwoboundingnodes(IPnodes)WH-IslandConstraintIsometimeswonder[whysheispopular]*[CPPopulari,[IPIsometimeswonder[CPwhy[IPsheisti]]]].____________________________________Cannowbeexplainedas:popularmovedacrosstwoboundingnodes(IPnodes)ComplexNPConstraint[CPWhatidid[IPJohnclaim[CPthat[IPMaryownsti]]]]?_________________________________[CP*Whatidid[IPJohnmake[NPtheclaim[CPthat[IPMaryownsti]]]]?_________________________________________________TheabovefewrulesarethengeneralizedintoSubjacencyPrinciplesandParameters（原理和参数）toexplainlanguagevariationUGconsistsofalistofabstractprinciplesEachprincipleisassociatedwithaparameterE.g.head-firstorhead-secondroseredorredrose（roseisthehead中心词）Whenthechildisplacedinalanguageenvironment,thelinguisticinputwillsettheparameterstosomevaluesIndifferentlanguageenvironments,parameterswillbesettodifferentvaluesSo,achildindifferentlanguageenvironmentswillacquiredifferentgrammarsV.Chomsky’sResearchMethodsGalileo（伽利略）styleofinquiryDonotconsideralldataConcentrateonsomeinterestingdataMakesimplificationsE.g.assumethereisvacuumBuildaprecisemodeltoexplainthesedataE.g.Free-fallRecalcitrant（顽抗的）datamaybeexplainedlaterE.g.afeatherdoesnotabideLawofFreeFallButcanbeexplainedTheprecisemodeladvancesourunderstandingChomsky’sGalileostyleofinquiryDonotconsideralldataManysentencescontradictsSubjacencyButChomskydoesnotconsiderthemConcentrateonsomeinterestingdataChomskyonlyfocusesonsomesentenceswhichleadstoSubjacencyMakesimplificationsE.g.assumethatinalanguagecommunity,allspeakersspeakthesamelanguage,withnoindividualdifferencesBuildaprecisemodeltoexplainthesedataUGprinciples,e.g.SubjancencyRecalcitrantdatamaybeexplainedlaterE.g.agrammaticalsentencethatviolatesSubjancencymaybeexplainedbyothermeansTheprecisemodeladvancesourunderstandingUGadvancesourunderstandingoflanguageThemethodofthestandardsciences“Thestandardmethodofthesciencesisnottoaccumulatehugemassesofunanalyzeddataandtotrytodrawsomegeneralizationfromthem.Themodernsciences,atleastsinceGalileo,havebeenstrikinglydifferent.Whattheyhavesoughttodowastoconstructrefinedexperimentswhichask,whichtrytoanswerspecificquestionsthatarisewithinatheoreticalcontextasanapproachtounderstandingtheworld”(Chomsky)Chomskyclaimstobeusingthismethod