<<

Toappearin:PeterO.Müller,IngeborgOhnheiser,SusanOlsen,andFranzRainer(eds.). Formation. An International Handbook of the Languages of Europe .Berlin/NewYork:De GruyterMouton.

GeertBooij

12. Word-formation in construction grammar 1.Introduction 2.Thehierarchical 3.Holisticpropertiesofwordstructure 4.Phrasallexicalitems 5.Formmeaningasymmetries 6.Conclusions 7.References Abstract The notion ‘construction’ that plays a central role in Construction Grammar, is an indispensable notion for the analysis of word-formation patterns. In the study of word-formation, we investigate the systematic correspondences between form and meaning at the word level. Constructional schemas provide an adequate format for expressing these systematic correspondences. Moreover, they are part of a hierarchical lexicon in which both complex and morphological patterns of various levels of abstraction can be specified. An important advantage of the Construction approach is that it can express the relevant similarities between morphological and phrasal lexical expressions, and the paradigmatic relations between morphological and phrasal schemas. Thus, lexical knowledge is characterized as a complicated network between words and phrasal expressions on a range of levels of abstractions, varying between individual words and completely abstract patterns. 1. Introduction

Wordformationisthedomainofinwhichsystematiccorrespondencesbetweenthe formandmeaningofcomplexwordsarestudied.ConsiderthefollowingsetsofEnglish adjectives:

1 (1) steady unsteady social unsocial suitable unsuitable stressed unstressed sympathetic unsympathetic Themeaningoftheadjectivesinthecolumnontherightcanbecircumscribedas‘notA’,where Aisthemeaningofthecorrespondingadjectiveinthecolumnontheleft.Thatis,thereisa systematiccorrespondencebetweenthepresenceof un andthemeaningcomponent‘not’. Therefore,weconsidertheadjectivesintherightmostcolumnascomplexwords,withtheword structure[ un [x]A]A,where xisavariableforaphonologicalstring.Thus,weassignmorphological structuretowordsonthebasisofsystematicparadigmaticrelationswithotherwords.

Thestructure[un [x]A]Aispartofthefollowingconstructionalschemainwhichthe correspondencebetweenformandmeaningofamorphologicallydefinedclassofwordsis specified: (2)

<[ un [x]Ai]Aj ↔[NOTSEM i]j> Theangledbracketsdemarcateaconstructionalschema.Thecorrelationbetweenformand meaningisexpressedbythedoublearrowedsymbol↔.Themeaningcontributionofthebase wordontherightofthearrowiscoindexedwiththerelevantpartoftheformalstructureonthe leftofthearrow.Theschemathusrepresentsthemeaningofthese un adjectivesasa compositionalfunctionofthatoftheirbasewords.Theformatoftheseschemasderivesfromthe ParallelArchitectureframework,asdevelopedinworkbyRayJackendoff(Jackendoff2002; 2009;2010;2011).Themeaning(SEM)ofthebasewordsisspecifiedindependentlyinthe lexicon,whereasthemeaningcontributionofaffixesisspecifiedinconstructionalschemas,since theirmeaningisnotaccessibleoutsideofthemorphologicalstructureinwhichtheyoccur.Note, moreover,thatthemeaningof un dependsonthekindofmorphologicalstructureitoccursin.In thestructure[ un [x]N]Vforinstance,themeaningof un is‘reversativeaction’,asinthedenominal verbsun-cork and un-root .Hence,themeaningoftheprefixun cannotbespecifiedinisolation ofthemorphologicalstructureofwhichitformsapart.

2 Wordformationpatternscanthusbeconsideredconstructionsatthewordlevel,andthe individualcomplexwordsthatinstantiatethesepatternsare(morphological)constructs.The constructionalschemain(2)differsfromtheformatoftheWordFormationRuleasusedin traditionalgenerativemorphology(Aronoff1976)inthatitisneutralastoproductionor perception.ThisschemaisadeclarativestatementthatcharacterizesasetofexistingEnglish complexadjectives,andatthesametimeindicateshownewadjectivesofthistypecanbeformed. Thistypeofmorphologicalknowledgecanbeusedbothinlanguageperceptionandinlanguage production,andtherefore,itisquiteappropriateformorphologicalregularitiestobeexpressedin declarativeform. Thenotion‘construction’hasbeenshowntobeessentialforapropercharacterizationof theofnaturallanguages,andthistheoreticalstanceisreferredtoasConstructionGrammar (Goldberg2006).Theuseofthenotion‘construction’inthedomainofmorphologyhasbeen arguedforinmymonograph Construction Morphology (Booij2010).InthisarticleIwillpresent anumberofargumentsinfavourofaConstructionMorphologyapproachtowordformation.In section2theconceptofthehierarchicallexiconisintroduced.Insection3holisticpropertiesof complexwordsareshowntobeanargumentforconstructionalschemasatthewordlevel. Section4arguesthatthereisnosharpboundarybetweenlexiconandgrammar,abasicideaof ConstructionGrammar,andthatconstructionalschemascanbeusedinordertoexpressthe parallelismsbetweenmorphologicalandsyntacticconstructs.Insection5wewillseethatwe needtheconceptofparadigmaticrelationshipsbetweenconstructionalschemasinorderto accountforformmeaningasymmetriesincomplexwords.Section6presentsasummaryofthe argumentationforaConstructionMorphologyapproachtowordformation. 2. The hierarchical lexicon Theacquisitionofwordformationpatternsisbasedonknowledgeofmemorizedcomplexwords. Onceapatternhasbeendiscovered,thiscanbeexpressedinaconstructionalschemathat dominatestheindividualinstantiationsofthisschema.Theschemaanditsinstantiationscoexist, asthediscoveryofaschemawillnotleadtoremovalfromone’smemoryofthesetofcomplex wordsthatformedthebasisoftheconstructionalschema.Thisinsightcanbeexpressedina hierarchicallexicon,inwhichabstractschemasdominatetheirinstantiations.Forinstance,the followingsubstructurecanbeassumedfortheEnglishlexicon:

(3) <[ un [x]Ai ]Aj ↔[NOTSEM i]j>

3

<[un [steady ]A]A↔[notsteady]><[un [social ]A]A↔[notsocial]><[un [suitable ]A]A↔[notsuitable]> Theinformationaboutthesethreeadjectivesisalmostcompletelyredundant,astheirproperties areinheritedfromthegeneralschemabywhichtheseadjectivesaredominated,andthemeaning oftheirindependentlyspecifiedbasewords.Theonlynonredundantinformationaboutthese complexadjectivesisthattheyexist,thatis,thattheyareconventionalizedcomplexwordsof presentdayEnglish. ThepartialstructureoftheEnglishlexiconspecifiedin(3)hasonlyonelevelof abstractionabovetheleveloftheindividualcomplexwords.However,itmightbenecessaryto havemorelevelsofabstraction.Aclearcaseforthisnecessitycanbefoundinthedomainof compoundinginGermaniclanguages.ThemostgeneralstatementaboutGermaniccompoundsis thattheyarerightheaded.Thiscanbeexpressedatthetopmostlevelofthesubstructureofthe lexiconforcompounds.However,thevariousclassesofcompoundsmayalsohavespecific propertiesoftheirown.Examplesarethatverbalcompoundingisunproductiveinlanguageslike Dutch,English,andGerman,andthatinDutchANcompounds,unlikeNNcompounds,thenon headtendstobesimplex.Forinstance,whereastheANcompound hoog-bouw ‘highbuilding, highrisebuilding’isacorrectcompoundofDutch,acompoundlike huizenhoog-bouw ‘houses highbuilding,veryhighrisebuilding’,withthecomplexadjective huizen-hoog ‘lit.houseshigh, veryhigh’asmodifier,isungrammatical.Hence,thefollowingpartialhierarchyisrelevantfor DutchANcompounds:

(4) <[X iY j]Yk↔[SEM jwithrelationRtoSEM i]k>

<[A iN j]Nk↔[SEM jhavingpropertySEM i]k> Condition:A i=simplex Thelowerschemaisaninstantiationoftheupperschema,withthecategoryvariablesspecifiedas AandN,andthenatureofthesemanticrelationRspecifiedas‘havingproperty’;moreover,a restrictiononthecomplexityofthenonheadconstituentapplies. Anotherargumentfortheuseofintermediatelevelsofabstractionforthedescriptionof regularitiesincompoundsisthatconstituentswithincompoundsmayhavelexicalized,yet productivemeanings.Thatis,theymayhaveboundmeaningsdependentontheiroccurrencein

4 compounds.AnexamplefromDutchistheuseofthenoun dood ‘death’asamodifierinNA compounds,withthemeaning‘very,toahighdegree’asin: (5) doodgewoon‘verynormal’ doodmoe‘verytired’ doodsimpel‘verysimple’ doodstil‘veryquiet’ Thisproductiveuseoftheboundmeaning‘very’ofthenoun dood canbeexpressedina subschemaforNAcompoundsthatisdominatedbytheNAcompoundschemathatinitsturnis dominatedbythegeneralschemaforadjectivalcompounds:

(6) <[[ dood ]Ni [ x]Aj ]Ak ↔[very iSEM j]k> Inthisschemaoneoftheslotsislexicallyspecified,andhencethisisaconstructional (Jackendoff2002).Thesemanticspecificationfor dood in(6)overrulestheliteralmeaning ‘death’of dood whenusedasanindependentword. Wordswithboundmeaningsareoftenreferredtoasaffixoids,astheyaresimilarto affixesinhavingboundmeanings.Bymakinguseoftheconceptof‘constructionalidiom’wecan avoidintroducinganewmorphologicalcategoryforwordconstituentsbesideswordsandaffixes. Aconstructionalidiomisaconstructionalschemainwhichatleastoneslotislexicallyfixed,and atleastoneslotisopen. Weneedthistypeofconstructionalschemaforanotherformofboundednessaswell. ConsidertheGermanword Macher ‘maker’,discussedindetailinJoeres(1995).Joeres’ observationisthatmacher withtheregularmeaning‘maker’isaveryproductiveconstituentof compounds,whereasithasalexicalizedmeaningwhenusedasanautonomouslexeme,namely ‘strongpersonalitywhoachievesalot’.Hence,Joeres(1995:151)concludedthatmacher canbe qualifiedasa‘Halbsuffix’,thatis,anaffixoid.Note,however,thatthat macher isnotone ,aswasthecasefortheaffixoidsdiscussedabove,butconsistsoftwo,theverbalstem mach-‘make’andtheagentivesuffix -er .ExamplesofthistypeofcompoundinginGermanare: (7) with A as first constituent : Fitmacher‘fitmaker’ Krankmacher‘illmaker’

5 Wachmacher‘awakemaker’ with N as first constituent : Babymacher‘babymaker’ Eismacher‘icemaker’ Programmmacher‘programmaker’ Differentfromwhatisatstakewiththenoun dood ,themeaningof macher inthesewordsis completelyregular:ithasthemeaning‘entitythatcausesorcreatessomething’.Thismeaning, however,isonlyproductivewithincompounds,andnotavailablefortheword Macher in isolation,whichonlyhasthelexicalizedmeaningmentionedabove.Inordertoaccountforits boundedproductivity,wemayassumethefollowingconstructionalforthisclassof compounds:

(8) <[[A i [ mach ]Vj]V er ]Nk ↔[who[CAUSETOBE] jSEM i]k

<[[N i [ mach ]Vj]V er ]Nk ↔[who[CREATE]jSEM i]k Thesestructurespresupposethattheer nounsarederivedfrompotentialbutnonexistingverbal compoundsofthetypeAVandNVrespectively.Forinstance, Fit-macher canbeseenasa derivationfrom fit machen ‘tomakefit’.Thus,itispredictedthatinthesecomplexwords, macher hastheexpected,fullyregularmeaningofcauserorcreator.Theschemasin(8)willbe dominatedbythegeneralschemafordeverbalnounsiner inGerman. Theschemaformatproposedheremakesitpossibletoaccountforthecooccurrenceof wordformationpatterns.In(8)wesawacaseoftheproductivecooccurrenceofverbal compoundinganddeverbaler derivationinGerman.Normally,verbalcompoundinginGerman isunproductive.Thatis,thisisacaseof‘embeddedproductivity’(Booij2010:Ch.3).Language usersmayusemorethanonewordformationprocessatthesametime,inordertoconstructa multiplycomplexword.Thereare,forexample,quiteanumberofEnglish un adjectiveslistedin VanDale’s English-Dutch dictionary forwhichthebaseadjectiveinable isnotlistedaswell, forinstance: (9) listed not listed unbeatable beatable uncommentable commentable uncrushable crushable

6 unsayable sayable Wecanaccountfortheformationof un-V-able adjectivesfromverbalbaseswithoutthe intermediateword,adeverbaladjectiveinable ,beinginexistence,byassumingthatword formationschemascanbeunifiedandhave‘alifeoftheirown’.Thefollowingunifiedschema accountsfortheadjectivesintheleftcolumnof(9):

(10) [un [Vable ]A]A Toconclude,theschemaformatisanappropriateformatforcharacterizingword formationprocessesandsetsofcomplexwordsaspartsofahierarchicallexicon. 3. Holistic properties of word structure Oneofthemotivationsforconstructiongrammaristhatconstructionsmayhaveholistic propertiesthatcannotbereducedtopropertiesoftheirindividualconstituents.Thisappliesto morphologicalconstructionsaswell.Aclearexampleisthatreduplicationstructuresoften indicatesomeformofincreaseofasemanticpropertydenotedbythebaseword.Forinstance,in Malaythepluralformsofnounsiscreatedbyaddingafullcopyofthebasenoun(e.g. buku ‘book’–buku-buku ‘books’).Themeaningofpluralitydoesnotderivefromoneofthe constituents buku ,butisevokedbythecopyingconfigurationassuch. AnexamplefromDutchisthesemanticinterpretationofnumeralsdiscussedindetailin Booij(2010:Chapter8).Thenumeral drie ‘3’and honderd ‘100’canbeconcatenatedinboth orders,butwithdifferentinterpretations: drie-honderd ‘300’, honderd-drie ‘103’.Hence,the natureofthesemanticrelationbetweenthetwonumeralconstituents(multiplicationoraddition) isdeterminedbytheorderinwhichthelowerandthehighernumeralappear.Thatis,the semanticinterpretationispartiallyaholisticpropertyofthecomplexnumeralasawhole. Aprototypicalexampleofholisticpropertiesinwordformationistheinterpretationof nominalVNcompoundsinRomancelanguages,whichdenoteagents,instruments,orboth,and sometimesothersemanticcategoriessuchasevents.HerearesomeexamplesfromItalian: (11) apriconcerto‘openconcert,openingact’(event) apriscatole‘opencans,canopener’(instrument) lanciafiamme‘throwflames,flamethrower’(instrument)

7 lavapiatti‘washdishes,dishwasher’(agentorinstrument) spazzaneve‘ploughsnow,snowplough’(instrument) portabagagli‘carryluggage,luggagecarrier/roofrack’(agentorinstrument) ThesemanticcomponentoftheseVNcompoundsthatisnotexpressedbyeitheroftheir constituentsisthemeaning‘agent/instrument/event’.Itisthismeaningthatisevokedbythe morphologicalconfigurationasawhole.NotethatthelexicalcategoryNofthesecompoundsis alsoaholisticpropertyasthesecompoundsareexocentric,andtheNconstituentisnotthehead ofthecompound,neitherformallynorsemantically.Theseholisticpropertiescanbecapturedby aconstructionalschemaofthefollowingtype:

(12) <[V iNj]Nk↔[Agent/Instrument/EventofACTION ionSEM j]k> Thisschemaspecifiesameaningcomponentforwhichnoexplicitconstituentisavailable. Inconclusion,holisticpropertiesofwordformationconstructionssupportthe ConstructionMorphologyapproachtowordformation. 4. Phrasal and phrase-based lexical items AnimportantargumentforConstructionMorphologyisthatitisabletoexpresstheparallelisms betweencomplexwordsandphrasalexpressionswithsimilarfunctions.Aclassofphrasallexical itemsthatisveryrelevantforaproperanalysisofthesyntaxmorphologyinteractionisformedby particleverbsinGermaniclanguages(Los,Blom,Booij,ElenbaasandvanKemenade2012). Theseparticleverbsfunctionasnamesforverbalconcepts(events,actions).Theyarephrasalin naturesincetheparticlecanbeseparatedfromtheverb.InGermanandDutch,forinstance,root clauseshaveawordorderinwhichthefiniteverbappearsinsecondposition,whereasthe particleappearsattheendoftheclause.Particleverbsarethereforephrasallexicalunits,but similarinfunctiontoprefixedverbs.AnexampleistheclassofDutchparticleverbswiththe particle door .Theseverbsexpresscontinuativeaspect: (13) verb phrasal verb eten‘toeat’ dooreten‘tocontinueeating’ fietsen‘tocycle’ doorfietsen‘tocontinuecycling’ zeuren‘tonag’ doorzeuren‘tocontinuenagging’

8 Thecrucialobservationisthattheword door hasaproductive,yetboundmeaningwhen combinedwithaverbintoaphrasalverb,namelythatofcontinuativeaspect.Thefollowing constructionalschemaexpressesthissemanticproperty:

(14) <[door iV j]k ↔[continue iSEM j]k> Thisschemaspecifiesredundantpropertiesofexistingparticleverbswith door ,andhownew particleverbsofthistypecanbecoined.Theschemain(14)isagoodexampleofa constructionalidiom(Jackendoff2002),aschemawithbothvariableandlexicallyfilledpositions. InprobablyallEuropeanlanguageswefindphraseswithcoordinationthatare semanticallytransparent,buthaveafixedorder,suchasEnglish salt and pepper, father and son, ladies and gentlemen .Theseconventionalizedexpressionshavetobelistedbecauseofthefixed orderinwhichthetwowordsappear.Ontheotherhand,theyinstantiateregularandproductive phrasalpatterns,andweshouldexpressthisinourdescription.Therefore,thesyntactic coordinationschema[NandN] NP ,whichisasubcaseofEnglishcoordination,dominatesits instantiations,phrasessuchas salt and pepper .Theonlynonredundantinformationconcerning these N and N phrasesisthattheyexist(thatis,areconventionalized),andtheorderinwhichthe twowordsappear.Thus,thereisnoprincipleddifferencebetweencomplexwordsandphrases withrespecttothedivisionoflabourbetweenstorageandcomputation.Thereismassive empiricalevidenceforprefabs,wordcombinationsstoredinmemory(ErmanandWarren2000; Kuiper,Egmond,KempenandSprenger2007;Sprenger2003;Sprenger,LeveltandKempen 2006;TremblayandBaayen2010;Wray2002;2008). Romancelanguagesusevarioustypesofphrasesasnamesforconcepts.Forexample,in Frenchthepatterns N à N and N de N (asin moulin à vent ‘windmill’and salle de bains ‘bathroom’respectively)areusedforconstructingnamesforentities.ThismeansthatFrenchNPs withtheform[N[à/deN] PP ]NP havealexicalstatus,andfunctionasconstructionalidioms(Booij 2009). Acorrectpredictionofthepositionthatbothwordsandphrasesarestoredaslexicalunits isthatwewillfindcompetitionandblockingeffectsbetweensynonymousphrasesandwords. Forinstance,theexistenceoftheDutchANcompound zuur-kool ‘sourcabbage,Sauerkraut’ impedesusingtheANphrase zure kool ‘sourcabbage’forthistypeofcabbage , andinversely,the existenceoftheANphrase rode kool ‘redcabbage’impedesthecoinageoftheANcompound rood-kool .Relatedlanguagesmaydifferinthechoicebetweensynonymousbutstructurally

9 differentoptions.Forinstance,whereasGermanischaracterizedbyaveryfrequentuseofAN compounds,DutchspeakerstendtopreferANphrases(asinGerman Rotwein ‘redwine’versus Dutch rode wijn ‘redwine’(Booij2002;Hüning2010)).Interestingly,ithasbeenshownthatthe choicebetweenANcompoundandANphraseinGermanforanewconceptdependsonthe familyofrelatedexpressions.Forinstance,sincetherearemanyANcompoundswith milch ‘milk’astheirhead,anewANexpressionforanewkindofmilkwillmostprobablybea compound(SchlückerandPlag2011).ThisshowsthatnewANexpressionsaremadeonanalogy toconstituentfamilies,thesetsofexistingANcompoundsorphrasesthatshareaconstituent. InPolish,nounsfollowedbyadenominalrelationaladjectiveareusedextensively,where EnglishwouldhaveNNcompounds(Szymanek2010):

(15) a. kość[[słoni] Now] Aa boneelephantADJ FEM .SG ‘elephantbone’

b. sok[[jabłk] Now] Ay juiceappleADJ MASC .SG ‘applejuice’ LexicallystoredcollocationalrestrictionsbetweenAandNinANphrasesmayalso affectinflection.InDutch,theprenominaladjectivethatisnormallyinflectedandendsinthe suffixe /ə/,maybeusedwithoutendingwhentheANphraseisalexicalizedname,asin het stoffelijk overschot ‘themortalremains’,wheretheexpectedschwaendingisomitted(Speelman, TummersandGeeraerts2009). Theconsequenceoftheseobservationsisthatthereisnosharpboundarybetween grammarandlexicon.Bothatthelevelofthewordandthelevelofthephrase,thegrammar containsschemasthatdominatetheirlexicalizedinstantiations.Thus,thelexiconbecomesa ‘construction’. Theabsenceofasharpboundarybetweenlexiconandsyntaxisalsomanifestedbythe factthatphrasesmayfeedwordformation.Itisinparticularcompoundingthatappearstobe permissiveinthisrespect.AsfarasDutchisconcerned,itisfarlesscommontousephrasalunits asbasesofderivation,althoughthisispossibleforafewsuffixes,suchasdenominaler ‘er’and denominalachtig ‘like’.Thedenominalsuffixer inDutchcombineseasilywithclassifying phrases:

(16) [[derdeklass] NP er] N‘thirdformpupil’ de [[17 eeuw] NP er] N ‘personlivinginthe17thcentury’

10 [[doehetzelv] Ser] N‘doityourselfer’ Inthelinguisticliteratureonefindsanextensivedebateontheconsequencesofphrasal compoundsforourviewofthearchitectureofthegrammar(Meibauer2007).Wieseclaimedthat “amechanismsuchasquotationmustbeheldresponsiblefortheexistenceofphrasal compounds”(Wiese1996).Meibauer(2007)objectedtothisclaim,andnotedthatoftentheuse ofphraseswithincompoundshasanexpressivefunction,anddoesnotalwaysleadto conventionalizationintolexicalunits.Expressivenessisnot,however,aformalpreconditionfor usingphrasesasbuildingblocksofwords,aswewillseebelow. Theexpressivefunctionofthistypeofwordformationisdominantwhensentencesare embeddedincompounds,asinthefollowingEnglishexamples: (17) Don’ttell,don’taskpolicy(USarmywayofdealingwithgaysoldiers) Onesizefitsalleducation( Boston Globe 6March2010) Iunderstandthewhole‘liveitup,you’reonlyincollegeonce’thing( Tufts Daily April7, 2010) Theeatyourspinachapproachtoeducation( Boston Globe 13March2010) Thesententialconstituentsinthesecompoundscanbecharacterizedas‘fictiveinteraction’ (PascualandJanssen2004),andtheyhaveindeedastrongexpressivecharacter.Someofthese compoundswithasententialconstituentareconventionalized,asisthecaseforthefollowing Dutchcompounds: (18) Godisdoodtheologie‘Godisdeadtheology’ blijfvanmijnlijfhuis‘lit.stayawayfrommybodyhouse,houseforthreatenedwomen’ Notallphrasalcompoundspossessthepropertyofexpressiveness,however.Whatone mayexpectisthatwhenaparticularphrasalpatternisusedsystematicallyforthepurposeof classification(thatis,fornamingconcepts),theresultingtypeofphrasalcompoundwillbequite normal,andwillnotcarryaparticularexpressivevalue.Thisisindeedthecaseforrightheaded phrasalcompoundswithANphrasesastheirleftconstituent,asisillustratedhereforDutch:

(19) [[blinde Adarm N]NP [ontsteking] N]N‘blindintestineinflammation,appendicitis’

[[centrale Averwarmings N]NP [monteur] N]N‘centralheatingtechnician’

11 [[derde A wereld N]NP [land] N]N‘thirdworldcountry’

[[ronde A tafel N]NP [conferentie] N]N‘roundtableconference’

[[vrije Atijds N]NP kleding] N]N‘freetimeclothing,casualdress’

[[warme Agolf N]NP [stroom] N]N‘warmgulfstream’ Intheseexamples,theANphraseisaconventionallexicalunit.However,itisnotthecasethat DutchANphrasescanonlyfeedcompoundingwhentheyareconventionalnamesforconcepts. ThecrucialpropertyofANphrasesaspartsofcompoundsisthattheycanfunctionasclassifiers. Hence,therearealsophrasalcompoundsinwhichtheleftANconstituentisnotaconventional lexicalunitbyitself.ThisisthecaseforthefollowingDutchphrasalcompounds:

(20) [[drie NUM sterren N]NP [hotel] N]N‘threestarshotel’

[[lange Aafstands N]NP [loper] N]N‘longdistancerunner’

[[twee NUM componenten N]NP [lijm] N]N‘twocomponentsglue’

[[vier NUM kleuren N]NP [druk] N]N‘fourcoloursprint’

[[witte Aboorden N]NP [criminaliteit] N]N‘whitecollarscriminality’ Forinstance,thereisnolexicalunit drie sterren ‘threestars’inDutch.Crucially,theseAN phrasesembeddedincompoundsdonotfunctionasreferringexpressions,butspecifythetypeof N,whereNistheheadofacompound.ThisisalsowhytheNPmodifierinthesecompounds cannotcontainadeterminer,asthiswouldimplyareferentialinterpretationofthemodifyingNP. ItisrevealingtocomparethisabsenceofadeterminertotheuseofANcombinationssuchas blinde darm ‘blindgut’whennotembeddedinacompound.Since darm ‘intestine’isacount noun,itmustbeprecededbyadeterminer,andthedeterminerlessphrase blinde darm istherefore illformed,unlessitisembeddedinaword.Yet,theseANsequencesarephrases,sincethe adjectiveisinflected.Theendingein blinde ,forinstance,isrequiredsince darm isasingular commonnoun,andDutchadjectivesagreeingenderandnumberwiththeirheadnoun. Agreementisaclearindicationofphrasalstatus,sincetherecannotbeagreementbetweenparts ofwords,asthiswouldgoagainstLexicalIntegrity,thedefiningpropertyofwordhood. Wethusobservethataparticulartypeofwordstructure,compounding,licensesa particulartypeofphrasalstructure:ANphrasesheadedbycountnounsbutwithoutdeterminers. Thisoncemoreunderscoresthepointthatschemasformorphologicalstructureandthosefor phrasalstructureareinterdependentintheformationoflexicalunits.

12 5. Form-meaning asymmetries Insection4wesawthatphrasescanfeedwordformation.Thisstrengthensthepositionthat morphologyandsyntaxcannotbecompletelyseparated.Butwhathappensifawordformation processdoesnottakephrasesasbasesforaffixation?Considerthefollowingexamplesfrom Italian: (21) chitarraelettrica‘electricguitar’ chitarristaelettrico‘electricalguitarist’ flautobarocco‘baroqueflute’ flautistabarocco‘baroquefluteplayer’ tennisdatavolo‘tennistable’ tennistadatavolo‘tabletennisplayer’ ThewordsequencesontheleftandtherightareNPsoftheformA+NorNPrepN.Thephrase chitarrista elettrico denotessomeonewhoplaystheelectricguitar.However,thephrase chitarra elettrica ‘lit.guitarelectric,electricguitar’isnottheformalbasefortheattachmentofista ,a suffixthatisusedinItaliantocreatepersonalnames.Thisisclearfromthewordorder(thesuffix isnotattachedattherightedge,* chitarra elettricista ),andfromthefactthattheadjective elettrico agreesinnumberandgenderwiththemasculinenoun chitarrista ,andnotwiththe femininenoun chitarra .Thisasymmetrybetweenformandmeaning(‘bracketingparadox’)isthe sameasthatinthefamousEnglishexample transformational grammarian discussedinSpencer (1988).InthecaseofEnglish,onemightconsiderthesuffixian tobeformallyattachedtothe phrase transformational grammar ,anoptionthatisavailableduetoEnglishwordorderandthe poorinflectionofEnglish.SuchananalysisisobviouslyimpossibleinthecaseofItalian,which onlyallowsforaphrasalinterpretation:

(22) [[chitarrista] N[elettrico] A]NP Thissuggeststhattheadjective elettrico hassemanticscopeoverapartofthecomplexphrasal head guitarrista ,andthattheinternalmorphologicalstructureofthemodifiedheadwordis accessible.Inexample(22),theadjective elettrico modifiesthenominalbaseoftheheadnoun.In thefollowingexamplesANphrasesusedasmodifiersareturnedintoAAsequences,inorderto fitthecanonicalphrasestructureofItalian(examplesprovidedtomebyDanieleVergilitto):

(23) energiasolare‘solarenergy’ impianto Nenergetico Asolare A‘solar powerplant’

13 chirurgiaNestetica A‘aestheticsurgery’ intervento Nchirurgico Aestetico A ‘aestheticsurgeryinvention’ ThesametypeofbracketingparadoxhasbeenobservedforPolish.InPolishrelational adjectivesfollowtheheadnoun.Hence,thestructuralparadoxisthesameasforItalian (Szymanek2010): (24) chirurgiaplastyczna‘plasticsurgery’ chirurgplastyczny‘plasticsurgeon’ pracafizyczna‘manualwork’ pracownikfizyczny‘manualworker’ Intheseexamplesthepostnominalrelationaladjectivesagreeingenderwiththeprecedinghead noun.Hencethevariationinformoftherelationaladjectives plastycznand fizyczn . Letuslookatthesefactsfromtheperspectiveofhowtoencodemeaning.WhenanItalian speakerknowsthelexicalphrase guitarra elettrica ,and(s)hewantstoconstructthelexicalunit fordenotingapersonplayinganelectricguitar,thesuffix -ista thatisusedforcreatingpersonal namescannotbeattachedtothephrase guitarra elettrica becausethissuffixisattachedtowords only:thenoun* gitarra-elettric-ista isillformed.Thealternativeisaphrasewithaformmeaning mismatchinwhichthe(stemformsofthe)relevantlexemes( guitarra and elettrico )thatforma lexicalcollocation,areinsertedintheavailableslotsinaschemathatcomplieswiththe

restrictionthatista doesnotacceptphrasalbases.Thisistheschema[[Nista ]NA] NP .The semanticpropertytobespecifiedisthattheAdjectivehassemanticscopeoverthebaseNonly. Productivepatternswiththiskindofasymmetrycanbeaccountedforbyspecifyinga paradigmaticrelationshipbetweentwophrasalschemas(Booij2010),where≈denotesa paradigmaticrelationship:

(25) <[N j Ak]NPi ↔SEM i>≈<[[N jista] N Ak]NPl ↔[PERSONwithrelationRto[SEM i]] l> Anotherexampleofformmeaningasymmetriesisformedbywordsderivedfromparticle verbsinScandinavianlanguages.IntheMainlandScandinavianlanguagesDanish,Norwegian, andSwedish,theparticlefollowstheverb,asisthecaseforEnglish.However,theparticle precedestheverbinpastandpresentparticipleswhichhavebothverbalandadjectivalproperties, andindeverbalnouns(Allan,HomlesandLundskaerNielsen1995;Faarlund,LieandVannebo 1992;HolmesandHinchcliffe2003).Herearesomeexamplesofparticiplesofparticleverbsin thesethreelanguages:

14 (26) particle verb past participle Danish faldened‘falldown’ nedfaldne(æbler)‘windfall(apples)’ Norwegian bringud‘deliver’ Postenbliverudbragt‘Themailisbeingdelivered’ Swedish köraom‘overtake’ denomkördalastbilen‘theovertakenlorry’ ringup‘phone’ uppringd‘phoned’ InthefollowingdeverbalnounsinNorwegian(Faarlund,LieandVannebo1992),weseeagain thereversaleffect: (27) particle verb nominalization bindeinn‘tobind(abook)’ innbinding‘bookcover’ ryddeopp‘totidyup’ opprydding‘tydingup’ skriveav‘tocopy’ avskrift‘copy’ trekkeopp‘topullup’ opptrekker‘corkscrew’ Thedeverbalsuffixescannotbeattachedtotheparticlesinceitisnotaverb,orpartofaverb.If thesuffixwereattachedtotheverb,wewouldgetformslike bind-ing-inn .Asthereareno particlenouns,particlescannotoccurafteranoun.AndsinceNorwegiancompoundsareright headed, bind-ing-inn cannotbeinterpretedasacompound.Therefore,thenominalizationshave theformofacompound,consistingofaparticlefollowedbythedeverbalheadnoun.Hence,the formalbaseofthesenominalizationsisnottheparticleverbassuch,eventhoughthemeaningof thesenominalizationsisacompositionalfunctionofthe(oftenidiomatic)meaningofthe correspondingparticleverb.Inthenominalizations,thelexicalcollocationoftheparticleandthe verbispreserved,butinsertedinthereverseorder,intheslotsoftheavailablecompoundtype[X

N] N,whichisthegeneraltemplatefornominalcompoundsinGermaniclanguages.Thereisa formmeaningasymmetry,sincetheparticleverbisnota(nondiscontinuous)subconstituentof thenominalization.Yet,themeaningofthesenominalizationsisregularandtransparentasthey relateinasystematic,yetparadigmaticwaytothecorrespondingparticleverbs.

15 Thesameexplanationcanbeinvokedforthereversaleffectinparticiplesofparticle verbs.Sinceparticiplesareadjectives,theonlypossibleshapeofadjectivalformsofparticle formsisthatofarightheadedadjectivalcompound. Thefollowingformmeaningasymmetrieshavebeenobservedinrelationtoparticle verbsinEnglish,Dutch,German,andItalian(Booij2010:189): (28) a.Englishparticleverbsmaycarrythesuffixeser anding ontheirverbalhead: look-er- on, runn-er-up, digg-ing up, switch-ing off the lights ; b.thepastparticipleofDutchparticleverbsisformedbyprefixing ge andsuffixingt/- d/-en tothestemformoftheverbalhead: aan-val ‘toattack’ - aan-ge-vall-en , op-bellen ‘tophoneup’op-ge-bel-d); ge nominalizationalsoappliestothehead: rond-spring ‘jumparound’ - rond-ge-spring ‘jumpingaround’; c.whenGermanparticleverbsundergonominalizationwiththeaffixcombination ge-e, thisaffixcombinationisattachedtotheverbalheadoftheparticleverb: herum-hops-en ‘tojumparound’ Herum-ge-hops-e‘jumpingaround’(Müller2003;2006); d.inItalian,nominalizingsuffixesareattachedtotheverbalhead(FrancescaMasini,pers. comm.): venire giù ‘tocomedown’la venuta giù ‘thecomingdown’, mangiare fuori ‘to eatout’la mangiata fuori ‘themealatarestaurant’; Inallcases,canonicalmorphologicalformsareusedforcreatingdenominalformsofparticle verbs.Notethat,sinceItalianhasleftheadedcompounds,theparticlecanappearafterthe deverbalnoun,unlikewhatisthecaseforGermaniclanguagessuchasNorwegian,wherea compoundlike[[ bind-ing ]inn]isillformed.Therelevantpatternscanbeaccountedforinterms ofparadigmaticrelationsbetweenschemas.Forexample,therelationbetweenGerman herumhopsen and Herumgehopse in(28c)canbeexpressedbyaparadigmaticrelationship symbolizedbythe≈betweentwoschemas:

(29) <[Part iV j]k↔[SEM] k>≈<[Part i[geVje]N]Nl↔[NOMSEM k]l

NOMstandsforthesemanticoperatorofeventnominalization.Thestructure[Part i[ ge Vje]N]N in(29)isarightheadedcompoundstructureofthetype[XN] N,inwhichtheheadisadeverbal nounwiththediscontinuousaffix ge..e .Whatisspecialaboutthestructureofthesecompoundsis thatitssemanticcorrelaterequiresreferencetotheofaparadigmaticallyrelated structure,thatofparticleverbs.In(29)wespecifythatthiseventnominalizationofGerman

16 particleverbsisformallyaheadoperation:itistheverbalheadthatisnominalized,butthe nominalizingelementhassemanticscopeovertheparticleverbasawhole. Inconclusion,theflexibilityinencodinglexicalunitsfordenotingconceptsisgreatly enhancedbyallowingforcertaintypesofasymmetriesbetweenformandmeaning,mismatches thatcanberesolvedbythelanguageuseronthebasisofparadigmaticrelationshipsbetween (morphologicallyorsyntactically)complexexpressions.Theseobservationsalsoshowthatthe internalstructureofalexicalunitisnotbydefinitionopaqueasaneffectofstorage:conventional lexicalunitsmaypreservetheirformalandsemantictransparency. 6. Conclusions Thenotion‘construction’thatplaysacentralroleinConstructionGrammar,isindispensablefor theanalysisofwordformationpatternsaswell.Inthestudyofwordformation,weinvestigate thesystematiccorrespondencesbetweenformandmeaningatthewordlevel.Constructional schemasprovideanadequateformatforexpressingthesesystematiccorrespondences.Moreover, theyarepartofahierarchicallexiconinwhichbothcomplexwordsandmorphologicalpatterns ofvariouslevelsofabstractionarespecified. AnimportantadvantageoftheConstructionMorphologyapproachisthatitcanexpress similaritiesbetweenmorphologicalandphrasallexicalexpressions,andtheparadigmatic relationsbetweenmorphologicalandphrasalschemas.Thus,lexicalknowledgeischaracterized asacomplicatednetworkbetweenwordsandphrasalexpressionsatarangeoflevelsof abstractions,varyingbetweenindividualwordsandcompletelyabstractpatterns. 7. References Allan,Robin,PhilipHolmesandTomLundskaerNielsen 1995 Danish, a Comprehensive Grammar .London/NewYork:Routledge. Aronoff,Mark 1976 Word Formation in .Cambridge,Mass.:MITPress. Booij,Geert 2002 Constructionalidioms,morphology,andtheDutchlexicon. Journal of Germanic Linguistics 14:301327. Booij,Geert 2009 Phrasalnames:aconstructionistanalysis. Word Structure 3:219240. Booij,Geert 2010 Construction Morphology .Oxford:OxfordUniversityPress.

17 Erman,BrittandBeatriceWarren 2000 Theidiomprincipleandtheopenchoice principle. Text 20:2962. Faarlund,JanTerje,SveinLieandKjellIvarVannebo 1992 Norsk Referansegrammatikk . Oslo:Universitetsforlaget. Goldberg,Adele 2006 Constructions at Work. The Nature of Generalization in Language .Oxford:OxfordUniversityPress. Holmes,PhilipandIanHinchcliffe 2003 Swedish, a Comprehensive Grammar .London/ NewYork:Routledge. Hüning,Matthias 2010 Adjective+Nounconstructionsbetweensyntaxandword formationinDutchandGerman.In:SaschaMichelandAlexanderOnysko(eds.),Cognitive ApproachestoWordFormation,195218.Berlin:DeGruyterMouton. Jackendoff,Ray 2002 Foundations of Language .Oxford:OxfordUniversityPress. Jackendoff,Ray 2009 Compoundingintheparallelarchitectureandconceptual semantics.In:RochelleLieberandPavolŠtekauer(eds.),TheOxfordHandbookof Compounding,105129.Oxford:OxfordUniversityPress. Jackendoff,Ray 2010 Meaning and the Lexicon. The Parallel Architecture 1975-2010 . Oxford:OxfordUniversityPress. Jackendoff,Ray 2011 FormalCxG:Constructionsintheparallelarchitecture.In:T. HoffmannandG.Trousdale(eds.), The Oxford Handbook of Construction Grammar ,xxxxxx. Oxford:OxfordUniversityPress. Joeres,Rolf 1995 Wortbildungen mit -macher im Althochdeutschen, Mittelhochdeutschen und Neuhochdeutschen .Heidelberg:UniversitätsverlagC.Winter. Kuiper,Koenraad,MarieElaineVanEgmond,GerardKempenandSimoneSprenger 2007 Slippingonsuperlemmas.Multiwordlexicalitemsinspeechproduction. The Mental Lexicon 2: 313357. Los,Bettelou,CorrienBlom,GeertBooij,MarionElenbaasandAnsVanKemenade 2012 Morphosyntactic Change: A Comparative Study of Particles and Prefixes .Cambridge: CambridgeUniversityPress. Meibauer,Jörg 2007 Howmarginalarephrasalcompounds?Generalizedinsertion, expressivity,andI/Qinteraction. Morphology 17:233259. Müller,Stefan 2003 ThemorphologyofGermanparticleverbs:solvingthebracketing paradox. Journal of Linguistics 39:275325. Müller,Stefan 2006 Phrasalorlexicalconstructions? Language 82:850883. Pascual,EstherandTheoJanssen 2004 Zinneninsamenstellingen.Presentatiesvan fictieveverbaleinteractie. Nederlandse Taalkunde 9:285310.

18 Schlücker,BarbaraandIngoPlag 2011 Compoundorphrase?Analogyinnaming. Lingua 121:15391551. Speelman,Dirk,JoséTummersandDirkGeeraerts 2009 Lexicalpatterningina constructiongrammar.Theeffectoflexicalcooccurrencepatternsontheinflectionalvariationin Dutchattributiveadjectives. Constructions and Frames 1:87118. Spencer,Andrew 1988 BracketingparadoxesandtheEnglishlexicon. Language 64: 663682. Sprenger,SimoneA 2003 Fixed Expressions and the Production of Idioms .Nijmegen: MaxPlanckInstitutfürPsycholinguisitk. Sprenger,Simone,WillemJ.M.LeveltandGerardKempen 2006 Lexicalaccessduring theproductionofidiomaticphrases. Journal of Memory and Language 54:161184. Szymanek,Bogdan 2010 A Panorama of Polish Word-formation .Lublin:Wydawnictwo KUL. Tremblay,AnnieandR.HaraldBaayen 2010 Holisticprocessingofregularfourword sequences:Abehavioralanderpstudyoftheeffectsofstructure,frequency,andprobabilityon immediatefreerecall.In:DavidWood(ed.), Perspectives on Formulaic Language. Acquisition and Communication ,15173.London:TheContinuumInternationalPublishingGroup. Wiese,Richard1996 Phrasalcompoundsandthetheoryofwordsyntax.LinguisticInquiry27: 183193. Wray,Alison 2002 Formulaic Language and the Lexicon .Cambridge:Cambridge UniversityPress. Wray,Alison 2008 Formulaic Language: Pushing the Boundaries .Oxford:Oxford UniversityPress. Geert Booij, Leiden (Netherlands)

19