generalizationlavalava

AbstractInthisstudyIpresentacomparativeandhistoricalanalysisof''fre-quentative''Bantuverb-stemreduplication,manyofwhosevariantshavebeendescribedforanumberofEasternandSouthernBantulanguages.
Whilesomelanguageshavefull-stemcompounding,wherethestemconsistsoftheverbrootplusanyandallsufxes,othersrestrictthereduplicanttotwosyllables.
Twoquestionsareaddressed:(i)WhatwastheoriginalnatureofreduplicationinProto-Bantu(ii)WhatdiachronicprocesseshaveledtotheobservedvariationIrstconsiderevidencethatthefrequentativebeganasfull-stemreduplication,whichthenbecamerestrictedeithermorphologically(byexcludinginectionalandulti-matelyderivationalsufxes)and/orphonologically(byimposingabisyllabicmaximumsizeconstraint).
Ithenturntotheoppositehypothesisandconsiderevidenceandmotivationsforaconictingtendencytorebuildfull-stemredupli-cationfromthepartialreduplicant.
Iendbyattemptingtoexplainwhythepartialreduplicantisalmostalwayspreposedtothefullerbase.
KeywordsPartialreduplicationBantuVerbstemDerivationInectionBisyllabicfoot1IntroductionAsAshton(1944,p.
316)succinctlyputsit,''REDUPLICATIONisacharacteristicofBantulanguages.
Itaffectssyllables,verbstems,words,andphrases.
''TraditionalBantugrammarsthusoftenincludesectionsshowingthatverbs,nouns,adjectives,L.
M.
Hyman(&)DepartmentofLinguistics,UniversityofCalifornia,Berkeley,CA94720-2650,USAe-mail:hyman@berkeley.
edu123Morphology(2009)19:177–206DOI10.
1007/s11525-009-9140-yORIGINALPAPERThenaturalhistoryofverb-stemreduplicationinBantuLarryM.
HymanReceived:2January2008/Accepted:2March2009/Publishedonline:11November2009TheAuthor(s)2009.
ThisarticleispublishedwithopenaccessatSpringerlink.
comnumeralsandevenpronounsanddemonstrativescanbereduplicatedwithspecicsemanticeffects.
Inthis,asinmostaspectsofBantugrammar,thestoryisoneofthemeandvariations:Thereduplicationsinquestionshowgreatsimilaritybothinstructureandinmeaning,butalsointerestingdifferences.
Nowherearethesedif-ferencesmorepronounced-ormoresignicant—thaninverbreduplication.
WhilesomeBantulanguagesexhibittotalreduplicationofthestemconstituentconsistingoftheverbroot+sufxes,othersplacemaximumsizeconstraintsontheredupli-cantand/ordisallowcertainsufxes,e.
g.
inectionalendings,fromappearingwithinit.
ThemajorgoalofthispaperistoattempttomakesenseoutofthisvariationtodeterminethenatureofverbreduplicationinProto-Bantu(PB).
Inordertodoso,Istartbycataloguingthevariousrestrictionstheca.
500daughterlan-guagesplaceonthereduplicantinSect.
2.
IthenconsiderthepossiblerelationoftotalreduplicationtofullwordrepetitioninSect.
3.
InSect.
4Iconsidertheevi-dencethatthedirectionofchangewasfromfull-stemtopartial-stemreduplication.
Whilethesimplestexplanationistoassume,alongwithotherscholars,thatpartialreduplicationderiveshistoricallyfromfullreduplication(Eulenberg1971;Bybeeetal.
1994;Niepokuj1997),inSect.
5Ipresentevidencethattotalverb-stemreduplicationcanbebuiltupfrompartialreduplication.
IthereforesuggestinSect.
6thatthechangeshavebeenbidirectionalandconcludeinSect.
7byaddressingthequestionofwhypartialreduplicantsdevelopontheleftofthefullbase.
2Pan-Bantuverb-stemreduplicationAsiswellknown,Bantulanguageshaveahighlyagglutinativewordstructure.
Inordertoappreciatetheissueswhichariseinverb-stemreduplication,thetraditionalinternalconstituentsoftheverbinProto-Bantu(PB)andmostofthedaughterlanguagesareshownin(1),asreconstructedbyMeeussen(1967):1brevpre-stemstembaseFV(="nalvowel")radicalextensionsThedifferentsubconstituentsareidentiedin(2),wheresomeofthecommonPBsufxesarealsoindicated:(2)a.
pre-stemsubject,negative,tense,aspect,objectprexesb.
radicalverbroot(-CVC-isthemostcommonshape)c.
extensionsderivationalsufxes(causative*-Is-i-,applicative*-Id-,reciprocal*-an-,passive*-f-);frozensufxesandpost-radicalunanalyzable''expansions''178L.
M.
Hyman123d.
FVobligatoryinectionalnalsufx(past*-I,subjunctive*-e,perfective*-id-e,imperfective*-ag-a;otherwiseFV=default*-a)AnexamplefromHayaisgivenin(3).
(3)ti-ba-ka-ki-[kom]-el-angan-ag-a'theyhavenevertieditforeachNEG-SUBJ-PAST-OBJ-tie-APPL-RECIP-HAB-FVother'(appl=applicative;HAB=habitual)Asseen,aBantuverbcanbequitelongandinvolvebothmultipleprexesandsufxes.
GiventhestructureoftheBantuverbin(1),threequestionsnaturallyarise.
First,whichmorphologicalconstituentin(4)wasavailableforverbreduplicationinPB(4)a.
thewholeword(prexes+root+extensions+FV)b.
theverbstem(root+extensions+FV)c.
theverbbase(root+extensions)d.
theverbroot(perhapswithalinkervowel,e.
g.
CVC-a-CVC-a)Besidesthesubconstituentsof(1),thereareotherpossibilitiesaswell.
AsseenintheHayaexamplein(3),theobjectprex(OBJ)occursclosesttotheverbstem.
InsomeBantulanguagestheOBJ+stemishasbeentreatedasa''macro-stem''con-stituentforthepurposeoftoneassignmentor,asweshallsee,reduplication.
Whilethequestionin(4)concernsthemorphologicalconstituentwhichservedasinputtoreduplicationinPB,i.
e.
whichpartsoftheverb(prexes,root,sufxes)couldpotentiallybecopied,asecondquestionconcernstheEXTENToftheredupli-cation:Couldtheconstituentinquestionbefullycopied,orwasreduplicationonlypartialNotonlydopresent-dayBantulanguagesdifferfromeachotheronthisscore,butthereisevenvariationwithinthesamelanguage,asseenintheTswanaexamplesin(5),wherereduplicationmarksthefrequentative(Cole1955,p.
217):(5)rek-el-a'buyfor'!
a.
rek-el-a+rek-el-a(Htonespreadingbuy-APPL-FV!
b.
rek-e+rek-el-anotshown)!
c.
rek-a+rek-el-aWhilethefullverbstemisreduplicatedin(5a),onlythersttwosyllablesappearinthepreposedreduplicant(RED)in(5b).
(5c)showsthatREDcanalsoconsistofa-CVC-verbrootplusthelinkervowel-a-.
Thereisnodifferenceinmeaningbetweenthesethreevariants.
Thenalquestionconcernswhichpartofthereduplicatedwordistheredupli-cantParticularlywhenreduplicationistotalitmaynotbeimmediatelyobviouswhetherthestructureisRED-baseorbase-RED.
Infact,threedifferentstructuresappeartobeneeded,asinTswana:Thenaturalhistoryofverb-stemreduplicationinBantu179123(6)a.
asinallverbstems,allbuttherstandlastvowelsofRED+stemareunderlyinglytoneless,includingthesecondverbroot,suggesting[[[rek]-el-a-rek-el]-a]b.
segmentaltruncationoccursonrstpart,suggestingREDispreposed(''prexed'')tothebase:[[rek-e][rek-el-a]].
c.
sufxesareoftentruncated,suggestingrootreduplication:[[rek-a-rek]-el-a]Whileitremainstobedeterminedexactlywhatshouldbereconstructed,itisreasonabletoassumethatsomekindofverbreduplicationwaspresentinPB.
CertainlythisistheimpressiononegetsfromthecomparativeandhistoricalBantumanuals.
Meinhof(1932,p.
46)specicallyreferstothekindofstemreduplicationwearemostlyconcernedwithhere:''Completeandincompletereduplicationoftheverb-stemoccurswithaniterative-intensiveforce,e.
g.
Yaolava-lava'gofromplacetoplace',from-lava'startearlyinthemorning',Swahili-cekaceka'laughcontinuously'fromceka'laugh'.
''Meeussen(1967,p.
88)alsopointsoutthatBantulanguagesoftenhavelexicalizedverbrootswithCV-reduplications:''Somereconstructions.
.
.
implyatypeofradicalwithinitialreduplication(-cvCVC-):-titim-'befrightened',-tetem-'tremble',-pepet-'winnow',-tftfm-'tremble,quake,thunder'.
Thetranslationssuggestmeaningswithacommonelement'movementtoandfro'.
''Thefewlexicalizedstem-orCV-reduplicationspresentintheca.
10,000BantureconstructionsinBastinandSchadeberg(2003),generallyhaveregionalorinfrequentattestations.
Someexamplesaregivenin(7).
(7)a.
*jag-a-jag-'berestless'*jag-'scratch,becrampedforroom'*dng-a-ding-'windroundseveraltimesorwrong'*dng-'twist,wrapup,surround'b.
*de-demb-'swing,hang,oat'*demb-'behungup,hover,swing'*mu-mun-'suckinmouth'*mfn-'suck'Still,theintuitiveappealofpositingsuchastructureinPBderivesfromthefactthatproductiveverb-stemreduplicationisfoundwithsimilarmeaningsinallBantuzones.
Representativeexamplesareprovidedin(8).
(8)a.
Duala(Cameroon):''todosomethingrepeatedlywithoutaim''(Meinhof1912,p.
67);''arepeated(orfrequent)andsenselessaction''(Ittmann1939,p.
195)1tomb-a-tomb-ane'passbyoverandoveragain'tomb-a`'passby'yeng-a-yeng-ane'wanderwithoutaim'yeng-a'walk'b.
Gunu(Cameroon):''acontinuousorsustainedaction''(Rekanga1989,p.
141)1Thereciprocalsufx-aneisrequiredinverb-stemreduplicationinDuala.
180L.
M.
Hyman123bel-a-bel-a'liecontinually'bel-a'lie'namb-a-namb-a'prepare(continousaction)'namb-a'prepare'c.
Lingala(DRC):''arepeated.
.
.
intensive.
.
.
(or)futileaction''(Guthrie1939/1966,p.
37)bet-a-bet-a'giveagoodbeatingto'bet-a'beat'lob-a-lob-a'talk,chatter'lob-a'speak'd.
Nyanga(DRC):''anintensivevalue.
.
.
doalot.
.
.
severaltimes''(Mateene1969,p.
82)i-k-a-ki-a'faireetrefairesanscesse'i-k-a'faire'i-but-a-but-a'enfanterunpeutrop'i-but-a'enfanter'e.
Rimi(Tanzania):''frequencyofactionwith.
.
.
attenuation,purposelessness,errative''(Olson1964,p.
161)ng-a-ng-a'trip,closefrequently'ng-a'trip,close'mank-a-mank-a'trot'mank-a'run'f.
Mbukushu(Namibia):''frequencyorintensicationofanaction''(Fisch1998,p.
123)korw-a-korw-a'alwaysbesick'korw-a'besick'kwat-a-kwat-a'toucheverything'kwat-a'touch'g.
Yao(Mozambique):''action.
.
.
isrealizedrepeatedlyorfrequentlyatlargeintervals''(Ngunga2000,p.
107)lim-a-lim-a'cultivatefrequently'lim-a'cultivate'lokot-a-lokot-a'pickuprepeatedly'lokot-a'pickup'.
h.
Venda(SouthAfrica):''anactionthatiscarriedoutfrequentlyorrepetitively.
Sometimes.
.
.
aimlessly,orindiscriminately.
''(Poulos1990,pp.
195–6)lim-a-lim-a'ploughsporadicallyhereandthere'lim-a'plough'dzul-a-dzul-a'sithereandthere'dzul-a'sit'Bycontrast,CV-reduplicationsareusuallylexicalized,sometimeswithan''intru-sive''-lV,asinYaogala-gaat-a'rollontheground',Nulu-Nuund-a'scrapeout'kolo-koosol-a'shell'(Ngunga2000,p.
114).
Whileitisnotclearwhetherthetworeduplicativepatternsarehistoricallyrelated,doubletsdoappearinsomelanguages,e.
g.
inKinyarwanda:ku-beera-beera~ku-be-beera'towalkaimlessly',ku-buuta-buuta~ku-buu-buuta'towalkbackbent'(Kimenyi2002,p.
260).
ThesharedpropertythatcanbeisolatedthroughoutBantuisthatverbreduplicationprimarilyoronlytargetsthestem.
Asweshallsee,prexalmaterialonlysporadicallymakesitintoRED,andonlywhenmotivatedbyphonologicalconstraints.
Althoughthereissomequestionaboutwhetherthepre-stemelementswereprexesorseparatecliticsorwordsinProto-Bantu(seeHyman2007;Nurse2007),itisclearthattheyAREprexesinmanyBantulanguages,e.
g.
inLuganda(HymanandKatamba2005).
Thus,thegeneralexclusionofprexesinreduplicationcouldhaveoneoftwoThenaturalhistoryofverb-stemreduplicationinBantu181123diachronicexplanations:First,ifMeeussen'sreconstructionin(1)iscorrect,reduplicationmayhaveoriginallytargetedasub-constituentoftheverbin(4b–d).
Alternatively,ifthepre-stemmarkerswerenotprexesatthetimeverbredupli-cationwasintroduced,theywouldhaveautomaticallybeenexcluded.
However,wewouldhavetoaddthatoncethesemarkersbecameprexes,thequestionwouldbewhytheydidnotlaterbecomeincorporatedintoRED.
Thissecondexplanationmostnaturallytstheclaimeduniversalthatreduplicationalwaysbeginsastotalcopyandthatchangesfollowtheuniversallyunidirectionalprocessoffull>partialreduplication(Niepokuj1997).
InthenextsectionIconsiderthelogicandevidenceforfullverbreduplicationinPB.
3ReduplicationversusrepetitionTorecapitulate,andsettingasidetheissueofthemacro-stemforthemoment,weareconsideringfourhypothesesconcerningtheoriginofverbreduplicationinBantu:(9)a.
PBreduplicationtargetedthewholeverb(prexes+root+extensions+FV)b.
PBreduplicationtargetedtheverbstem(root+extensions+FV)c.
PBreduplicationtargetedtheverbbase(root+extensions)d.
PBreduplicationtargetedtheverbrootonly(e.
g.
-CVC-a-CVC-)Thereareseveralintuitiveargumentsforwhole-verbreduplication,bothuniversalandBantu-specic.
First,thereisthepossibilityofderivingreduplicationfromwordrepetition,asinYoutalk,talk,talkallthetime!
Second,thereisthepossibilityofmaintainingunidirectionalitiy,i.
e.
*full>partialreduplication(Niepokuj1997).
Third,thereissynchronicevidencethatatleastsomecasesofpartialreduplicationshouldbetreatedas''morphologicaldoubling''(InkelasandZoll2005;cf.
Steriade1988).
AsEulenberg(1971,p.
73)putsit,casesofso-calledpartialredupli-cationaresimplyphonological[andmorphological]reductions,sometimesdrastic,fromcasesoffullreduplications.
''Concerningthepossiblediachronicprocessrepetition>reduplication,manyBantulanguagesareknownforrepeatingfullwordsandphrasesinspontaneousdiscourse.
ExamplesfromTotelaareprovidedin(10)(TheraCrane,personalcommunication):(10)a.
Aboba-ku-tutulukaku-kulaku-kulamuzitheythey-NARR-come.
outNARR-cleanNARR-cleanvillageku-kulamuzi'TheycameoutandcleanedupNARR-CLEANvillagethewholevillage.
'b.
Ku-mu-busa-busaku-mu-busa-busaba-ku-mu-taNARR-her-greet-greetNARR-her-greet-greetthey-NARR-her-tell'Theygreetedher(enthusiastically,overandover)andtheytoldher.
.
.
'182L.
M.
Hyman123Whilethereisanobvioussemanticlinkbetweenrepetition,asintheaboveexamples,andthefrequentativemeaningsseenintheexamplesin(8),therearepotentialdifferences.
Fortune(1982,p.
49),forexample,statesthat''themeaningofthereduplicatedRdiffersfromthatoftheRrepeated,''providingexamplessuchasthefollowingfromShona:Rova!
Rova!
'Hit!
Hit!
'versusRov-a-rov-a'hitalloverindiscriminately!
'.
Intherstcaseeachverbgetsitsowntonalassignment,whileinthesecond,thereisasingletoneassignmentoverthewholeverb.
Onceweleavetheverb,wendthattherearewidespreadcasesoffullwordreduplication(orisit''repetition''),forexample,thewidespreaddistributivenumeralconstructionexempliedin(11)fromKanyok,whichdropsnalvowels(Mukash-Kalel1982,p.
151):(11)a.
ba`-b`dy'two'!
ba`-b`dyba`-b`dy'twobytwo,twoeach'b.
ba`-sa`t'three'!
ba`-sa`tba`-sa`t'threebythree,threeeach'Asseen,theclass2pluralprexba-iscopiedalongwiththenumeralstem.
IamunawareofanyBantulanguagewhichallowsformslike*ba`-b`dy-b`dyand*ba`-sa`t-sa`t.
Eventhereduplicationofnounsandadjectivesismoreamenabletoallowingaprextobecopiedintothereduplicant.
InKinande,nounreduplicationassignsmeaningslike'arealX'or'agoodexampleofanX'versus.
otherBantulanguageswherethemeaningwouldbepejorative(e.
g.
'alousyexampleofanX').
Asseenin(12a),anounprexwillnotbecopiedifthestemisbisyllabic,whichalsoestab-lishesthatREDispostposedtothebase(MutakaandHyman1990,pp.
77–80):2(12)a.
kf-gflf'leg'!
kf-gflf-gflf'arealleg'*kf-gflf-kf-gflfmf-ska'girl'!
mf-ska-ska'arealgirl'*mf-ska`-mf-ska`b.
ka-tI'stick'!
ka-t-ka-tI'arealstick'ri-bwe'stone'!
ri-bwe-ri-bwe'arealstone'c.
m-bflI'goat'!
m-bfl-m-bflI'arealgoat'n-zoka'snake'!
n-zoka-n-zoka'arealsnake'd.
mu-heruki'bride'!
*mu-heruki-herukiki-tembekali'tree(sp.
)'!
*ki-tembekali-tembekaliIn(12b),however,theprexiscopiedwhenthestemismonosyllabic,asitisin(12c),wherethenasalisnon-syllabic,unlessinitial.
ThegeneralizationisthatKinanderequiresabisyllabicRED:eitherthestemin(12a)ortheprex+stemin(12b,c).
(12d)showsthatlongerstemscannotbereduplicated.
Ifthefullstemwerereduplicated,REDwouldhavethreeorfoursyllables.
Truncatedsolutionssuchas*mu-heruki-heruor*mu-heruki-rukiarenotacceptablebecauseofMORPHEMEINTEGRITY:KinandeprohibitsthemappingonlypartofamorphemeintoRED.
2Unlikeverbreduplication,tonesarecopiedinnounreduplicationinKinandeandgenerallyinBantu.
TheKinandeformsarecitedwithoutnalboundarytones.
Thenaturalhistoryofverb-stemreduplicationinBantu183123ThesameREDr–risinforceinKinandeverbreduplication,butwithacrucialdifference.
Asseenin(13a),wheretheverbstemismonosyllabic,thenounclass5innitiveprexrI-cannotbecopiedintoREDaswaspossibleinthecaseofnounsin(12b).
(13)a.
rI-sw-a'togrind'!
(rI-)swa-swa+sw-a'.
.
.
alittlehereandthere'r-tw-a'tocut'!
(r-)twa-twa+tw-ab.
rI-hfm-a'tobeat'!
(rI-)hfm-a+hfm-ar-tfm-a'tosend'!
(r-)tfm-a+tfm-ac.
rI-hfm-Ir-a'tobeatfor/at'!
(rI-)hfm-a+hfm-Ir-ar-tfm-Ir-a'tosendfor/to'!
(r-)tfm-a+tfm-Ir-aAsaresult,theverbstemmustcopytwiceinordertolloutthebisyllabicREDrequirement.
Thebisyllabicstemreduplicatesasexpectedin(13b),whereasredu-plicationispossiblein(13c)onlybecausetheapplicativesufx-Ir-canbetrun-cated.
Aswillbeseeninthefollowingsections,similar''complications''areobservedinotherBantulanguages.
InSects.
4and5weconsidertwohypotheses:rst,thatPBhadfullverb-stemreduplicationwhichhasbeensubjecttotruncations,andsecond,thatPBhadverb-rootreduplicationwhichhasbeensubjecttoaugmentations.
4HypothesisI:full>partialverb-stemreduplicationDespitethesemanticandstructuralsimilaritiesofverbreduplicationwithinBantu,thereareimportantformaldifferencesinthephonologicalsizeofpreposedREDanditsmorphologicalcontents.
First,itshouldbenotedthatsomeBantulanguagesrequirefullverb-stemreduplication.
AsseenintheCiyaoexamplesin(14)(Ngunga2000,pp.
105–107),bothderivationalextensionssuchasapplicative-il-/-el-andnalinectionalendingssuchasperfective-il-earereduplicated.
(Reduplicationsareshownwithoutprexmorphemes,sincetheselatterdonotcopy.
)(14)a.
root+-a:telek-a!
telek-a+telek-a'cookfrequently'b.
root-APPL-a:telec-el-a!
telec-el-a+telec-el-a'cookfor(s.
o.
)frequently'c.
root-PERF:dim-il-e!
dim-il-e+dim-il-e'cultivatedmanytimes'WhilelanguageslikeCiyaoandLugandamustfullyreduplicatetheverbstem,truncationofsufxesinREDisquitewidespread.
Asseenin(15),Odden(1996,pp.
130–145)hascaughtKikereweintransition:184L.
M.
Hyman123(15)a.
ku-lim-il-an-a'tocultivateforeachother'i.
ku-lim-il-an-a+lim-il-an-a(fullstemreduplication)ii.
ku-lim-il-a+lim-il-an-a(reciprocal-an-is''truncated'')iii.
ku-lim-a+lim-il-an-a(applicative-il-andreciprocal-an-aretruncated)iv.
*ku-lim-an-a+lim-il-an-a(applicative-il-truncated)b.
a-lim-l-e'hecultivated'i.
a-lim-il-e+lim-l-e(fullstemreduplication)ii.
a-lim-a+lim-l-e(inectional-ileis''truncated'',-adefault)c.
ku-kalaang-a'tofry'i.
ku-kalaang-a+kalaang-a'tofryanyoldway'ii.
*ku-kala+kalaang-a(fullrootmustreduplicate)Fullverbstemreduplicationisshowninthe(i)examples.
In(15a.
ii)therstofthetwoderivationalextensionshasbeentruncated,while(15a.
iii)hastruncatedboth.
Theungrammaticalformin(15a.
iv)showsthatitisnotpossibletotruncatethesecondextension(reciprocal-an-)withouttruncatingtherst,i.
e.
onceaninputsufxhasfailedtobecopiedinRED,nosufxestoitsrightcanbefurtherconsidered.
Therelevantconstraintisthat''thereduplicantmustcorrespondtoacontiguoussubstringofthederivational[i.
e.
base]stem''(Odden1996,p.
137).
(15b.
i.
)againshowsfullreduplication,thistimeincludingtheperfectiveending-il-e,whichhowevercanbetruncated,asin(15b.
ii),whereREDappearswiththedefaultFV-a.
(15c.
i.
)exempliesthefullreduplicationofalongverbroot,while(15c.
ii)showsthattruncationisnotpossibleiftheresultisaviolationofmorphemeintegrity:''partialcopyingofmorphemesisblocked''(Odden1996,p.
138).
ThusfarwehaveestablishedtwotendenciesthatcontributetoREDbeinglessthantotalreduplication:thetendencyforREDtoconsistoftwosyllablesandthetendencyforREDtoexcludesufxalmaterial(otherthanthedefaultFV-a).
Concerningtherst,thereactuallyarethreedifferentsituations,dependingonwhetherthetwo-syllableREDis(i)aminimum,(ii)amaximum,or(iii)both.
Asseenin(16),whichshowsreexesofstemscontainingthePBroots*-gf-'fall'and*-dIm-'cultivate',allthreesituationsareattested:(16)(i)REDr–r(ii)REDr–r(iii)REDr–re.
g.
Sukumae.
g.
Kinyarwandae.
g.
Ndebelegw-a-gw-a+gw-agw-aa+gw-aan-aw-a-yi+w-a'fallhere&there'lIm-Il-a+lIm-Il-arim-aa+rim-ir-alim-a+lim-el-a'cultivateforhere&there'(i)InSukuma,RED''isnotmaximallydisyllabicalthoughminimallyitmustbeatleastdisyllabic''(Matondo2003,p.
133).
Asseen,REDmaydoublyreduplicateamonosyllabicsteminordertolloutthetwo-syllableminimum.
AnThenaturalhistoryofverb-stemreduplicationinBantu185123alternativeistocopyaprecedingprex,e.
g.
(gf-)gw-a-gw-a+gw-a~gf-gw-a+gf-gw-a'tofallhere&there'(Matondo2003,p.
122).
ThereisnoupperlimitonthesizeofRED,e.
g.
leembeel-el-nij-iw-a+leembeel-el-nij-iw-a('becalm'+APPLICATIVE+SIMULTANEOUS+PASSIVE).
However,whenaverbstemcontainsonlyoneproductivesufx,itmayoptionallybetruncated,e.
g.
lIm-Il-a+lIm-Il-a~lIm-a+lIm-Il-a'cultivateforhere&there'(Matondo2003,pp.
129–130,154).
(ii)KinyarwandaisunusualbothinlimitingREDtorootmaterialonlyandinitstreatmentofsubminimal-CV-roots.
WhileKimenyi(2002)showsthat/gu-a/gw-a'fall'cannotreduplicateas*gw-aa+gwa,Fide`leMpiranya(per-sonalcommunication)pointsoutthatitispossibleforgw-atoredupli-cateasgw-aa+gw-aan-a,wherethebasehasbeenaugmentedbywhatlookslikethereciprocalextension-an-(cf.
/pfu-a/pf-a'die'!
pf-aa+pf-aan-a'wasteaway',i.
e.
'diealittlebithereandthere').
3Asaresult,REDcanbeeithermonosyllabicorbisyllabicinKinyarwanda.
(iii)ThethirdsituationisthemostcommonandisrepresentedbyNdebele,whichhasanabsoluterequirementthatREDconsistoftwosyllables.
Whenthestemismonosyllabic,adummysecondsyllable-yithusllsoutthebisyllabictemplate.
WhilethetendencytowardsabisyllabicREDisaphonologicalcondition,itsome-timesgoeshandinhandwiththesecondtendencytoexcludebothextensionsandinectionalendingsotherthandefault-a.
Theadditionalexamplesin(17)showthatNdebele'sbisyllabicREDdoesnotrespectmorphemeintegrity(Hymanetal.
2009):(17)a.
lim-a'cultivate'!
lim-a+lim-a'.
.
.
alittlehereandthum-a'send'!
thum-a+thum-athere'b.
nambith-a'taste'!
nambi+nambith-athembuz-a'gofromwifetowife'!
thembu+thembuz-aLongrootsthereforecanbetruncated,asin(17b).
Whentheverbrootoccurswithaproductivederivationalextension,REDoccursintwoforms,asin(18).
(18)a.
lim-el-a!
lim-e+lim-el-a'cultivatefor/at'(applicative-el-)!
lim-a+lim-el-ab.
lim-is-a!
lim-i+lim-is-a'makecultivate'(causative-is-)!
lim-a+lim-is-a3Oneisremindedoftheobligatorysufxationofreciprocal-aneinDualareduplicationseenin(8a).
Reciprocal-an-isalsosometimesaddedtoconvertamonosyllabicstemintoabisyllabicword,asinthesingularafrmativeimperativeinNdebele:/m-a/'stand'!
m-an-a(Sibanda2004)withthevariantyi-ma(Downing2001,p.
36).
186L.
M.
Hyman123Eithertherstvowelofthe-VC-extensioniscopied,ortheextensionistruncatedanddefault-aappearsinstead.
However,although(17b)andtherstvariantsof(18a,b)showthatNdebeledoesnotrespectmorphemeintegrity,monomorphemicverbbasesof3+syllablesstillcannotreduplicatewith-a:*namb-a+nambith-a,*themb-a+thembuz-a.
Thegeneralizationisthatreplacementbydefault-aispossibleonlyifalloftherootmaterialhasbeenexhaustivelymappedintoRED.
Whiletherstvariantsof(18a,b)showthatthevowelofaderivationalsufxcanbecopiedintoRED,(19)showsthatREDcannotcontaininectionalmaterialfromthebase:(19)a.
lim-e!
lim-a+lim-e*lim-e+lim-e(subjunctive-e)b.
lim-i!
lim-a+lim-i*lim-i+lim-i(negative-i)c.
lim-ile!
lim-a+lim-ile*lim-i+lim-ile(perfective-ile)Asobserved,default-aisrequiredtolloutthebisyllabicREDinsuchcases.
In(20)IsummarizethepropertiesofNdebeleREDintermsoftheBantuverbstemaccordingtoDowning(1999,2003):20mets-IExtendedDerivational-stemInectionalnalsuffix(IFS)MinimalD-stem[=root]Extensionse.
g.
lim--el-,-is--e,-i,-ile,-aMustcopyMaycopyCannotcopyWhattheseexamplesshowisthatthereareTWOscalesforparingdownthereduplicant,startingwiththefull(inected)verbstem,asin(21).
(21)a.
Phonologicalscale:full>foot(r–r)>syllable(>mora>tone>)b.
Morphologicalscale:I-stem>ExtendedD-stem>rootConcerningthemorphologicalscale,Ndebelenicelycapturesthethree-waydis-tinction:AllrootmaterialforwhichthereisroomisobligatoryinRED,derivationalmaterialisoptional,andinectionalmaterialisprohibited.
AlogicalextensionofthisistorestrictreduplicationtoCVC-roots,asisnearlythecaseinKinyarwanda.
SinceKimenyi's(2002)studymostlyconcernslexicalizedverb-stemreduplication,in(22)IhavereplacedhisexampleswithrelativelyproductiveonesprovidedbyFide`leMpiranyi(personalcommunication)toshowthatnon-syllabicextensionssuchaspassive/-u-/-w-andcausative/-i-/-y-donotoccurinREDevenifthereisroomforthem:44In(22b),r!
zbeforecausative-i-.
Mpiranyapointsoutthattherearelexicalizedexceptionswherethecausativeiscopied:rwaar-a'besick'!
/rwaar-i-a/rwaaz-a'takecareofasickperson'!
rwaaz-a+rwaaz-a'takecareofasickpersonorsituationwithalotofeffort'.
Thenaturalhistoryofverb-stemreduplicationinBantu187123(22)a.
rim-w-a'becultivated'!
rim-aa+rim-w-a'becultivated(*rim-w-aa+rim-w-a)severaltimes'b.
kwiiz-a'spread(tr.
)'!
kwir-a+kwiiz-a'spread(tr.
)/kwiir-i-a/(*kwiiz-a+kwiiz-a)allover'cf.
kwiir-a'spread(intr.
)!
kwiir-a+kwiir-a'spread(intr.
)allover'Itdoes,however,seematleastmarginallyacceptableforaproductive-VC-extensiontoappearwithaCV-rootinRED,e.
g.
gw-iir-a'fallfor/at'!
gw-iir-a+gw-iir-a,gu-ush-a'causetofall'!
gu-ush-a+gu-ush-a.
WhileKinyar-wandadoesnotallowreduplicationofunanalyzablestemsofthreeormoresyllables,orthosewhosemorphologyisfrozenorunproductive,KikuyuhasimposedDowning's(1999)canonicalCVC-astemindeanceofmorphemeintegrity(Peng1991;Mugane1997,p.
12):(23)a.
kor-a'grow'!
kor-a+kor-acin-a'burn'!
cin-a+cin-ab.
koor-a'pullout'!
koor-a+koor-abuut-a'depose'!
buut-a+buut-ac.
bocor-a'beindented'!
boc-a+bocor-ahccrer-a'bequiet'!
hccr-a+hccrer-ad.
ciaerer-a'encircle'!
cia-a+ciaerer-ahwererek-a'tilt'!
hwer-a+hwererek-aNeitherKinyarwandanorKikuyuallowinectionalendingsinRED.
Concerningthephonologicalscalein(21a),wehaveyettoillustratethereductiontoasyllable.
Asseenin(24),Lengolaexpressesthehabitualbymeansofverb-stemreduplication(Stappers1971,p.
268):(24)a.
i-kul-a'acheter'i-kul-a+kul-a(CVC-a!
CVC-areduplication)i-"on-a'regarder'i-"on-a+"on-ai-tum-a'envoyer'i-tum-a+tum-ai-lmb-a'chanter'i-lmb-a+limb-ab.
i-"-a'manger'i-"-^a+"i-a(CV-a!
Ca-reduplication)i-b^i-a'parler'i-b-^a+bi-ai-ki-a'faire'i-k-a+ki-ac.
i-kpet-a'couper'i-kp-a+kpet-a(CVC-a!
C-a-reduplication)i-gbok-a'trouver'i-gb-a+gbok-ai-/am-a'crier'i-/-^a+/am-a188L.
M.
Hyman123Asseen,the-CVC-verbrootsin(24a)fullyreduplicatealongwiththeFV-a.
In(24b),wheretheroothastheshape-Ci-,thevowelistruncatedinRED.
Theformsin(24c)showthattheREDofsome-CVC-rootshasalsobeentruncatedtoC-a-.
AsimilarstorycomesfromKanyok,whichcontrastsstem-andCV-reduplica-tion(Stappers1986a,p.
17;Mukash-Kalel1982,pp.
151–2,personalcommuni-cation):(25)a.
frequentative:'toutletempsetdemanie`redesordonnee''dim'cultiver'!
dim+dimow'selaver'!
ow+owtum'envoyer'!
tum+tumand'creuser'!
and+andb.
imperfectiveaspect(progressive,durative)dim'cultiver'!
dii+dimow'selaver'!
ow+owtum'envoyer'!
tuu+tumand'creuser'!
and+andSinceKanyokhaslostmostnalvowels,thefrequentativeformsdim-dimandtum-tumin(25a)areequivalenttodim-a+dim-aandtum-a+tum-ainotherBantulanguages.
In(25b),weseethatCVCrootshaveaCVV-REDintheimperfectiveaspect.
BothLengolaandKanyokthusexploitstemreduplicationformarkingas-pect.
WhilethetwoREDpatternsmayatrstseemunrelated,thereis,inturn,anobvioussemanticrelationbetweenfrequentative'allthetime,hereandthere'andimperfectiveaspect.
Ihypothesize,therefore,thatbothconstructionsin(25)havethesamesource.
Thisisconrmedinthecaseof-VC-roots.
Asseenintherightcolumnof(25a,b),thefrequentativeandimperfectivehavethesamereduplicatedroot—andthereisnovowellengtheningintheimperfective(i.
e.
*oow-ow).
Wemighthypothesizethatthefrequentativeoriginallyimposedatwo-syllableminimumonRED.
Thisconstraintwassubsequentlyrelaxedinthesemanticsplitthatgaverisetothereduplicatedimperfectiveaspect(whichisclearlyaninnovation).
Furtherevidenceisseenfromthewaysinwhichsubminimal-CV-rootsreduplicatein(26).
(26)a.
/tu/tw'piler'!
tw-aa-tw-aa-tw(frequentative)b.
!
tw-aa-tw(frequentativeorimperfective)Asseen,thehistoricalFV-aappears(lengthened)inRED,whichinthefrequentativecanoptionallyproduceatriplicatedstructure.
Boththetotalreduplicationof-VC-rootsandthepresenceof-aa-suggestthattheREDofimperfectiveCVV-redu-plicationwasoriginally*CVC-a.
Thechangeof*CVC-atoCV-reduplicationisalsoseeninBomanominalizedhabituals,whichStappers(1986b,p.
40)describesashavinganimpliedpejorativesense.
Thus,whenhabitual/repetitiveverbformssuchasin(27a)arenominalized,REDdevelopsintoaconsonant+highvowel,asin(27b).
(27)a.
la-a!
la-a-la-a'usually/alwaysgoonwalks'b.
sa-a!
i-s-sa-a'customarydoing'za:-e!
i-zi-za:-e'customaryeating'c^um-a!
i-cu-c^um-a'customarybuying'Thenaturalhistoryofverb-stemreduplicationinBantu189123kw^a=-a!
i-ku-kw^a=-a'customaryloving'kcb-a!
i-ku-kcb-a'customaryweaving'WhileattestedonlyinNorthwestBantu,Ci-/Cu-reduplicationisofcoursewell-knownfromWestAfrica.
5Furtherreductionofasingle-syllableREDisseenin(28)fromBaa,anotherNorthwestBantulanguage(ArogaBessongandMelcuk1983,p.
500):(28)innitive3sg+durativereduplicationRED!
,withVlengthri--ka`n'ecrire'a-ka`N-ka`n^a:-Nka`nri--so`o'laver'a-so`o-so`o^a:-so`ori--kal-'dire'a-ka-kal-^a:-kal-ri--la`n-'sefendre'a-la`-la`n-^a:-la`n-Asseen,durativereduplicationismarkedbyamonosyllabicRED.
Asanalterna-tive,REDmayundergosegmentaldeletion,whichArogaBessongandMelcukterm''contraction''.
Inthelastcolumnof(28)weseethatthesubjectacquiresthelowtoneoftheREDtobecomeahigh-lowfallingtoneonalongvowel.
Ifthiscon-tractionreplacedtheearlierconstruction,therewouldbenosynchronicevidencethatthedurativehadoriginallyinvolvedreduplicationratherthanthelossofanaspectual(C)Vprexwithlowtone.
Thelaststepsinthephonologicalevolutionofreduplicationare,thus,reanalysisandloss.
5HypothesisII:partial[fullverb-stemreduplicationIntheprevioussectionwesawthatsomeBantulanguageshavefull-stemverbreduplication,whileothersreduplicatelessthanthefullstem.
Insomecasesthetruncationsor''contraction''areclearinnovationsandlookratherrecent.
ThesimplesthypothesisisthatPBreduplicatedthefullverbstem,whichwaslaterpareddowntomorphologicallyandphonologicallysimplerstructures.
HypothesisIthusclaimsaunidirectionalityfromabiggertoasmallerRED.
Thefullstoryis,however,abitmorecomplex.
Therearecounter-tendencieswhichcanhavetheeffectofENLARGINGthereduplicant,i.
e.
ingoingfromasmallertoabiggerRED(cf.
HurchandMattes2005).
Ifcorrect,thiswouldmeanthattherehasbeenbidirectionalchange,whichnaturallycouldcomplicatethetaskofdeterminingwhatthePBsituationinfactwas.
5Stappersdoesn'tgiveenoughexamplestobecertain,butexceptfori-ku-k^a=-a'customaryplanting',theCu-REDoccurswhentherooteitherhasaroundvowelorbeginswithCw.
Ontheotherhand,rootswithaCi-REDhavebothaninitialcoronalconsonantandthevowel/a/or/e/.
190L.
M.
Hyman123AsIshallnowdocument,thesecounter-tendenciestakethefollowingshapes:(29)a.
REDmayincludeafxes(whichareotherwisebarred)tomakeREDbisyllabicb.
REDmayincludeafxes(whichareotherwisebarred)becausetheysyllabifywiththerootc.
REDmayincludeafxes(whichareotherwisebarred)whenbase-reduplicantfeaturalnon-identitywouldotherwiseresultAsweshallsee,theresultisthatREDmaybeenlargedinoneoftwoways:First,morestemmaterial,specicallysufxes,maybecomeincorporatedintoRED.
Sec-ond,materialoutsidethestem,specicallyprexes,maybecomeincorporatedintoRED.
Whereasthersthastodowithhowmuchoftheverbstemiscopiedinreduplication,thesecondhastodowiththescopeofreduplicationandwhetheritcan''see''prexalmateriallyingtotheleftofthestem.
Inthefollowingsubsec-tionswewillfocusrstontheissueofprex-incorporationintoREDandthenconsidercaseswhereinectionalsufxesareexceptionallycopied.
5.
1Prex-incorporationTherstsituationtobeconsiderediswhenaprexisexceptionallycopiedwhenitisneededtomakeREDbisyllabic.
Thishappensonlyinthecaseofsub-minimalverbroots,whichareeither/-CV-/or/-C-/,dependingonthelanguage.
In(30)weseehowthesubminimalroot-dl-'eat'isreduplicatedinNdebele:(30)a.
(uku-)dl-a'toeat'!
(u-ku-)dla-yi+dlab.
(uku-)zi-dl-a'toeatthem'!
(u-ku-zi-)dla-yi+dlac.
(uku-)zi-dl-a'toeatthem'!
(u-ku-)zi-dla+zi-dlad.
(uku-)zi-bon-a'toseethem'!
(u-ku-zi-)bon-a+bon-a*(u-ku-)zi-bo(n-a)+zi-bon-aIn(30a),theverbstemdl-a'eat'ismonosyllabic.
RecallthatNdebelerequiresabisyllabicRED.
Inordertolloutthetemplate,adummysyllable-yiisaddedtothemonosyllabicstem.
Sincetheclass15augment+prexsequence/u-ku-/isnotavailable,(30a)representstheonlywaythatsubminimaldl-acanberedupli-cated.
Thesamedla-yiREDisobservedin(30b),wheretheclass10objectprexzi-'them'hasbeenadded.
However,(30c)showsthatzi-mayalternativelybeitselfreduplicated.
Whatthismeansisthatwhenanothersyllableisrequired,Ndebelespeakerscan''goup''tothemacro-stemlevel,theconstituentwhichconsistsoftheobjectprex+stem.
Asseenin(30d),thisstrategyisnotavailableiftherootisanylonger,i.
e.
ifitis-CVC-.
Ifzi-isincluded,theresultingREDwillviolateoneoftwootherwiseinviolableconstraints:(i)*zi-bon-a+zi-bon-ahasatrisyllabicRED;(ii)*zi-bo+zi-bon-ahasaREDwhichfailstoparseasmuchoftherootaspossible,specicallythe/n/of/-bon-/'see'.
Aswesawwithrespecttosufxalmaterial,anobjectprexcanappearinREDonlyiftherootmaterialisexhausted.
Theveequallyacceptableoutputsin(31),allThenaturalhistoryofverb-stemreduplicationinBantu191123ofwhichderivefromu-ku-zi-dl-el-a'toeatthemfor/at',showthatanobjectprexmayappearinREDevenifanextensionoccurswiththeconsonantalroot:(31)a.
(u-ku-zi-)dl-el-a!
(u-ku-zi-)dl-el-a+dl-el-aeat-APPL-FV!
(u-ku-zi-)dl-a-yi+dl-el-a!
(u-ku-zi-)dl-e-yi+dl-el-ab.
(u-ku-)zi-dl-el-a!
(u-ku-)zi-dl-a+zi-dl-el-aOBJ-EAT-APPL-FV!
(u-ku-)zi-dl-e+zi-dl-el-aIn(31a),wherereduplicationoccursatthestemlevel,therearethreedifferentpossibilities.
First,thefullstemdl-el-a'eatfor/at'canbecopied.
Second,theapplicativeextension-el-canbetruncated,inwhichthereduplicantconsistsoftheroot+defaultFV-a(dl-a)plusthedummysyllable-yi.
Thisrealizationcorre-spondstolim-a+lim-el-ain(18a).
Thethirdstem-levelreduplicationconsistsoftherootdl-plusthevowel[e]ofapplicative-el-followedagainbythedummy-yi.
Thisrealizationcorrespondstolim-e+lim-el-ain(18a).
Turningto(31b),hereweseetwoalternativeswherereduplicationhasgoneuptothemacro-stemlevel.
Inbothcasestheobjectprexzi-iscopied:Intherstrealization,applicative-el-istruncatedandthedefaultFV-aappears.
Inthesecondrealization,the[e]ofapplicative-el-isparsed.
TheNdebelefactsshowhowextra-stemmaterial,namelytheobjectprex,canbecopiedinREDincasetherootmaterialhasbeenexhausted.
Crucially,thereisnorequirementtoexhauststem-levelmaterialbeforemovinguptothemacro-stem.
Wethereforeneedtoextendthethree-waydistinctionmadein(20)asfollows:(i)rootmaterialmustcopy;(ii)anextensionORobjectprexmaycopy;(iii)inectionalendingsmaynotcopy.
InNdebele,theonlypre-stemmaterialwhichcanbecopiedinREDistheobjectprex,andonlyifthereisasecondsyllableslotavailableforit.
InKiheheanyprexmayappearinREDwhichsyllabieswiththeroot(OddenandOdden1985;Odden2001).
Asseenin(32a),REDgenerallyexcludesprexes:(32)a.
objectprex:ku--gul-a!
(ku--)gul-a+gul-a'tobuyabitofthem'b.
innitiveku-:ku-tov-a!
(ku-)tov-a+tov-a'tobeatabit'c.
subjectprex:tu-gul-iite!
(tu-)gul-iite+gul-iit-e'weshoppedabit'AlthoughKihehehasfull-stemreduplication,theaboveformswouldbeungram-maticalwithprexcopying,e.
g.
*(ku-)-gul-a+-gul-a.
In(33),ontheotherhand,weseethatprexeswhichsyllabifywiththebasearecopied:(33)a.
object:ku-mw-iimb-il-a!
(ku-)mw-iimbil-a+mw-iimb-l-a'tosingabittohim'b.
innitive:kw-imb-a!
kw-imb-a+kw-imba'tosingabit'c.
subj+obj:n-gw-itite!
n-gw-itite+n-gw-itite'Ipoureditabit'192L.
M.
Hyman123Theseexamplesallinvolveavowel-initialrootbeforewhichtheclass1objectprex/mu-/isrealizedmw-in(33a)andtheinnitiveprex/ku-/!
kw-in(33b).
Asaresultofthisfusion,pre-stemmaterialiscopiedinRED.
Thisismoststrikingin(33c),whereboththerstpersonsubjectprexn-andtheclass3objectprex/gu-/(!
gw-)arecopied,presumablyreectingthattheinitialsyllableis[bNWi].
Thefactthatboththeinnitiveprexandsubjectprexescanbecopiedshowsthatthemacro-stemisirrelevantinKihehe.
ThesamepointcanbemadefromSwati,whereZiervogel(1952,p.
81)reportsthecopyingofasubjectprexwhichfuseswitha-VC-root,e.
g.
/"a-ev-a/'theyhear'!
"-ev-a!
"-ev-a+"-ev-a.
ThecasesjustexamineddemonstratetwodifferentphonologicalmotivationsforincludingprexesinRED:Aprexmayeitherprovidethesecondsyllableinthecaseofsubminimalroot,oritmaybecopiedbyvirtueofbeingsyllabiedwiththeroot.
ThesignicanceofNdebeleandKiheheisthattheyprovideinnovativemodelsbywhichverb-stemreduplicationcouldbecomemacro-stem-orevenfull-wordreduplicationbyanalogy.
Inotherwords,thereisawaytoenlargethedomainofreduplication.
5.
2IncorporationofInectionalsufxesInSect.
4wesawthatthereisatendencyforfull-stemreduplicationtoundergomorphologicalandphonologicalreduction.
Mostofthecaseswhereverb-stemreduplicationislessthantotalfallintotwoclasses:(i)REDisidenticaltoaPHONO-LOGICALconstituentattheleftedgeofthebase,e.
g.
therstsyllableorbisyllabicfoot.
(ii)REDisidenticaltoaMORPHOLOGICALconstituentattheleftedgeofthebase,e.
g.
theroot,withpossiblellers(default-a,dummy-yietc.
).
Thereare,however,caseswherehereneitheroftheseconditionsismet,withmorphophonemicprocessescausingtheREDandbasetobecomephoneticallydissimilar.
Whenthishappens,acounter-tendencytobuildtowardsfull-stemreduplicationsometimesshowsup.
Asarstillustration,letusreturntoNdebele,which,itwillberecalled,disallowsinectionalmaterialtobecopiedintoRED.
Ofconcernherearethecomplicationsconcerningtherealizationoftheperfective(Hymanetal.
2009).
Asseenin(34a),whenthebaseisa-CVC-root,theperfectiveisformedbysimplesufxationof-il-e:(34)a.
ba-gan-il-e'theybecamebetrothed'ba-dal-ile'theycreated'b.
ba-hamb-el-an-il-e!
ba-hambel-ain-e!
ba-hamb-el-en-e'theyagreed'ba-beth-an-il-e!
ba-beth-ain-e!
ba-beth-en-e'theyclashed'c.
ba-thath-il-e!
ba-thaith-e!
ba-theth-e'theytook'ba-sal-il-e!
ba-sail-e!
ba-sel-e'theyremained'Thenaturalhistoryofverb-stemreduplicationinBantu193123However,whenthebaseislonger,manyBantulanguagesshowfusionor''imbrication''ofperfective-il-e.
Intheexamplesin(34b),the[i]of-il-eisimbricatedbeforethelastconsonant(herethe/n/ofthereciprocalextension-an-),andthe[l]isdeleted.
Theresulting[ai]sequenceisattestedinotherBantulan-guages,butmonophthongizesto[e]inNdebele.
Thederivationsin(34c)showthatsome-CVC-rootswhichwerehistoricallybimoraicexceptionallyundergoimbri-cationaswell.
Sincethis[e]fusesinectional[i]withnon-inectional[a],thequestionishowitisviewedinthereduplicationprocess.
Asseenin(35a),ifthefused[e]appearsinthethird(orlater)syllableoftheimbricatedstem,therewillbenocomplications:(35)a.
ba-hamb-el-en-e!
(ba-)hamb-e+hamb-el-en-e'theyagreedabit/ba-hamb-el-an-il-e/~hamb-a+hamb-el-en-ehereandthere'b.
ba-beth-en-e!
(ba-)beth-a+beth-en-e'theyclashedabit/ba-beth-an-il-e/~beth-e+beth-en-ehereandtherec.
ba-theth-e!
(ba-)thath-a+theth-e'theytookabithere/ba-thath-il-e/~theth-a+theth-eandthere'*(ba-)theth-e+theth-ed.
*ba-thath-il-e(ba-)thath-a+thath-il-eAswesawin(18),therearetwovariants:therstcopiesthe[e]ofapplicative-el-,whilethesecondtruncates-el-andusesthedefaultFV-a.
Thetwovariantsarethereforeasexpected.
Nowconsiderthetwovariantsin(35b).
WhiletherstREDhastruncatedallsufxalmaterialtoproducebeth-a,thesecondRED,beth-e,hascopiedthersttwosyllablesofthebase.
Recallthatthis-eresultsfromthefusionofthe[a]ofthereciprocalextension-an-withthe[i]ofperfective-il-e.
Inotherwords,thefrontfeatureof[i]hasbeencopiedeventhoughNdebelenormallyprohibitsinectionalmaterialfromRED.
Thesameoptionsareseenin(35c),wherethefrontfeaturecanberealizedontherootsyllableofRED.
TheinterpretationwhichHymanetal.
(2009)givetodoubletssuchasin(35b,c)isthatNdebelespeakersvaryinwhethertheytreatthefused[e]asinectionalornon-inectional:Whilethefrontfeaturederivesfromunderlying/-il-/,thevowelslotonwhichitisrealizedbelongseithertoaderivationalsufxortotheroot.
Thethirdstarredformin(35c)showsthatitisnotpossiblefortheFV-etoappearinREDsinceitisunambiguouslyinectional.
Notein(35d)thatalthough/thath-il-e/obligatorilyundergoesimbricationinthesimplexform,imbricationcanbeoptionallyblockedinthecorrespondingreduplication(Sibanda2004).
Inthiscaseinectional-il-eisnotcopied,andthath-isrealizedidenticallyinREDandinthebase.
Thesignicanceofthesefactsistwofold.
First,formssuchasbeth-a+beth-en-eandespeciallythath-a+theth-eagainprovideawayinwhichasmallerdomain(theverbbaseorderivationalstem)mayreachouttoincludematerialfromalargerone(theinectionalstem),whichisotherwiseprohibited.
ThisisalsoseenincloselyrelatedSwati,which,likeNdebele,normallydisallows-il-einRED,butmaycopy-il-whenitfollowsa-VC-root(Ziervogel1952,p.
81):194L.
M.
Hyman123(36)a.
en-'beovergrown'+PERF!
en-il-e!
en-il-a+n-il-eb.
om-'becomedry'+PERF!
om-il-e!
om-il-a+m-il-eTheexpectedformsareen-a+yen-il-eandom-a+yom-il-e,andindeedZiervogelmentionsthelatterasavariant.
(-VC-rootsoftenalternatewith-yVC-inBantu.
)In(36)theinitialrootvowelisexcludedfromtheprocess,leavingn-il-eandm-il-eastheinputstoreduplication.
Since-n-and-m-aresubminimal,-il-isincorporatedintothebaseasifitwereaderivationalextensionratherthanpartoftheinectionending-il-e.
However,inectional-emaynotbecopiedintoRED(*en-il-e+n-ile,*om-il-e+m-il-e).
Thesecondsignicanceofimbricationisthatoutputslikethath-a+theth-eandsal-a+sel-e(from/sal-il-e/!
sel-ein(34c))nowproduceformsinwhichthereduplicantsandbasesstarttodriftapartfromeachother.
SuchformsaretheonlyoneswheretherstvowelofREDisdifferentfromtherootvowel.
Tomakethetwopartsmoresimilar,thealternativeformstheth-a+theth-eandsel-a+sel-ebe-comeacceptable.
Thelogicalendpointofsuchastrategywouldbetomakethereduplicantandthebaseexactlyidentical,i.
e.
tohavetotalcopy.
While*theth-e+theth-eand*sel-e+sel-earenotacceptableinNdebele,otherBantulan-guagesseemtohavetakenthestepofre-introducinginectionalendingsintoRED.
OnesuchcaseconcernsLuvale(Lwena)aboutwhichHorton(1949,p.
101)writes:''TheFrequentativeformindicatesthattheactionorstateisrepeatedanumberoftimesorappliestoanumberofsubjects.
.
.
.
Thisderivativeisformedbyreduplicationofthestem.
Originallythesereduplicatedstemsweretreatedassingleverbs,withsingleinectionalsufxes.
Today,inthecaseofdisyllabicstemsintheperfect,bothpartsofthereduplicatedformundergomutation.
Whatthismeansisthattheearlierstructurein(37a)involvingrootreduplicationisbeingreplacedbyfullstemreduplicationin(37b).
(37)a.
[[ROOT-a-ROOT]INFL]b.
[[ROOT-INFL][ROOT-INFL]]Intheperfect,Luvaleusesapatterncommonincentral-westernBantu(Gregoire1979),wherebytheFVis-eafter-CaC-roots,asin(38a),otherwiseacopyoftherootvowel,asin(38b).
(38)a.
tu-na-tal-e'wehavelookedfor'tu-na-mbat-e'wehavecarried'b.
tu-na-het-e'wehavearrived'tu-na-mon-o'wehaveseen'tu-na-hik-i'wehavestirred'tu-na-tumb-u'wehaveplanted'c.
tu-na-lis-a'wehavecausedtoeat'(/li-is-/'eat-CAUS')tu-na-ci-many-is-a'wehavenishedit'(/man-is-/'nish-CAUS')tu-na-hambakan-a'wehavepassedby'Thenaturalhistoryofverb-stemreduplicationinBantu195123Asseenin(38c),derivedandlongerverbstemstaketheFV-a.
Hortonindicatesthatreduplicationswereoriginallytreatedasonelongbase,thustaking-aintheperfect,asin(39a).
(39)a.
(va-na-ci-)tal-a+tal-a[[tal-a-tal]-a]'theyhavelookedforit'b.
(va-na-ci-)tal-e+tal-e[[tal-e][tal-e]]However,headdsthatspeakers''nowfrequently''useinnovativeformssuchasin(39b),wheretheFV-eappearsalsoinRED.
Thatthemotivationistoderivetotalidentitybetweenthetwopartsofthereduplicatedverbstemisconrmedintheremotepasttensein(40).
(40)a.
(va-mu-)vet-a+vet-el-e[[vet-a-vet]-el-e]'theyeachofthemstruckhimb.
(va-mu-)vet-el-e+vet-el-e[[vet-el-e][vet-el-e]]againandagain'Asindicated,theoldervariantin(40a)showsareduplicatedROOT-a-ROOTfollowedbyasingleinection.
However,Hortonpointsoutthatspeakers''nowoften''produceformssuchasin(40b),wheretheperfectiveending-el-eoccursinRED.
Thispointstotherstoftwoconceptualconicts,thatbetweenthetwodesirableanalysesin(41),whereBASEisusedintheBantusenseofROOT+possiblederivationalextensions:(41)a.
[[BASE]i+[BASE]i]+INFL]reduplicatedbase+singleinectionb.
[BASE+INFL]i[BASE+INFL]ireduplicatedcompoundstemwithinectionAsshownin(41a),speakerswouldliketoreduplicatethebasewithasingleinection.
ThiscanresultinatruncatedRED,asinvet-a+vet-el-e.
Ontheotherhand,theywouldlikeareduplicatedcompoundstemwherethetwopartsareidentical,asinvet-el-e+vet-el-e.
Sincebothstructuresaremotivated,thediachronicimplicationisthateithershouldbeabletochangeintotheother.
AsecondconceptualconictconcernsthequestionofwhetherthereduplicatedverbstemconsistsofoneBASEortwoInMambwe,theperfectiveisrealized-il-eafterCVC-roots(whosenalconsonantmayundergomutation),asin(42a),butisimbricatedafterlongerroots,asin(42b)(Halemba1994):(42)a.
land-+il-e!
lanz-il-e'talk'sent-+il-e!
sens-il-e'gnaw'lil-+il-e!
liz-il-e'cry'lot-+il-e!
los-il-e'dream'fum-+il-e!
fum-il-e'goout'b.
pongan-+il-e!
pongin-e'beunrecognizable'pelem-+il-e!
pelim-e'disobey'vwitik-+il-e!
vwitik-e'bewitch'196L.
M.
Hyman123folol-+il-e!
folwil-e'scratch'simul-+il-e!
simwil-e'run'Ofrelevancehereishowtheperfectiveisrealizedonreduplicatedverbstems,i.
e.
whetherthelatteraretreatedasonelongbaseortwoshortones.
Bothpossi-bilitiesareattested.
Mostlexicalizedreduplicationswhichhavenotransparentsimplexbasearetreatedasasinglebaseandhenceundergoasingleimbricationasin(43).
(43)a.
[fuk-a-fuk]+il-e!
fuk-a-fwik-e'berestless,dgety'b.
[tang-a-tang]+il-e!
tang-a-ting-e'beembarrassed,perplexed'c.
[tuw-a-tuw]+il-e!
tuw-a-twiw-e'beslow,lazy'Ontheotherhand,mostproductiveorsemanticallytransparentreduplicationsaretreatedastwobases,onlythesecondofwhichisvisibletotheperfectivemor-phology.
Wherethesecondbaseisshort,thereisnoimbrication,asseenin(44).
(44)a.
[lap-a][lap]+il-e!
lap-a+laf-il-e'swearoftenwithoutgoodreason'(cf.
lap-a'swear')b.
[cit-a][cit]+il-e!
cit-a+cis-il-e'doasonewishes,notfollowingrules'(cf.
cit-a'dobadthings')c.
[kow-a][kow]+il-e!
kow-a+kov-il-e'fumbleeverything'(cf.
kow-a'catch,grab,clingto')Eventhoughthenalconsonantoftenundergoesmutationto[f]or[s]before-il-e,themotivationfornotimbricatingisthatthereisgreateridentitybetweenREDandthebase,whichwillhaveanidenticalinitialCV(andsometimesCVC).
Still,whileIexpectthat(44)representsthewaythatproductivelycreatedreduplicationswouldberealizedintheperfective,aswellasmostofthosewhicharelexicalizedbuttransparent,thereareafewofthelatterwhichexceptionallyundergoimbrication,asin(45).
(45)a.
[cew-a][cew]+il-e!
cew-a+ciw-e'lookallaround'(cf.
cew-an-a'look,winkate.
o.
')b.
[suk-a][suk]+il-e!
suk-a+swik-e'berestless,dgety'(cf.
suk-an-a'dget,wrigglebody')Thenaturalhistoryofverb-stemreduplicationinBantu197123Infact,thereisthefollowingminimalpairin(46).
(46)a.
[kap-a][kap]+il-e!
kap-a+kaf-il-e'blink'(cf.
kap-awil-a'blink',kap-a-iz-y-a'causetowink,blink')b.
[kap-a][kap]+il-e!
kap-a+kip-e'rustle,asdryleaves'Asseen,(46a)doesnotundergoimbrication,astherearerelatedverbswiththerootkap-.
(46b),ontheotherhand,doesnothaveanyobviouslyrelatedverb(otherthanthederivativekap-a+kap-ol-a'causetorustle')andhenceundergoesimbrication.
Wehavenotaddressedthequestionofwhichperfectivestrategyisolder.
Beforewedo,considerhowlongerreduplicatedverbstemsformtheperfectivein(47).
(47)a.
simul-a+simul-a!
simwil-e+simwil-e'runhereandthere'b.
pongan-a+pongan-a!
pongin-e+pongin-e'becompletelyunrecognizable'c.
folol-a+folol-a!
folwil-e+folwil-e'scratchallover'Asseen,bothstemsareimbricated(cf.
thecorrespondingunreduplicatedverbstemsin(42b)).
Thereasonisclear:Iftheoutputshadbeen*simul-a+simwil-e,*pongan-a+pongin-e,and*folol-a+folwil-e,therewouldhavebeenasignicantdifferencebetweenthetwopartsofthereduplication.
Asinthecaseoflap-a+laf-il-e'swearoftenwithoutgoodreason'versus*lap-a+lif-ein(44),thepreferredformistheonewherethevowelsofREDandthebaseareidentical.
Twostrategiesareemployedtoachievethiseffect:Imbricationunderappliesin(44),wheretherootsareCVC-,butoverappliesin(47),wheretheinputsareCVCVC-.
That-il-eisnotexpectedtobecopied,isseennotonlyfromtheCVC-a-CVC-il-ebasesin(44),butalsofromlongerverbbasessuchasin(48).
(48)a.
sinteek-+il-e!
sintees-il-e'cauterizeoveralargearea'b.
sukook-+il-e!
sukoos-il-e'wrigglecontinuouslyone'sbody'c.
peelook-+il-e!
peeloos-il-e'givemuchwithoutmeasure'Whatthesehaveincommonisthelengthoftheirpenultimatevowel,whichblocksimbrication.
Asseennowin(49),-il-eisnotcopiedinthecorrespondingredupli-cations:(49)a.
sinteek-a+sintees-il-e'cauterizeallover'b.
sukook-a+sukoos-il-e'wrigglecontinuouslyone'sbody'Finally,reduplicatedsubminimalrootsarenotconsistentwithrespectto-il-e:198L.
M.
Hyman123(50)a.
sy-a!
s-il-e'leave'sy-a+sy-a!
s-il-e+s-il-e'leavethingslyingabout'sy-a+sy-a+sy-a!
s-il-e+s-il-e+s-il-e'leavethingslyingallovertheplace'b.
fw-a!
fw-il-e'die'fw-a+fw-a!
fw-a+fw-il-e'dieinbignumbers'c.
pw-a!
pw-il-e'dryup'pw-a+pw-a!
pw-il-e'getdryfrequently'Asseen,-il-eiscopiedin(50a),butnotin(50b).
Curiously,Halemba(1994,p.
692)providesthesame-il-eformforpw-a+pw-aashedoesforpw-a.
Since-il-eisregularlycopiedintoREDonlywhenimbricated,itwouldseemlogicaltoconsiderthisaninnovation,asinNdebele.
Again,weseethatthedomainofreduplicationcanbeextended''upwards''toincorporatetheinectionalending.
Notmentionedthusfaristhepossibilityofasemanticconictbetweentheperfectivityof-il-eandthefrequentativemeaning.
Attheveryleast,reduplicatedperfectivesmayoccurlesscommonlythannon-perfectives.
AnindicationofthiscomesfromBemba:''Generallytheuseofthe-ILEtensesisavoided,butwhentheyareusedthetensesarerepeated.
vanSambeek1955,p.
91)WhileBembarequiresfullverbstemreduplication,twodifferenttonepatternsarereportedbySharman(1963,p.
82–83):51a.
(bá-a-)[lim-íné+lím-íné]'theyhoedbyfitsandstarts'Hb.
(bá-a-)[lim-íné]+[lim-íné]'theyhoedenormously(orscrappily,badly)'HHIn(51a)asingleHtonesufxislinkedfromthesecondtothelastvowelofthereduplicatedstem.
In(51b),however,therearetwosufxalHtoneswhicharelinkedtothesecondandnalvowelofeachpartofthereduplicatedstem.
In(51a),asinmostBantulanguages,speakerstreatthereduplicatedverbasasinglestemwhichreceivesoneintonationalpatternjustlikeotherstems.
In(51b),twoinec-tionaltonepatternsareassignedtothereduplicatedverb.
ThisisextremelyrareinBantu,theonlywell-knowncasebeingChichewa(CarletonandMyers1996,HymanandMtenje1999).
6Examplesinvolvingthehortativeareobservedin(52).
(52)a.
ti-tandiz-e!
ti-tandiz-etandiz-e'let'shelphere&there'b.
ti-vundikir-e!
ti-vundikir-evundikir-e'let'scoverhere&there'c.
ti-khululukir-e!
ti-khululukir-ekhululukir-e'let'spardonhere&there'Asseen,asingleHtoneappearsontheFVofeachstem.
HymanandMtenje(1999)presentargumentsthatthesecondstemconstitutesaseparatephonologicalword,6Subiyamayalsocopytoneinverb-stemreduplication(JoyceMathangwane,pers.
comm.
)Thenaturalhistoryofverb-stemreduplicationinBantu199123whichagainmakesChichewaverbreduplicationquitedifferentfromotherBantulanguages.
SinceChichewaisnotaparticularlyconservativeBantulanguage(e.
g.
ithaslost-il-eandthenominal''augment''andhasinnovatedcontrastivetonesonverbextensions),itislikelythatitisinnovativeinitsverb-stemreduplicationaswell.
Inotherwords,Chichewaisanextremecaseofrebuildingfullreduplicationfromanearliersystem.
Wenextturntoconsiderwhythismayhavehappened.
6Anewhypothesis:bidirectionalityIntheprecedingsectionsIhavedocumentedsomeofthevariationontherealizationofREDinverbreduplicationindifferentBantulanguages:Someimposeabisyllabiccondition,somebaninectionalmaterial,someallowprexestocreepin,someunder-oroverapplyimbricationofperfective-il-e,andsoforth.
Attheoneendisfull-stemreduplication,attheother,abisyllabic(morerarelymonosyllabic)maximum.
Itisnothardtoexplainwhysuchvariationexists,sincebothfull-andpartial-reduplicationaremotivatedbyconictingprinciples.
Consider,rst,full-stemreduplication.
Inthiscasethetwopartsofthereduplicatedverbareexactlyidentical,asexpectedofreduplication,especiallyatitsearly''iconic''stage.
Thisidentitythusavoidsthephonologicalormorphologicalmismatchproblemswhichariseinpartialreduplication.
Finally,full-stemreduplicationismoreinkeepingwiththereduplicationprocessesaffectingotherpartsofspeechinBantu:reduplicatednoun-andadjectivestemsarerarelytruncated,andnumeralstypicallyshowfull-wordreduplication,aswasseeninKanyokin(11)above.
Ontheotherhand,full-stemreduplicationhasdisadvantages.
Therstisthe''effort''problem:thestemcanbequitelong,henceawkwardtorepeatintoto.
Recallfrom(14)thatCiyaohasfull-stemreduplication.
Italsoallowsquitecomplexverbstems,astheoctosyllabicexamplein(54)shows(Ngunga2000):(53)taam-uk-ul-igw-aasy-an-il-a'causeeachothertobeunseatedfor/at'beseated-IMPOSITIVE-REVERSIVE-PASS-CAUS-RECIP-APPL-FVIfreduplicated,theresultwouldbe16syllables,whichisconsiderableoverkillwhenoneconsidersthatfromasemioticpointofview,REDrealizedonlyonesememe,e.
g.
{frequentative}.
Intermsofmarkingthefrequentativeconstruction,repeatingalloftheextensionsandaninectionalFVdoesnomorethan,say,copyingtherstCV-.
Thereisalsothesemanticissuethatverb-stemreduplicationtypicallyisconcernedwithadiminutionorintensicationofthelexicalrootmeaning('dosomethingalittlebithere&there,perhapsaimlesslyorbadly'),whichultimatelycanbecomedisassociatedfromtheoriginalmeaning,aswehaveseeninsomeoftheglosses.
Fromasemanticpointofviewderivationalsufxesareschizophrenic.
Totheextentthattheyarefrozenorcontributeunpredictablytothelexicalmeaning,itwouldmakesensetocopythemaswell.
Wheretheyhaveaproductive,grammaticalfunction,e.
g.
licensingacausativeorapplicativeargu-ment,asintheCiyaoexamplein(53),itwouldseemlessmotivatedforthemto200L.
M.
Hyman123appearinRED.
Incaseswherethesemanticsoftheextensionisspecicallytargeted,theextensionmayitselfberepeated,aswhencausative-is-isusedwithanintensivefunctioninShona(Dembetembe1978,p.
43):(54)far-a'behappy'bat-a'hold'nak-a'begood'far-s-a'betoohappybat-s-a'holdfast'nak-is-a'beverygood'far-s-s-a'beexcessivelyhappy'bat-s-s-a'holdvery'beextremelygood'rmly'Sincereduplicationoriginallytargetsthelexicalmeaningoftheroot,itmakeslesssenseforthetense/aspect/moodendingstoberepeatedinRED.
Interestingly,itisinthoselanguageswhichhavemostreducedRED(e.
g.
toasinglesyllable)thatimperfectiveorhabitualaspectualmeaningsareobserved.
Thereis,however,noreasontothinkthatcopyingoftheinectionalendings-i,-e,or-il-econtributestothesemanticdrifttowards''moregeneralmeanings''(Bybeeetal.
1994).
WhatthismeansisthatREDissubjecttoaconictbetweenIDENTITYversus.
ECONOMY,i.
e.
betweentheexpressivedemandsoffull-stemidentityvs.
thevariousphoneticshortcutsthathavebeenobserved.
InthiscontextitshouldbenotedthatmostnounshavebisyllabicstemsinBantu,whichalsoproduceatwo-syllableRED,aswasseenin(12a).
Giventhatbothfullandreducedreduplicationaremotivatedbyconictingconcerns,Iwouldlikeatthispointtoadvancethehypothesisin(55).
(55)Hypothesis:thehistoricaldevelopmentofREDgoesinbothdirectionsa.
full>partialb.
partial>full(55a)recapitulatesthegeneralassumptionthatpartialreduplicationderiveshis-toricallyfromfullreduplication(Eulenberg1971;Bybeeetal.
1994;Niepokuj1997;butcf.
HurchandMattes2005).
Inprevioussectionswehaveobservedbothfullverb-stemreduplicationaswellasvariousreducedversionsofREDpredictedbythephonologicalandmorphologicalscalesin(21).
However,wehavealsoseencaseswhereprexesandinectionalendingsbecomeincorporatedinREDinordertofulllminimalityortoenhancetheidentitybetweenREDandthecorrespondingfullstem.
Asmentionedearlier,eitheroftheseconsiderationscouldinprincipleprovidethemodelforextendingthedirectionin(55b).
Infact,thereisreasontospeculatethattheoriginalreduplicationappliedonlytotheverbroot.
7Themainargumentisthatverbreduplicationissowidespreadin7Tak(2003)arguesforanoriginalbisyllabicREDinPBonthebasisofitseconomyofeffort,itsfrequencyintheworld'slanguages,andthenaturalnessofreplacingitwithfullreduplication(whichissaidtoconstituteanimprovementinlearnability).
Whilemyemphasishasbeenonestablishingthemorphologicalconstituentsthatcouldbecopied,Takismostlyconcernedwiththebisyllabicconstraint(whichdoesnotnecessarilyruleoutderivationalorinectionalmaterialinRED).
Aswillbeseen,webothrecognizethattherearemotivationsforchangeineitherdirection:Ashorter,e.
g.
bisyllabic,REDissuperiortoalongerRED,butanexactcopyismoretransparentthanatruncatedRED.
Thenaturalhistoryofverb-stemreduplicationinBantu201123Niger-Congothatithadtopredatetheevolutionofatleastsomeoftheverbmorphology.
InSect.
2Ialludedtothecontroversysurroundingthequestionofwhetherpre-stemmorphemeswereprexes,clitics,orseparatewordsinPB.
Inaddition,thereareargumentsthatatleastsomeoftheinectionalendingshaveashallowhistory.
InHyman(1993,pp.
21–23)Ipresentevidencewhichsuggeststhat''perfective*-id-,justlikeimperfective*-ag-,wentthroughacliticstage,andinfactisonlynowbeingfullyincorporatedasaregularsufxwithintheBantuverbstem.
''8Asecondargumentisthattherearelexicalizedreduplicatedverbswhichcopyonlytherstsyllable.
AsseeninthefollowingexamplesfromKinyarwanda(Kimenyi2002,pp.
254–255,260),someofthesehavemeaningssuspiciouslysimilartotheverb-stemreduplicationsemanticssurveyedin(8),specically,'doalittlebithere&there,perhapsaimlessly(56a)orbadly(56b)':(56)a.
be-beer-a'goaroundaimlessly'(~beer-a+beer-a)jaa-jaab-a'walkaroundaimlessly'hwi-hwis-a'gossip'b.
buu-buut-a'walkbent(oldage)'(~buut-a+buut-a)de-demaang-a'stutter'se-serez-a'miscutnails'Asmentioned,thesemanticsconcernstheactionoftheverb,nottheextensionsorinection.
Ifthesereduplicationspre-datetherequirementthatallverbsendinaFV(Gregoire1979),theearlierformsmightwellhavehadabuttingconsonants.
Thetwoformsof'walkbent'mightthenhavecomefrom*buut-buutand*buut-a+buut-a,respectively.
Thesuggestion,then,isthatthePBderivedverbin(57a)rstreduplicatedasin(57b).
(57)a.
PB*dIm-Id-an-a'cultivateforeachother'cultivate-APPL-RECIP-FVb.
PB*dIm-a+dIm-Id-an-a'cultivateforeachotheralittlebithere&there'Ifcorrect,thismeansthatKikerewelim-il-an-a+lim-il-an-afrom(15)wouldhavetobeaninnovation,andthatlim-il-a+lim-il-an-amighthavetwosources:partialbuild-upfromlim-a+lim-il-an-aorpartialbuild-downfromlim-il-an-a+lim-il-an-a.
In§7weconcludebyconsideringwhyverb-stemreduplicationisrobustlyprexalinBantu.
8Asexpected,oatingHtoneprexesarenotreduplicated,e.
g.
theHassignedbytheinnitiveprexku-inChichewa.
WhatwouldbeparticularlypersuasivewouldbealanguagewhichreduplicatesarootHtone,butnotaninectionalHtonesufx.
Iamunawareofanysuchcases.
202L.
M.
Hyman1237ConclusionIntheprecedingsectionsIhavedocumentedtheconsiderablevariationfoundinverb-stemreduplicationinBantu.
WhileitseemsoverwhelminglikelythatPBhadverb-stemreduplication,determiningtheshapeofthePBreduplicantislessclear.
Bothfull-stemreduplicationandroot-onlyreduplicationweresaidtohaveadvan-tages.
Copyingthefullstem(root+sufxes)hastheadvantageofguaranteeingidentityofthetwoparts.
Ontheotherhand,full-stemreduplicationcanbecomequiteunwieldy.
WethushaveseenthatREDcanbesubjecttorestrictionsthatareeithermorphological(e.
g.
don'tcopyinectionalendings)orphonological(e.
g.
maximumoftwosyllables).
Phonologicalinnovationsmay,however,alsomotivatetheincorporationofinectionalprexesandendings,therebybuildingupREDonbothperipheries.
Inallcaseswenotetheprimacyoftherootinreduplication:AswasschematizedforNdebelein(20),non-rootmaterialcanappearinREDonlyifalloftherootmaterialhasbeenexhausted.
Thispresumablyhastodowiththefunctionoftheconstruction,whichistocommentonthequantity,quality,orextentinvolvedincarryingoutthelexicalrootsemantics.
Thesupremacyoftherootmayalsoexplainwhyverb-stemreduplicationisprexal.
Whenevertruncated,REDisclearlypreposedtothefullstem.
Hymanetal.
(2009)andDowning(2004)reportaspeakerofBukusu,whoomittedproductiveextensionsinpreposedRED,asin(58a),butvariablytruncatedtheroot-initialsyl-lableofpostposedRED,asin(58b).
(58)a.
lim-il-a'cultivatefor/at'!
lim-a+lim-il-arem-er-a'cutfor/at'!
rem-a+rem-er-ab.
kacul-a'chat,talk'!
kacul-a+cul-amulix-a'ash'!
mulix-a+lix-aOfthe334quadrisyllabicorlongerverbstemsinKhisaetal.
(2000),Ifound46lexicalizedreduplicationsofwhichthefollowingthreecontainbasesofthreesyl-lables:(59)a.
sob-a+soban-a'beofunevenlengthorsize'b.
kali+kalikan-a'complicateoneself,dostubbornly'(cf.
kalikan-a'complicate')c.
mulix-a+mulix-a'twinkle'(cf.
mulix-a'ash(oflightning),toower')WhilethersttwoshowapreposedRED,thethirdexhibitstotalreduplication(conictingwiththeoutputin(58b)).
Nolexicalizedverbswerefoundoftheshapeofkacul-a+cul-a,suggestingthatthismaybeaninnovation.
ExceptforthisBukusucase,sufxalverb-stemreduplicationappearstobeavoidedinBantu.
Thereasonforthisisnothardtounderstand.
Thenon-occurringformsin(60)areintendedtobereminiscentofthebisyllabicREDinNdebele,butwhichisinsteadpostposed:Thenaturalhistoryofverb-stemreduplicationinBantu203123(60)a.
lim-el-a!
lim-el-a+lim-e'cultivatefor/at'!
lim-el-a+lim-alim-is-a!
lim-is-a+lim-i'makecultivate'!
lim-is-a+lim-alim-is-el-a!
lim-is-el-a+lim-i'makecultivatefor/at'!
lim-is-el-a+lim-ab.
lim-el-a!
lim-el-a+m-el-a'cultivatefor/at'lim-is-a!
lim-is-a+m-is-a'makecultivate'lim-is-el-a!
lim-is-el-a+s-el-a'makecultivatefor/at'In(60a)thestemmaterialismappedleft-to-rightontothepostposedbisyllabicRED.
Asseen,theresultisan''entrapment''ofproductivederivationalsufxes(appli-cative-el-andcausative-is-)insidetheverb.
Aworseoutcomeisseenin(60b),wherethesegmentalmaterialofthebasestemismappedrighttoleft.
Asseen,thereisnotonlyinternalentrapmentof-el-and-is-,butalsofailureofrootmaterialtobecopiedintoRED.
Thersttwoexampleshaveonlythenal/m/of/lim-/'cultivate',whilethethirdexamplehasnorootmaterialinRED.
WhileNelson(2002)suggeststhatthereisaprexingpreferenceforREDingeneral,Hymanetal.
(2009)arguethattheunmarkedpositonforREDis''oppositeedgeafxation'',asin(61).
(61)a.
Thereduplicantwilltendtobepreposedwhenthebasehasasufxingstructureb.
ThereduplicantwilltendtobepostposedwhenthebasehasaprexingstructureSincetheverbstemhasasufxingstructure,REDispreposed.
Sincenounshaveaprexalstructure,REDispostposedinword-levelnominalreduplication(recalltheKinandeexamplesseenin(12a)).
BydevelopingatruncatedREDontheoppositesideofafxation,theresultisthatafxesdonotgetentrappedasin(60).
Thenalquestionconcernstheroleofinectioninverb-stemreduplication.
Wehaveseenseveralexampleswheretherehasbeenanoptionorarequirementthattheinectionalendingsotherthan-abeexcludedfromRED,insomecasesevenwhenderivationalsufxescanbecopied.
Thesameissueactuallyarisesincasesoflooserreduplicationorwordrepetition,infactinEnglish.
ThefollowingnumberofhitsonGooglewereobtainedonSeptember4,2007:(62)a.
''was/weretalkingtalkingtalking'':596hits''Iwasn'tmad,Iwasconfused.
.
.
everyonewastalking;talking;talkingatmeandIcouldn'tunderstandawordtheyweresaying.
.
.
''b.
''was/weretalktalktalking'':169hits''shewastalk-talk-talkingawayonhercellphone,holdingittoherearwithonehandandgesticulatingwildlywiththeother.
.
.
''Althoughthenumbersdiffer,withtheverbtalkitispossibletorepeattheinectedparticipletalking,asin(62a),orexpressitonceonlyonthelastofthethreeverbs,204L.
M.
Hyman123asin(62b).
IalsoqueriedGooglewithrespecttothethirdpersonsingular-s,wherethenumericaldifferencesturnedouttobemuchmoredramatic:(63)a.
''he/shetalkstalkstalks'':959hits''Hetalks;talkstalksaboutactinginsomefashion.
''b.
''he/shetalktalktalks'':9hits''Asalwaysinsuchsituations,Isaynothingandhetalk-talk-talks.
''WhilewearealongwayfromNdebeleverb-stemreduplication,thesamequestionarises:WhichisolderIfthedirectionofchangeisfromwordrepetitiontoredu-plication,asitappearstobeinEnglish,thentheinectedformisolder.
Ihave,however,triedtoarguethatbothdirectionsmakesenseinBantu,whereonecanalsoclaimthatrootreduplicationexpandedtoincludederivationalandinectionalmaterial.
Additionalresearchwillhopefullyprovidedenitiveevidenceconcerningtheoriginanddevelopmentofverb-stemreduplicationinBantu.
OpenAccessThisarticleisdistributedunderthetermsoftheCreativeCommonsAttributionNon-commercialLicensewhichpermitsanynoncommercialuse,distribution,andreproductioninanymed-ium,providedtheoriginalauthor(s)andsourcearecredited.
ReferencesArogaBessong,D.
P.
,&Mel'cuk,I.
A.
(1983).
Unmode`leformeldelaconjugaisonba(a`l'indicatif).
BulletinoftheSchoolofOrieintalandAfricanStudies,46,477–528.
Ashton,E.
O.
(1944).
Swahiligrammar(includingintonation).
London:Longman,Green&Co.
Bastin,Y.
,&SchadebergT.
(Eds).
(2003).
BantuLexicalReconstructions3.
Tervuren:RoyalMuseumforCentralAfrica.
OnlineDatabase,http://www.
metafro.
be/blrBybee,J.
,Perkins,R.
,&PagliucaW.
(1994).
Theevolutionofgrammar:Tense,aspectandmodalityinthelanguagesoftheworld.
Chicago:UniversityofChicagoPress.
Carleton,T.
,&MyersS.
(1996).
TonaltransferinChichewa.
Phonology,13,39–72.
Cole,D.
(1955).
AnintroductiontoTswanagrammar.
London:Longman,Green&Co.
Dembetembe,N.
C.
(1978).
ReduplicationinShonaanditssemanticeffect.
InE.
J.
M.
Baumbach(Ed.
),SecondAfricanlanguagescongressofUNISA(pp.
20–58).
Pretoria:UniversityofSouthAfrica.
Downing,L.
J.
(1999).
MorphologicalconstraintsonBantureduplication.
LinguisticAnalysis,29,6–46.
Downing,L.
J.
(2001).
UngeneralizableminimalityinNdebele.
StudiesinAfricanLinguistics,30,33–58.
Downing,L.
J.
(2003).
Compoundingandtonalnon-transferinBantulanguages.
Phonology,20,1–42.
Downing,L.
J.
(2004).
Bukusureduplication.
InC.
Githiora,H.
Littleeld,&V.
Manfredi(Eds.
),TrendsinAfricanlinguistics5(pp.
73–84).
Lawrenceville,NJ:AfricaWorldPress.
Eulenberg,J.
B.
(1971).
ConjunctionreductionandreduplicationinAfricanlanguages.
InC.
-W.
Kim&H.
Stahlke(Eds.
),PapersinAfricanlinguistics(pp.
71–80).
Edmonton:LinguisticResearch,Inc.
Fisch,M.
(1998).
Thimbukushugrammar.
Windoek,Namibia:OutofAfricaPublishers.
Fortune,G.
(1982).
Shonagrammaticalconstructions,PartII.
Harare:MercuryPress.
Gregoire,C.
(1979).
Lesvoyellesnalesalternantesdanslaconjugaisonafrmativedeslanguesbantouescentrals.
JournalofAfricanLanguages&Linguistics,1,141–172.
Guthrie,M.
(1939/1966).
Grammaireetdictionnairedelingala.
Westmead:GreggPressLtd.
Halemba,Fr.
A.
(1994).
Mambwe-Englishdictionary.
Ndola,Zambia:MissionPress.
Horton,A.
E.
(1949).
AgrammarofLuvale.
Johannisberg:WitwatersrandUniversityPress.
Hurch,B.
,&MattesV.
(2005).
U¨berdieEntstehungvonpartiellerReduplikation.
InG.
Fenk-Oczlon&C.
Winkler(Eds.
),SpracheundNatu¨rlichkeit.
Gedenkbandfu¨rWilliMayerthaler(pp.
137–156).
Tu¨bingen:Narr.
Hyman,L.
M.
(1993).
ConceptualissuesinthecomparativestudyoftheBantuverbstem.
InS.
S.
Mufwene&L.
Moshi(Eds.
),TopicsinAfricanLinguistics(pp.
3–34).
Amsterdam:Benjamins.
Thenaturalhistoryofverb-stemreduplicationinBantu205123Hyman,L.
M.
(2007).
ReconstructingtheProto-Bantuverbalunit:internalevidence.
InN.
C.
Kula&L.
Marten(Eds.
),BantuinBloomsbury:SpecialissueonBantulinguistics.
SOASWorkingPapersinLinguistics15(pp.
201–211).
London:UniversityofLondon.
Hyman,L.
M.
,Inkelas,S.
,&Sibanda,G.
(2009).
MorphosyntacticcorrespondenceinBanturedupli-cation.
InK.
Hanson&S.
Inkelas(Eds.
),Thenatureoftheword:EssaysinHonorofPaulKiparsky(pp.
273–310).
Cambridge,MA:MITPress.
Hyman,L.
M.
,&Katamba,F.
X.
(2005).
ThewordinLuganda.
InF.
K.
E.
Voeltz(Ed.
),StudiesinAfricanlinguistictypology(pp.
171–193).
Amsterdam:JohnBenjamins.
Hyman,L.
M.
,&MtenjeA.
(1999).
Prosodicmorphologyandtone:ThecaseofChichewa.
InH.
vanderHulst,R.
Kager,&W.
Zonneveld(Eds.
),Theprosody–morphologyinterface(pp.
90–133).
Cambridge:CambridgeUniversityPress.
Inkelas,S.
,&ZollC.
(2005).
Reduplication:Doublinginmorphology.
Cambridge:CambridgeUniversityPress.
Ittmann,J.
(1939/1978).
Grammaireduduala(L.
A.
Boumard,Trans.
).
Douala:Colle`geLibermann.
Khisa,W.
,WongP.
,&LoweJ.
B.
(2000).
Bukusuwordlist.
FilemakerProTMDatabaseavailablefrompresentauthor.
Kimenyi,A.
(2002).
AtonalgrammarofKinyarwanda:Anautosegmentalandmetricalanalysis.
Lewiston,N.
Y.
:E.
MellenPress.
Mateene,K.
(1969).
Essaidegrammairegenerativeettransformationnelledelalanguenyanga.
Kinshasa:PressesUniversitairesduZa¨re.
Matondo,M.
D.
(2003).
ToneandprosodicmorphologyinKisukuma.
Doctoraldissertation,UniversityofCalifornia,LosAngeles.
Meinhof,C.
(1912).
DieSprachederDualainKamerun.
Berlin:DietrichReimer.
Meinhof,C.
(1932).
IntroductiontothephonologyoftheBantulanguages(N.
J.
vanWarmelo,Trans.
,Revised).
Berlin:DietrichReimer/ErnestVohsen.
Meeussen,A.
E.
(1967).
Bantugrammaticalreconstructions.
AfricanaLinguistica,3,81–121.
AnnalesduMuseeRoyaledel'AfriqueCentrale.
SerieIN-8,SciencesHumaines.
N.
121.
Tervuren.
Mugane,J.
(1997).
AparadigmaticgrammarofGkuyu.
Stanford:C.
S.
L.
I.
Mukash-Kalel,T.
(1982).
Lekanyok:languebantoueduZaire:Phonologie,morphologie,syntagmatique.
Doctoratde3e`mecycle.
INALCO,Paris.
Mutaka,N.
,&HymanL.
M.
(1990).
SyllablesandmorphemeintegrityinKinandereduplication.
Phonology,7,73–119.
Nelson,N.
(2002).
The''prexing''preferenceinreduplication.
ProceedingsofthewestcoastconferenceonFormalLinguistics,21,320–333.
Ngunga,A.
(2000).
PhonologyandmorphologyoftheCiyaoverb.
Stanford:C.
S.
L.
I.
Niepokuj,M.
(1997).
ThedevelopmentofverbalreduplicationinIndo-European.
JournalofIndo-EuropeanStudiesMonograph,24.
Washington,D.
C.
:InsitutefortheStudyofMan.
Nurse,D.
(2007).
DidtheProto-BantuverbhaveasyntheticorananalyticstuctureInN.
C.
Kula&L.
Marten(Eds.
),BantuinBloomsbury:SpecialIssueonBantulinguistics.
SOASWorkingPapersinLinguistics15(pp.
239–256).
London:UniversityofLondon.
Odden,D.
(1996).
PatternsofreduplicationinKikerewe.
InD.
Dowtyetal.
(Eds.
),Papersinphonology.
WorkingPapersinLinguisticsNo.
48(pp.
111–148).
Columbus,OH:OhioStateUniversity.
Odden,D.
(2001).
Re-revisitingreduplicationinKihehe.
Handout,Uit,18May2001.
Odden,D.
,&OddenM.
(1985).
OrderedreduplicationinKihehe.
LinguisticInquiry,16,497–504.
Olson,H.
S.
(1964).
ThephonologyandmorphologyofRimi.
Hartford:HartfordSeminarFoundation.
Peng,L.
(1991).
RootandfootinKikuyureduplication.
Ms.
UniversityofArizona.
Poulos,G.
(1990).
AlinguisticanalysisofVenda.
Capetown:ViaAfrikaLtd.
Rekanga,J.
-P.
(1989).
Essaidegrammairegunu.
Doctoraldissertation,UniversiteLibredeBruxelles.
vanSambeek,J.
(1955).
ABembagrammar.
Capetown:Longmans,Green&Co.
Sharman,J.
C.
(1963).
Morphology,morphophonologyandmeaninginthesingle-wordverbformsinBemba.
Doctoraldissertation.
UniversityofSouthAfrica.
Sibanda,G.
(2004).
VerbalphonologyandmorphologyofNdebele.
Doctoraldissertation,UniversityofCalifornia,Berkeley.
Stappers,L.
(1971).
Esquissedelalanguelengola.
AfricanaLinguistica,5,257–307.
Stappers,L.
(1986a).
Kanyok:eineSprachskizze.
AfrikanistischeArbeitspapiere,Sondernummer.
Stappers,L.
(1986b).
Boma.
EineSprachskizze.
Hamburg:HelmutBuskeVerlag.
Steriade,D.
(1988).
ReduplicationandsyllabletransferinSanskritandelsewhere.
Phonology,5,73–155.
Tak,J.
-Y.
(2003).
VerbalreduplicationinsomeBantulanguages:Adiachronicapproach.
Eoneohag,35,261–292.
Ziervogel,D.
(1952).
AgrammarofSwazi.
Johannesberg:WitwatersrandUniversityPress.
206L.
M.
Hyman123