LINGUISTIC AREAS AND LANGUAGE HISTORY

Author(s): SARAH GREY THOMASON

Source: Studies in Slavic and General Linguistics, Vol. 28, Languages in Contact (2000), pp. 311-
327
Published by: Editions Rodopi B.V.
Stable URL: http://www.jstor.org/stable/40997175 .
Accessed: 02/06/2014 03:53

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

Editions Rodopi B.V. is collaborating with JSTOR to digitize, preserve and extend access to Studies in Slavic
and General Linguistics.

http://www.jstor.org

This content downloaded from 90.210.187.178 on Mon, 2 Jun 2014 03:53:38 AM

All use subject to JSTOR Terms and Conditions
Languages in Contact,editedbyD.G. Gilbers,J. Nerbonne,and J.Schaeken (= Studiesin Slavic
vol.28), 31 1-327.Amsterdam
and GeneralLinguistics, - Atlanta,
GA: Rodopi,2000.

LINGUISTIC AREAS AND LANGUAGE HISTORY

SARAH GREY THOMASON

Linguisticareas,or Sprachbünde, have been thetopicof a verylargeamountof

researchformore thana century.1 But althoughthereare numerousvaluable
studiesof particular linguistic areas and of particular features withincertainlin-
guisticareas,there is stilllittleconsensus on the general nature of thephenome-
non. This paper is a preliminary attempt to characterize the notion 'linguistic
area'. Section§1 belowbeginswitha definition of thetermand a justification of
thedefinition.I will also statemyposition,withreasons,on severalcontroversial
issuesin thisdomain,andthenarticulate whatseemto me to be themostimpor-
tanthistoricalquestions about linguisticareas:How do linguistic areasarise?And
how do theareal structural features originateand diffusethrough thearea? The
sectionconcludeswithan outlineof the crucialrequisitesfordetermining that
contact-induced changehas occurred;thisoutlinesetsthestagefortheattempt, in
§2, to interpretthearealfeatures of fiverepresentative Sprachbünde historically.
Section3 is a briefconclusion.Not surprisingly, giventheimmensecomplexity
and diversityone findsin thecontactsituations thatcompriselinguistic areas,no
simpleanswersto the 'how' questionsare possible;butcomparingdifferent lin-
guisticareasat leastshowswhatsomeofthemanypossibilities are.The mostim-
portant (thoughnotveryneat)conclusion,however,is thatattempts to findvery
generalsocial and/orlinguistic principlesof convergence in a linguisticarea are
doomed- notonlybecause everySprachbund differsfromeveryotherone,but
also becausetheconditionsof contactin largeSprachbünde will inevitably vary
overtimeand space. In otherwords,Sprachbund is nota uniform phenomenon
socially,orhistorically.
linguistically,
1. A definition
and itsramifications
A linguistic
areais a geographical
regioncontaining a groupofthreeormorelan-
guages thatsharesome structuralfeaturesas a resultof contactratherthanas a
resultof accidentor inheritance
froma commonancestor.Threepointsin this

1 This
paper is a revisedversionof the one presentedin November1999 at the Conferenceon
LanguageContactsin Groningen.I am grateful to membersof theaudiencethere,and also to my
studentsin a LanguageContactcoursetaughtat theLinguisticInstitute
at theUniversity
of Illinois
in July1999,forhelpfulcommentson earlierversions.

This content downloaded from 90.210.187.178 on Mon, 2 Jun 2014 03:53:38 AM

All use subject to JSTOR Terms and Conditions
3 12 SARAHGREYTHOMASON

definition requiresomejustification, becausenotall specialistswouldagreewith

them.
First,whymusttherebe at leastthreelanguagesbeforea regioncountsas a
Sprachbund? Whynotjusttwolanguagesin contact?The mostobviousreasonis
thatsubsuming two-language contactsituations undertherubric'linguisticarea'
wouldmeanthatalmosteverycontactsituation in theworldthatinvolvessignifi-
cantstructural interference wouldbe a linguistic area; and althoughthereareim-
portantsimilarities betweeninterference in two-language contactsituationsand
interference in morecomplexcontactsituations, thereare also important differ-
ences. Structural interference in manyor mostSprachbünde is multidirectional,
forinstance, whilestructural interference in manyor mosttwo-language contacts
is unidirectional:so, forinstance,it is clearthatthechangesthatformedthenet-
workofsharedfeatures intheBalkanSprachbund didnotall originate in thesame
language,whilestructural interference in theRomani-Russian contactsituation is
all fromRomanito Russian,notviceversa(at leastas faras Russianas a wholeis
concerned).But some linguistic areas,suchas theEthiopianhighlands,seem to
haveunidirectional interference resulting fromlanguageshift(see discussionbe-
low); and in sometwo-language contacts, suchas Uzbek and Tadzhikin thefor-
mer USSR, structural interference can be foundin both languages,thoughit
didn'tnecessarilyhappensimultaneously or throughout the contactregion(see
Comrie1981: 51, 163; fora broaderviewofTurkic-Iranian contacts,see e.g. Jo-
hanson1992, 1998). Froma historicallinguist'sviewpoint,perhapsthe major
reasonforconsidering two-language contactsseparately fromSprachbünde is that
in thegreatmajority of cases thesourceofa sharedfeature is easierto determine
whenonlytwolanguagesareinvolved.
Second,whytheinsistenceon structuralfeatures in thedefinition of a lin-
guistic area? Again the motive is to avoid an all-inclusivedefinition: if shared vo-
cabularyby itself
were enough to establish a linguisticarea, then the entire world
wouldbe one hugelinguistic area,thanksto suchwidelysharedwordsas email,
hamburger, democracy, pizza, Coca Cola, and television. Usingvocabularyas a
sole criterion wouldtherefore trivializethenotionof a linguisticarea,and we'd
needto inventa newtermforthoserather specialcontactsituations thathavetra-
ditionally been called linguistic areas.
Third,whymustthe sharedfeatures be due to contact?The answerto this
question is thatthat's the whole point the concept.Languagesall over the
of
worldsharenumerousfeatures thatdo notsignalanykindof historicalconnec-
tion;"accidentalsimilarity" is theusual coverterm,thoughit mustbe used with
cautionbecause some sharedfeatures are due to linguistic universaisof various
kinds.Examplesof featuresthatare widelysharedwithouthavinga common
historical sourcearetheexistenceof a phonemeIM,thelackof clickphonemes,a
nounvs. verbdistinction, SOV wordorder,exclusiveuse of suffixes (no prefixes

This content downloaded from 90.210.187.178 on Mon, 2 Jun 2014 03:53:38 AM

All use subject to JSTOR Terms and Conditions
 LINGUISTICAREASAND LANGUAGEHISTORY 3 13

orotheraffixtypes),andpresenceof subjectagreement inflectionon verbs.None

of thesefeatures, withthepossibleexceptionof thenounvs. verbdistinction, is
foundin everylanguagein theworld,butall are commonin widelydistantand
unrelatedlanguages.The othernon-contact sourceof sharedfeaturesis inheri-
tancefroma commonancestor.Sharedstructural features due to inheritance are
foundin themembersof everylanguagefamilyin theworld;to takejust one of
manyexamples,Salishanlanguagesof thePacificNorthwest regionof theUS
and neighboring Canadianprovincesinherited such featuresas labializeddorsal
stops,a globalizedlateralaffricate,lexicalsuffixes, verb-initialwordorder,and a
weaknounvs. verbdistinction fromProto-Salish. The conceptoftheSprachbund
was putforward preciselyin orderto focuson sharedstructural features thatarose
outofcontactrather thanthrough inheritance.
The definition aboveincludesall thecontactsituations traditionallyconsidered
to be Sprachbünde and excludescontactsituations thatare notgenerallyconsid-
eredtobe Sprachbünde. In additionto theserelatively uncontroversial definitional
points,however,thereare severalgeneralissueson whichspecialistsdisagree.I
willdiscussfiveofthesequestionsbriefly.
Do all thelanguagesin a Sprachbund haveto be unrelatedto each other?The
answerto thisfirstquestionis clearlyno. In a largeSprachbundit is virtually
certainthatsomeof thelanguageswill be relatedto each other,and it's possible
thatall of themwillbe. It's easyto see whyone mightwantto focuson changes
in unrelated languagesin a linguisticarea:withrelatedlanguages,distinguishing
changesdue to drift fromchangesdue to contactmaybe verydifficult. Butmeth-
odologicalconvenience cannotbe a validcriterion forSprachbund status,andthe
factthatrelatedlanguagesaremostoftenspokenin contiguousterritories makes
theirinclusioninthesamelinguistic areaall toolikely.One consequenceofthisis
that,in a Sprachbund, demonstrably relatedlanguageswill sharefeatures fromall
threepossiblesources- "accident",inheritance, anddiffusion.
How manysharedfeatures areneededfora regionto countas a linguistic area?
The shortanswerto thisquestionis thatno figurecan be given.But althougha
fewscholarshavearguedthatin principlea singlesharedfeature is enough- Ma-
sica, forinstance,refersto "thelimiting case, the area defined by a singletrait"
(1976: 172) - mostwouldagreethatseveralfeatures areneeded.Campbellet al.
(1986: 533) are certainly correctin asserting thattherecan be no specificlimit
thatwouldpermitus to distinguish putativelinguistic areas"definedon thebasis
of severalfeaturesfromthosebased on a singlesharedtrait";but this surely
doesn'tmeanthatone is forcedto accepta singlefeatureas sufficient evidence
forSprachbund status.Theyrefer(p. 532) to theold questionofhowmanygrains
of sandit takesto makea heap,butthevitalpointis thatit certainly takesmore
thanone or two grains,thoughno precisenumbercan be given.In otherwords,
theproblemis one of fuzzyboundaries, a familiar issue in historicallinguistics:

This content downloaded from 90.210.187.178 on Mon, 2 Jun 2014 03:53:38 AM

All use subject to JSTOR Terms and Conditions
3 14 SARAHGREYTHOMASON

one featureclearlydoes notmakea Sprachbund, two dozen featuresclearlydo,

andtherequisitenumberof features lies somewhere in between.Nor is theprob-
lemwitha one-feature "Sprachbund" merelyone of triviality, pace Campbellet
al. (p. 532). The mainproblemis thata one-feature Sprachbund wouldbe wildly
unrealistic It is difficult
historically. to imaginea processof diffusion thatwould
spreadexactlyone structural feature fromlanguageto languagewithina largere-
gion;in all well-understood contactsituations, diffusion of one structuralfeature
is alwaysaccompaniedby diffusion of at leasta fewothers,evenwhenjust two
languagesareinvolved.
It is conceivablethata Sprachbund could developwitha sizable numberof
sharedarealfeatures and then,withloss of contact,all thediffusedfeatures but
one couldvanishfromall thelanguages.Butwithsucha historicalscenario,and
withoutevidencethatthereused to be moresharedfeatures, no responsiblehis-
toricallinguist wouldbe likelyto claimtheareaas a Sprachbund. Moreover,con-
siderthesinglesharedfeatureitself(again assumingthatthereis no old docu-
mentation to provideevidencethatthe languagesof theproposedSprachbund
once sharedmorefeatures). If it is a markedfeature, whyis it theonlydiffused
innovation orrelic?Mosthistorical linguists wouldarguethatmarkedfeatures are
less likelyto diffuseand morelikelyto disappearthanunmarked features.But if
thesinglesharedfeatureis unmarked, how could one possiblytellwhetherit's
dueto contactornot?It couldeasilyhavearisenvia independent changein all the
languages that haveit.2
The nexttwoquestionsconcernthedistribution of thearealfeatures. First,do
all thesharedfeatures have to be in all thelanguagesof the Sprachbund? No,
surely not: if theanswer to thisquestion were yes, thetotal number ofthe world's
linguistic areaswouldimmediately shrinkfrommanyto zero,becausethereis no
Sprachbund inwhichall thearealfeatures arefoundin all thelanguages.The rea-
sonsforthisareeasyto find.Forone thing,innovations aresureto spreadamong
thelanguagesof a Sprachbund differentially - some changes spreadfarther than
others,and if changesstartin different places there will inevitably be different
patterns of spread.And supposethattwo languagesin a Sprachbundacquirea
certainfeatureby borrowing froma thirdlanguagein the area,but thatone of
theselanguagessubsequently losesthefeature through internal change.It is likely
tobe impossibletoprovethatthefeature everexistedin thatlanguage,andyetthe
languagemayotherwise be a definite memberoftheSprachbund.
The nextquestionis, do thesharedstructural features thatcharacterize a par-
ticularlinguistic areahaveto be confined to thearea?Againtheansweris clearly
no. Supposethata languageX is boththesourceofan arealfeature anda member

2 It
may be useful,as Campbell et al. suggest,to distinguishstronglinguisticareas fromweak
ones. But I believethat,in practiceas well as in principle,severalsharedfeaturesare neededeven
to establisha weak linguisticarea.

This content downloaded from 90.210.187.178 on Mon, 2 Jun 2014 03:53:38 AM

All use subject to JSTOR Terms and Conditions
 LINGUISTICAREASAND LANGUAGEHISTORY 3 15

ofa languagefamilythatalso has membersoutsidetheSprachbund. If thefeature

is inherited in X and stillpresentin itssisterlanguagesoutsidethearea,thenthe
feature is obviouslynotconfined to thelinguistic area;butifit spreadswidelyto
other(unrelated)languagesin thearea,it wouldcountas an arealfeaturetoo. In
addition,speakersof someof thearea's languagesare likelyto have some social
contactsbeyondthearealboundaries, andbythatmeanstheremaybe limiteddif-
fusionofarealfeatures to outsidelanguages.Andfinally, a featuremayspreadvia
contactwithina Sprachbund and also occurin neighboring languagesoutsidethe
areaevenwithoutdiffusion, especiallyifit's unmarked. So a criterionthatinsists
on exclusivity of arealfeatures is as worthless as a criterion thatrequiresuniver-
salityofarealfeatures withintheSprachbund.
Takingall thesedistributional considerations intoaccount,we would predict
neitheruniversality nor exclusivity of areal features withina Sprachbund.It is
therefore hardlysurprising thatinspectionof linguisticareas aroundthe world
supportthisprediction; forinstance,as Campbellet al. observedwithrespectto
theBalkan Sprachbund, "fewBalkan isoglossesbundleat the [linguisticarea's]
borders;some fail to reach all the Balkan languages,while othersextendbe-
yond..."(1986: 561).
The finalquestionconcernsthenatureof thelinkbetweenlanguagecontact
and contact-induced languagechange:is a Sprachbund inevitablewhenthreeor
morelanguagesarein intimate contactfora longtime?Much of theliterature as-
sumesa 'yes' answerto thisquestion;butin myopinionthequestioncan'tbe an-
sweredwithconfidence at ourpresentstateofknowledge, becausetheonlycom-
plex contact situations that have been studied intensively farare thoseinvolv-
so
ing contact-induced changes. I doubt if a Sprachbundis inevitableeven under
long-term intimate contact, however. The main reasonforthisbeliefis thatcon-
tact-induced is
change demonstrably not inevitable in intensetwo-language con-
tactsituations, and I can see no reasonwhyaddingmorelanguagesto a contact
situationshouldchangethepicture.3 The basic problemwithpredicting thata
Sprachbund must arise under certain contact conditionsis thatculturalattitudes
may,andsometimes do,inhibit lexicaland/or structural interference.
All thepointsof controversy discussedso far,thoughimportant issues,areul-
timately less interesting thanthetwovitalopenhistorical questionsaboutSprach-
bünde:How do linguistic areasarise?And how aretheirlinguisticfeatures to be
interpreted The
historically? answerto theformer questionis thatlinguistic areas
emergethrough diversesocial processesand institutions (e.g. traderelations,ex-
ogamy,andwar);theanswerto thesecondquestionis thatthehistorical interpre-
tationsvaryas muchas linguistic areasvary.The nextsectionillustrates thisdi-

3MontanaSalishandatleastsomeother
languagesoftheUS Northwest, forinstance,
havebor-
rowedfrom eachother,buthavevirtually
no loanwordsandlittleorno borrowedstructure
from
inspiteofwellovera hundred
English, yearsofintimate
contact.

This content downloaded from 90.210.187.178 on Mon, 2 Jun 2014 03:53:38 AM

All use subject to JSTOR Terms and Conditions
3 16 SARAHGREYTHOMASON

andhighlights
versity, theproblemsone encounters in trying
to answerthesetwo
questions,througha surveyoffivelinguistic
areas.
Beforebeginning thesurvey,though,we need to set the stageby listingthe
requisitesforestablishing changehas occurred.For simplic-
thatcontact-induced
ity'ssake,thelistbelow assumesjust twolanguagesin contact,X (theproposed
receivinglanguage)andY (theproposeddonorlanguage),buttheprincipleis the
sameformorecomplexcontacts. Therearefourrequirements:
(1) Establishthattherewas contactintimateenoughto permitcontact-induced
structural
change.
(2) Findseveralindependent in X andY - ideally,features
sharedfeatures in dif-
ferent
grammatical subsystems.
(3) Provethatthesharedfeatures
werenotpresent inpre-X.
(4) Provethat thesharedfeatures
were in
present pre-Y.

Note thatrequirement (1) is easy to satisfyifthereare loanwordsfromY in X.

But theremightnotbe any:ifY speakersshifted to X, interferencefeaturesin X
are morelikelyto be phonologicaland syntactic thanlexical. Worse,if all Y
speakersshifted to X, and ifY had no close relatives, it mightbe impossibleto
identifya sourcelanguageforthe suspectedinterference features.The crucial
pointto be madehereis thatifrequirements (l)-(4) can'tbe satisfied,thenitwill
be impossibleto makea convincing case forcontact-induced change.This does
notmeanthata givenfeature is notdue to theinfluence of anotherlanguage;but
itwon'tbe possibleto distinguish betweena contactoriginforthefeature and an
ancientor recentinternal origin(see Thomason1986, 1993 formoredetailedar-
guments in supportofthisclaim).4

linguisticareas
2. A surveyoffive"representative"
The word"representative" needsshudderquotesin thiscontextbecauseit isn'tat
all clearwhatwould countas a trulyrepresentative Sprachbund. The ones out-
linedbriefly in thissectionarediversegeographically butitmay
andhistorically,
wellbe thatchoosingfivedifferent conclu-
areaswouldresultin quitea different
sion(notthatmyconclusionswill turnoutto be veryconclusive).Still,thissur-
vey at leastsuggestswhatsortsof factorsneedto be consideredin thehistorical
interpretationoflinguisticareas.The fiveareasthatwillbe discussedaretheBal-
kansof southeastern Europe,theSepikRiverBasin in New Guinea,The Pacific
Northwest ofNorthAmerica,theEthiopian highlandsinAfrica,andSouthAsia.

4 Because of I omitherediscussionof MULTIPLECAUSATION, in whichan exter-

space limitations,
nal sourcecombineswithan internalsourceto producea particularchange; in a morecomplete
studyofthechangesthatbringaboutSprachbünde, mustbe considered.
multiplemotivations

This content downloaded from 90.210.187.178 on Mon, 2 Jun 2014 03:53:38 AM

All use subject to JSTOR Terms and Conditions
 LINGUISTICAREASAND LANGUAGEHISTORY 3 17

2.1 The Balkan Sprachbund

The Balkanpeninsulais theworld'smostfamouslinguistic area,andtheone that
has receivedthemostattention fromscholarsoverthelongestperiodof time.Its
majorlanguagesare Rumanian(a Romancelanguage);Bulgarian,Macedonian,
and southeastern dialectsof Serbian(all are Slavic languages);Albanian;Greek;
perhaps Balkan dialects of Romani(an Indie language);and dialectsof Turkish
thatare spokenin theBalkans.All of theselanguagesexceptforTurkishbelong
totheIndo-European languagefamily.
Arealfeatures are commonin theBalkans,with,as notedabove,varyingdis-
tributions withinthelanguagesof theSprachbund. Here are a fewtypicalexam-
ples: there are many Turkish and Greek loanwords in (other)Balkan languages;
among the more widespread structuralBalkanisms are thepresenceof a highor
midcentralvowel,vowelharmony, thepartialor totalloss of theinfinitive, post-
posed articles,pleonasticobjectmarkers,a mergerof the dativeand genitive
cases,a future construction formed withtheverbwant,and a perfect construction
formedwiththeverbhave.Arealfeatures withmorelimiteddistribution within
theSprachbund area changeofunstressed to
[o] [u] (in Bulgarian,Rumanian,and
Albanian), a Slavic diminutive suffix-ica (in Greek,Rumanian,and Albanian),a
plural suffix borrowed into Arumanian from Greek,the replacement of dative
feminine pronouns with dativemasculine pronouns in Macedonian as a resultof
Albanianinfluence, and a vocativecase in Rumanianas a resultof Slavic influ-
ence.
The listof Balkanismscould easilybe extended:it mustbe emphasizedthat
theseareonlyexamplesofthewholecomplexof arealfeatures (see, forinstance,
Sandfeld1930 and Lindstedtin thisvolumeformoredetailand further refer-
ences).But itmustalso be emphasizedthata completelistwouldnotamountto
massiverestructuring in anyof theBalkanlanguages:therehas been significant,
butby no meansextreme, contact-induced structuralchangein themost-affected
Balkanlanguages.For instance,theamountof changein Macedonian,whichis
generallybelieved to be the most Balkanized of all the languages in the
Sprachbund, wouldprobablyfallintocategory4, thesecondhighestcategoryof
'ordinary' structural in theborrowing
interference, scale proposedin Thomason
andKaufman(1988: 74-76).
The crucialquestionis, whatare thesourcesof theareal featuresin theBal-
kans?Severalwritershave arguedformultiplecausation(see especiallyJoseph
1983,in severalpassageslistedunder'causation'in theindex,and Lindstedt, this
volume),butin thepresentcontextthefocusis on theexternal motivations forthe
innovations. In spiteof variousproposalsforsourcelanguages(e.g. Greek),there
is littleagreement amongBalkanologistsaboutthehistoricaloriginsof mostof
themostfamousBalkan features; butit is at leastclearthatthenumerousarea-
wideandlocal contactsituations wereextraordinarily complex.The history ofthe

This content downloaded from 90.210.187.178 on Mon, 2 Jun 2014 03:53:38 AM

All use subject to JSTOR Terms and Conditions
3 18 SARAHGREYTHOMASON

Balkansis famouslyturbulent. Most notoriously, fivehundredyearsof Turkish

invasionand conquestled directly and indirectly to large-scalemultilingualism,
promoted or at leastfacilitated
bymovements of smallgroupsofpeopleandeven
entiresmallcommunities fromregionto regionandsmall-scalelanguageshiftsin
certainregionsand at different times.Multilingualism was presumablyrather
symmetrical over the area as a whole: withthepossibleexceptionof Turkish
speakersduringperiodsof Turkishrule,area-wideone-waybilingualismof the
sortthatoftenaccompaniesasymmetrical dominancerelationsis unlikelyto have
obtained.
The sourcesof lexicalfeatures arerelatively easyto establish.We can be cer-
tain,forinstance,aboutGreekand Turkishloanwordsin theBalkan languages.
The sameis trueof a fewof thestructural features; Latinhad a perfectconstruc-
tionwithhave,forexample,so Romanceis a plausiblesourceforthatBalkan
feature.Thingsaremuchless clearformostofthestructural butit seems
features,
atbestriskyto assumea singlesourceforthem.It is muchmorelikely,giventhe
populationmovements and theresulting intimate contacts,thatfeaturesarosein
differentplaces at different timesand then,as is commonin linguisticareas,
spreaddifferentially withintheSprachbund. It is also likelythatmostofthediffu-
sion of features was via borrowing, i.e. incorporation of featuresfromone lan-
guageintoanother bybilingualspeakers, rather thanvia imperfect learningduring
a processof grouplanguageshift,because apparently no large-scalelanguage
shiftstookplace duringtherelevantperiod(veryroughly,1000-1800CE). But
thenon-lexicalBalkanisms,especially,could in principlebe due eitherto bor-
rowingor to shift-induced andthepresenceof numerousloanwords
interference,
fromlanguagesthatwereveryunlikelyto be thesourceofmostBalkanismsdoes
nothelptoresolvethepuzzleoforigins.

2.2 The Sepik RiverBasin

Papua New Guineahas longbeen famousforarealphenomena:see, forinstance,
Arthur Capell's comment,withreference to thenon-Austronesian languagesof
the centralhighlands,thatalthoughneighboring languageshave different vo-
cabularies,theirgrammatical "recurwithalmostmonotonousregularity
features
fromlanguageto language"(citedin Wurm1956: 451). WilliamFoley,in his
book on Papuanlanguages,observesthat"Papuanlanguagesare generallyin a
stateof permanent contactwitheach other"(1986: 210). Foley system-
intimate
atically explores the topic of contact-induced change in Papuan (non-
Austronesian) languagesof New Guineaand includesa veryusefulanalysisof
one smallSprachbund comprising threeneighboring languages- Yimas, Alam-
-
blak,and Enga whichbelongto threedifferent languagefamilies,all spokenin
theSepikRiverBasin of northern Papua New Guinea(1986: 263-267). Foley's
detailedstudyof the featuressharedby theselanguagesmakesit clearthatthe

This content downloaded from 90.210.187.178 on Mon, 2 Jun 2014 03:53:38 AM

All use subject to JSTOR Terms and Conditions
 LINGUISTICAREASAND LANGUAGEHISTORY 3 19

Sprachbund is an old one and that,as in theBalkans,diffusion has been multi-

He also showsclearlythatnotall thesharedfeatures
directional. can be shownto
havediffused at all.
All threelanguageshavepalatalconsonants, withEnga as thepresumedsource
because such consonantswere demonstrably inherited by Enga but not by the
otherlanguages;andall threehavecomplextensesystems(withat leasta present,
a future,and threepasts),a feature thatseemsto be old and perhapsinherited in
bothEnga andAlamblakbutprobablyinnovative inYimas.
Yimas and Alamblaksharethe largestnumberof features.Four of these,a
particularpluralpronominal a typeof temporaladverbialclause in which
suffix,
an oblique suffix-n is added to theinflected verb,an elaboratesystemof verb
compounding, and a causative construction,have no detectablesources,though
Foley believes that at leastsome ofthem are due to diffusion.One sharedfeature,
boundadverbialformsin theverb,probablyreflects diffusion fromAlamblakto
Yimas; anda sixthfeature, thepresenceofmorethanone centralvowel,seemsto
havebeeninherited bybothlanguages.
Yimas and Enga sharetwo features not foundin Alamblak:one of these,a
causativeformedwith-(a)sa, is an Enga interference featurein Yimas, but the
sourceof theotherone, an indirect causativeformedwith'say', is obscure.Fi-
nally,Alamblakand Enga sharea switch-reference construction thatdiffused
fromEngato Alamblak.
In otherwords,it is possibleto establishsourcesforsome,butnotall, of the
featuresthatare sharedby two or all threelanguagesin thissmall Sprachbund.
It's noteworthy thatYimas is alwaysa recipientlanguageof non-inherited fea-
turesthatit shareswithone or bothof theotherlanguages,neverdemonstrably
thesourcelanguage;beyondthis,however,thereappearsto be no definite infor-
mationaboutspecificprocessesof diffusion. In thecase of sharedfeatures inher-
itedfromtherespectiveproto-languages, ancientdiffusion is possiblefromone
proto-language to the other; butthe chances of such
establishing ancientdiffusion
range from slim to none. The greatvirtue of Foley's studyis thatit lays out the
difficulties
witha historical of a
analysis Sprachbund so clearly.

2.3 The PacificNorthwest

In thenorthwesternUS statesWashington andOregonand in neighboring British
Columbia,together withlimitedcontiguousareasfarther there
inland, is a well-
knownbutas yetunderstudied area.
linguistic The three
corelanguage familiesof
thisSprachbundare Salishan(about21 languages),Wakashan(6 languages),and
Chimakuan(2 languages).Smallernumbersof thearealfeatures are also shared
byotherlanguagesin theregion:Tsimshian, Chinookanand Sahaptianlanguages,
theisolateKutenai,andto a slightextentnearbyAthabaskan languages.As in the
highlandsof New Guinea,the level of multilingualism was apparently always

This content downloaded from 90.210.187.178 on Mon, 2 Jun 2014 03:53:38 AM

All use subject to JSTOR Terms and Conditions
320 SARAHGREYTHOMASON

highin theNorthwest, so thatheretoo one mayreasonablyspeakof permanent

intimate contactamongneighboring languages.
Froma historicalviewpoint, thePacificNorthwest Sprachbundis the Sepik
RiverBasin writlarge.Most of the widespreadareal featuresmustbe recon-
structed forall threeofthecoreproto-languages (Kinkade1997). The moststrik-
ingofthesearelabializeddorsalconsonants, a velar/uvular distinctionin dorsals,
lateralobstruents suchas lateralaffricates anda voicelesslateralfricative, /ts/
af-
fricates,a verycommonsound changefromvelarsto alveopalatals,complex
wordstructure withmanysuffixes butrelatively fewprefixes, minimalcase sys-
tems,possessivepronominal affixesaddedto a possessedword,verb-initial sen-
tentialwordorder,sentence-initial negation,thepresenceof a yes/noquestion
particle,a weak lexicalnoun/verb distinction, pairsof rootsreferring to singular
vs. pluralactionsor states,optionalpluralmarking, distributivepluralsformed by
reduplication, numeralclassifiers (e.g. 'person'vs. 'non-person' categories),anda
systemof lexical suffixes(withconcretemeaningslike 'hand' and 'roundob-
ject'). It seemsveryunlikelythatall thesefeatures - whichincludemanythatare
certainly independent ofeachotherandseveralthatarehighlymarkedin termsof
theirdistribution in theworld'slanguages- are accidentally sharedby all these
languagefamilies;but if therewas diffusion betweentwo or moreof the core
families'parentlanguages,it cannotbe established, at leastnot on thebasis of
current knowledge.Thatis: diffusion maybe suspected, butthere'sno directevi-
denceto support a diffusionalhypothesis.
Otherarealfeatures withintheSprachbund have limiteddistribution. Two of
the most strikingof these,strikingbecause they are extremelyrare cross-
linguistically outsidethisarea,arethepresenceof severalpharyngeal consonants
anda soundchangethatreplacednasal stopswithvoicedoralstops.What'sespe-
ciallystartling aboutthesetwofeatures is thattheyappearin non-contiguous ar-
eas, a factthatwould standin theway of a straightforward diffusion originhy-
pothesisevenifwe knewwhereeach of thefeatures appearedfirst.Most of the
otherlimitedareal features, whichare quitenumerous, also lack a clear source,
thoughwe knowat leastthattheywerenotinherited fromtherespective proto-
languages.Only a few,such as the lack of elaboratesyllable-initial consonant
clustersin theSalishanlanguageComox(due to interference fromtheWakashan
languageKwakwala)and a nonglottalized lateralaffricateborrowedby theChi-
makuanlanguageQuileutefromWakashan,can be traceddefinitely to a particu-
larsource.Andas withtheSepikRiverBasin Sprachbund, theprocessesofdiffu-
sion throughwhichthe PacificNorthwest Sprachbundarose cannotbe deter-
mined.

This content downloaded from 90.210.187.178 on Mon, 2 Jun 2014 03:53:38 AM

All use subject to JSTOR Terms and Conditions
 LINGUISTICAREASAND LANGUAGEHISTORY 321

2.4 The Ethiopianhighlands

The Africanlinguistic areathathas receivedthemostattention in theliteratureis
the EthiopianSprachbund,more specificallythe languagesof the Ethiopian
highlands(see e.g. Leslau 1945, 1952,Hetzron1975,Moreno1948,Little1974,
and Ferguson1976). Thereare manylanguagesin theregion,threequartersof
themmembersoftwobranchesoftheAfro-Asiatic languagefamily;of these,the
greatmajority belong to the Cushitic branch of the family(includingtheso-called
Omoticlanguages),but thereare also severalSemiticlanguages.Besides the
Afro-Asiatic languages,therearesomelanguagesin theregionthatbelongto the
proposed Nilo-Saharan family, butsincetheseare notdiscussedin theliterature
on thelinguisticarea,I cannotcommenton whether, or how, theyfitintothe
Sprachbund.
The areallinguistic features havevarying distributions, as is typicalof linguis-
ticareas:a fewfeatures are area-wide, butmosthave localizeddistributions. The
languagesspoken in the southern part of the area have significantly more of the
features.Among the areal features are the presence of labialized dorsal stops,al-
veopalatalconsonants, prothetic glides before mid a
vowels, separate futuretense,
a causativeformation witha doubleaffix(prefixes in EthiopieSemitic,suffixes in
a
Cushitic), negativeperfectformation, lack of a dual number category, optional
ratherthanobligatory pluralmarkingon nouns,SOV word orderwithVerb-
Auxiliary,Adjective-Noun, andRelativeClause-Nounwordorders, postpositions,
andsubordination by means ofnon-finite gerundconstructions. Besidestheseand
otherstructural features, thereare also sharedlexicalfeatures - manywords,in-
cluding some quite basic terms (e.g. kin terms,numerals, bodyparts),and
and
also derivationalsuffixes anda vocativeparticle.
Froma historical viewpoint, theEthiopianSprachbund differs strikinglyfrom
theotherthreeareaswe have examined:themajorinterference hereseemsto be
unidirectional,fromCushiticto Semitic,andtheprocesswas apparently imperfect
learning that occurred when some (groupsof) Cushiticspeakersshiftedto the
Semiticlanguage(s)spokenbynewcomers to theregion.(This statement requires
a hedge,becauseto dateno systematic historical researchhas beencarriedouton
thenon-Semitic languagesofthehighlands; itmaywellbe thatsomeinterference
fromSemiticto Cushiticwill be found,and/orinterference betweenthe Afro-
Asiaticand thenon-Afro-Asiatic languages.) It is at least certain thatEthiopie
Semitichas numerouslexicaland structural features thatarelikeCushiticandun-
likeolderSemitic.The age of thefeatures in Cushiticis less clear,however;it's
for
quitepossible, instance, thatthe Cushitic languagesacquiredthefeatures from
non-Afro-Asiatic languages rather than by inheritance from Proto-Cushitic. Still,
thefeatures aremorewidespreadin theCushiticlanguagesof theareathanin the
Semiticlanguages,and thisfact,together withspecificstructural considerations,
makesan immediate Cushiticsourcelikely.Thereis also sociolinguistic evidence

This content downloaded from 90.210.187.178 on Mon, 2 Jun 2014 03:53:38 AM

All use subject to JSTOR Terms and Conditions
322 SARAHGREYTHOMASON

thatsup-
thatCushiticspeakersdid shiftin numbersto Semitic,a circumstance
portstheanalysisofshift-induced
interference.

2.5 SouthAsia
Like the Balkans,thoughto a lesserextent,SouthAsia is a well-knownand
much-studied linguisticarea (see e.g. Emeneau1980 and Masica 1976). Unlike
theBalkans,however,its statusas a linguistic area is a matterof considerable
controversy; thoughfew authorsdeny flatlythatit should be classed as a
Sprachbund, thearealstatusof severalof themostfamousfeatures is vigorously
disputedand,even more,thehistoricalinterpretation of theemergenceof areal
featuresin thesubcontinent is disputed(in e.g. Hock 1975, 1984). I will notat-
temptto resolvethiscontroversy here,butitshouldbe notedthattherearedoubts
aboutsomeofthefeatures.
The Sprachbundcompriseslanguagesbelongingto at least threedifferent
families:Dravidian,Indie(a sub-branch ofIndo-European), andMunda(a branch
of Austro-Asiatic). A fewof thearealfeatures are also sharedby theisolateBu-
rushaskiandby someIranianlanguages(IranianandIndietogether forma branch
ofIndo-European) .
As is typicalof linguistic areas,some features are morewidespreadthanoth-
ers,and someofthearealfeatures areclearlyolderthanothers.Amongthemost
prominent ancientarealfeatures arethepresenceofretroflex consonants, aggluti-
nationin noun declension,a particular echo-wordformation, a quotativecon-
struction,absoluteconstructions whichdifferfromthe typicalIndo-European
type,thesyntaxof a discourseparticle(Indieapi, Dravidian*-wm),SOV word
order,morphological causatives,anda 'secondcausative'construction.
Languagecontactin theIndiansubcontinent has a verylongand a verycom-
plex history,datingback overthreethousandyearsor more.Multilingualism is
thenormtodayin somepartsof thearea,forinstancein thevillageof Kupwar,
where,accordingto Gumperzand Wilson 1971, grammatical convergencehas
been extreme. In Kupwar,theprocessby whichtheIndielanguagesMarathiand
especiallyUrduandtheDravidianlanguageKannadahave convergedgrammati-
callywas apparently borrowing, not shift-induced interference. And it was cer-
tainlymultidirectional:Kannadais thesourceof somefeatures and Marathiis the
sourceof others.Thereis everyreasonto believethatthissortof convergence
happenedelsewherein SouthAsia, so thatKupwarmightbe a miniature reflec-
tionoftheSprachbund as a whole.
However,thesituation seemsto havebeendifferent fortheancientarealfea-
tures.The majorityof thoseare reconstructable forProto-Dravidian butnot for
Indie,whichmeansthatDravidianhas almostsurelyinfluenced Indie,not(in an-
cienttimes)vice versa.Significantly, thereareveryfewold Dravidianloanwords
in Indielanguages;this,takentogether withthestructural interference,pointsto a

This content downloaded from 90.210.187.178 on Mon, 2 Jun 2014 03:53:38 AM

All use subject to JSTOR Terms and Conditions
 LINGUISTICAREASAND LANGUAGEHISTORY 323

interference.
process of shift-induced There is good evidencethatDravidian
speakerswerein SouthAsia whenIndiespeakersarrivedthere,and it is at least
verylikelythatmanyofthemshifted to theIndielanguage(s)oftheinvaders.The
SouthAsian Sprachbund thereforeresemblestheEthiopianSprachbundin pre-
fortheearliestperiodofintimate
senting, a pictureofunidirectional
contact, inter-
ferencevia imperfect
learningofa targetlanguagebyshifting speakers.

3. Conclusion
As we have seen,it is oftenpossibleto establisha sourcelanguageor language
familyfora particular arealstructural feature in a Sprachbund, butveryoftenno
sourcecan be establishedor,in manycases, evenguessedat. For thesefeatures,
theshortanswerto thequestion'wheredo thefeatures come from?',therefore, is
a largequestionmark:we don'tknow.The bestchancesforestablishing sources
forarealfeatures willbe in linguisticareasthatarerelatively simplesociolinguis-
tically,with(mostly)unidirectional ratherthanmultidirectional In
interference.
practice,as faras I can tellfroma reviewofnumerous areasaroundthe
linguistic
world,theseare cases wheretherehas been large-scaleshiftby speakersof one
groupof relatedlanguagesto a different groupof relatedlanguages- as (appar-
ently)in theEthiopianhighlandsand ancientSouthAsia - so thatmostinterfer-
encefeatures aredue to imperfect learningof thetargetlanguagesby theshifting
speakers.Unfortunately forthosewho yearnforeasy solutionsto historical puz-
zles, Sprachbündethatare relativelysimplesociolinguistically are much less
commonthanthemorecomplexkinds.
Even in sociallymorecomplexlinguisticareas,however,a longerand more
substantialanswercan be givento the'wherefrom?'question.Therearefourob-
viouspossiblesourcesforarealfeatures whoseorigincan'tbe tracedto anyofthe
languagesof theSprachbund. First,theycouldall havebeen inherited froma re-
moteproto-language fromwhichall the languagesof the Sprachbundare de-
scended.But ifno geneticrelationship can be establishedamongsome or all of
the languagesin the area, the putativeall-encompassing proto-language must
have been veryremoteindeed,beyondthe reachof the ComparativeMethod;
methodologically, therefore, thisfirstpossibilityis worthless,and mustbe in-
cludedunderthesecondpossibility.
The secondpossibility is thatthearealfeatures couldbe "accidentally"shared
(and herewe mustremember thatin thiscontextthetermincludestheoperation
of variouskindsof universallinguistic tendenciesas well as genuineaccident).
Thisis unlikelyforhighlymarkedfeatures likethePacificNorthwest pharyngeals
andnasal-lessconsonant inventories, in spiteofthefactthatbothfeatures seemto
be independent innovations in at leastpartoftheirpresentterritory. The possibil-
ityof accidentis muchgreaterforuniversally unmarkedfeatures, whichcould
easilyariseas independent innovations inneighboring languages.

This content downloaded from 90.210.187.178 on Mon, 2 Jun 2014 03:53:38 AM

All use subject to JSTOR Terms and Conditions
324 SARAHGREYTHOMASON

The othertwo possibilities bothinvolvecontact-induced change.One is that

each feature arosethrough internal changein somelanguage(or proto-language)
in the area and thenspreadto some or all of the otherlanguages(or proto-
languages).Thisis ofcoursepossibleevenwhen,as in theBalkans,we knowthat
theproto-language lackedthefeature, becauseitcouldhavearisenaftertheproto-
languagesplitintotwoor moredaughter languagesbutbeforedocumentation of
thedaughter languages.Ifthishappenedin one oftheproto-languages beforeany
splitsaffected thelanguagesin contact, contact-induced changecouldproducethe
situation we findin thePacificNorthwest: a numberofarealfeatures thatmustbe
reconstructed forall threecorefamilies'proto-languages. Thispossibility, unfor-
tunately, is not amenableto testingor proof,unlessfurther historicalresearch
permitstheestablishment of broadergeneticrelationships amongan area's lan-
guagesandhenceearlierproto-language reconstructions.
The fourth and lastpossibility is thatan arealfeature mayarisethrough a pro-
cess of 'negotiation' - in thiscase,as a misperception by semi-bilinguals ofan L2
structure. This misperception could thenspreadnot onlyto the misperceivers'
ownlanguagebutalso totheL2 andbeyond.An exampleofthisprocessin a two-
languagecontactsituation is thefixingofstresson thepenultin a northern dialect
of Serbo-Croatian - whichlikeotherSerbo-Croatian dialectsoriginally had free
stress- underthe influenceof Hungarianspeakerswho realizedthatSerbo-
Croatiandidn'thave initialstress(as in Hungarian)butnevertheless assumeda
fixedstresspattern and settledon thepenultas its location(Ivie 1964; see Tho-
mason1997 forfurther discussionofnegotiation as a mechanism ofinterference).
The problem,of course,is thatin thevastmajority of linguisticareasthereis
no hope of distinguishing betweenthethirdand fourth originscenariosforhis-
torically mysterious areal features, evenif "accident"can reasonablybe consid-
eredrelatively unlikely groundsof plausibility.
on This is simplyone morein-
stanceof an uncomfortable truth:historicallinguists,like otherhistoricalscien-
tists,areforcedto deal withlimitations on hypothesis testingthatareimposedby
gaps in thehistorical record.This does notmeanthatwe shouldstoplookingfor
solutionstopuzzles;itdoesmeanthatwe shouldbe ableto recognizewhenwe've
reachedthelimitsofhistorical knowledge, so thatwe don'tgo beyondtheminto
historical fantasy.
Specifically, thefactthatwe can listpossibleexplanations forunsourcedareal
features is notan indication thatwe can expectto establishsourcesforall ofthem
eventually. Ultimately, thereasonforthisis thatwe can't meetone or bothof
requisites (3) and (4) for provingthatcontact-induced changehas occurred- that
is, we can't provetheabsenceof a sharedfeaturein one or moreof theproto-
languagesand/orthepresenceof the featurein theotherproto-language(s). Of
coursethishappensin historical investigations oftwo-language contactsituations
as well,butit seemsto be a worseproblemforSprachbünde (thoughthisimpres-

This content downloaded from 90.210.187.178 on Mon, 2 Jun 2014 03:53:38 AM

All use subject to JSTOR Terms and Conditions
 LINGUISTICAREASAND LANGUAGEHISTORY 325

sionmightbe due to thefactthatthereis less researchon linguisticareasthanon

two-language contacts).
A finalconcludingremarkis in order.Even in thestrongest Sprachbünde, the
often-cited'tendencytowardisomorphism' rarelyifeverleads to massiveoverall
convergence.Even in the Kupwarcase, a mini-Sprachbund withinthe larger
SouthAsia linguisticarea,onlysixteenfeatures,all ofthemmorphosyntactic, are
discussedin GumperzandWilson'sfamous1971 article,andthetotalamountof
changein any one of the languagesis not all thatradical(see Thomasonand
Kaufinan1988: 86-88 fordiscussion).Thereare probablymanyreasonsforthe
lack of massiveoverallconvergence in linguistic
areas,all of themsocial rather
thanlinguistic.But surelya majorfactoris thatthe'other-directed'attitudes that
promote convergence (presumablyin conjunctionwithcognitivefactors havingto
do withease of processingof severallanguages)are counterbalanced by a 'self-
directed'worldview thatpromotesmaintenance of one's own cultureand lan-
guage(Foley 1986: 27 etpassim).Bothof theseattitudes are displayedin an old
Croatiansayingthatcelebratesmultilingualism:
Kulïko jezïkou clovïg znâ,
taîïkoclovïg vajâ.5

University
ofMichigan

REFERENCES
Lyle,Terrence
Campbell, Kaufinan,andThomasC. Smith-Stark
1986 "Meso-America as a linguistic
area",Language62,530-570.
Comrie,Bernard
1981 ThelanguagesoftheSovietUnion.Cambridge: Cambridge Press.
University
Emeneau,Murray B.
1980 Languageandlinguistic area(ed.byAnwarDil). Stanford:
Stanford
University
Press.
Charles
Ferguson, A.
1976 "TheEthiopian languagearea",in: M. LionelBender,J.DonaldBowen,R.L.
Cooper,andC.A. Ferguson (eds.),LanguageinEthiopia,63-76.Oxford:Ox- '
fordUniversityPress.
Foley,William
A.
1986 ThePapuan languagesof New Guinea.Cambridge: CambridgeUniversity
Press.
Gumperz,John J.,andRobert Wilson
1971 "Convergence andcreolization:a casefrom theIndo-Aryan/Dravidian
border",
in: Dell Hymes(ed.),Pidginization and Creolization
ofLanguages,151-167.
Cambridge: Cambridge Press.
University

5"However a person that'showmucha personis worth."

manylanguages knows,

This content downloaded from 90.210.187.178 on Mon, 2 Jun 2014 03:53:38 AM

All use subject to JSTOR Terms and Conditions
326 SARAHGREYTHOMASON

Hetzron,Robert
1975 "Geneticclassification and EthiopieSemitic", in: JamesBynonandThodora
Bynon(eds.),Hamito-Semitica, 103-127.TheHague:Mouton.
Hock,HansHenrich
1975 "Substratum influence on (Rig-Vedic)Sanskrit?", Studiesin theLinguistic
Sciences5/2,76-125.
1984 "(Pre-)Rig-Vedic convergence ofIndo- AryanwithDravidian? Another lookat
theevidence", StudiesintheLinguistic Sciences14/1,89-107.
Ivic,Pavle
1964 Balkanlinguistics. Lecturecoursetaught at theLinguisticInstituteoftheLin-
guisticSociety ofAmerica, Indiana University, August1964.
June-
Johanson,Lars
1992 StrukturelleFactor en in Türkischen Sprachkontakten. Stuttgart:FranzSteiner
Verlag.
1998 "Code-copying inIrano-Turkic",LanguageSciences20,325-337.
Joseph,BrianD.
1983 Thesynchrony anddiachrony oftheBalkaninfinitive:a study inareal,general,
andhistorical Cambridge:
linguistics. Cambridge UniversityPress.
Kinkade, M. Dale
1997 "Theemergence ofsharedfeatures inlanguages ofthePacificNorthwest". Pa-
perpresented at thesymposium ThePacificNorthwest as a Linguisticand
CulturalArea,American Association fortheAdvancement of ScienceAnnual
Meeting, Seattle.
Leslau,Wolf
1945 "Theinfluence ofCushiticon theSemitic languagesofEthiopia: a problemof
substratum", Word1,59-82.
1952 "Theinfluence ofSidamoontheEthiopie languages oftheGurage", Language
28,63-81.
Jouko
Lindstedt,
2000 "Linguistic Balkanization:contact-induced changeby mutualreinforcement".
Thisvolume, 231-246.
GretaD.
Little,
1974 "Syntactic evidenceof languagecontact: Cushiticinfluence in Amharic", in:
RogerW. ShuyandCharles- James N. Bailey(eds.),Towardstomorrow's lin-
267-275.Washington,
guistics, DC: Georgetown University Press.
Masica,ColinP.
1976 Defining area:SouthAsia.Chicago:University
a linguistic ofChicagoPress.
Moreno, Martino Mario
1948 "L'azione del cuscito sul sistemamorfologico delle lingue semitiche
dell'Ethiopia",RassegnadiStudiEtiopici 7, 121-130.
Sandfeld,Kristian.
1930 Linguistique balkanique:problèmes Paris:Librairie
etrésultats. C. Klincksieck.
Thomason, SarahG.
1986 "On establishing externalcausesof languagechange",in: SoonjaChoiet al.
(eds.),Proceedings of theSecondEasternStatesConference on Linguistics,
243-251.Columbus: TheOhioStateUniversity.
1993 "On identifying thesourcesof creolestructures: Comments on thepapersby
Singler and Lefebvre", in: Salikoko Mufwene (ed.), Africanisms in Afro-

This content downloaded from 90.210.187.178 on Mon, 2 Jun 2014 03:53:38 AM

All use subject to JSTOR Terms and Conditions
 LINGUISTICAREASAND LANGUAGEHISTORY 327

American LanguageVarieties, 280-295.Athens, GA: University of Georgia

Press.
1997 "Onmechanisms in:StigEliassonandErnstHâkonJahr
ofinterference", (eds.),
Languageanditsecology:Essaysinmemory ofEinarHaugen,181-207.Ber-
lin:de Gruyter.
SarahG.,andTerrence
Thomason, Kaufman
1988 Languagecontact, andgenetic
creolization, linguistics.
Berkeley:University of
California
Press.
Wurm,Stefan
A.
1956 "Comment on question:
'Arethereareasofaffinitégrammaticale as wellas of
affinité
phonologiquecuttingacrossgenetic
boundaries?'",in:F. Norman (ed.),
Proceedings ofthe7tnInternationalCongress ofLinguists,450-452.London:
Permanent International
Committee ofLinguists,SectionB4.

This content downloaded from 90.210.187.178 on Mon, 2 Jun 2014 03:53:38 AM

All use subject to JSTOR Terms and Conditions