A Model of the Mechanisms of Language Extinction and Revitalization Strategies to Save

Endangered Languages
Author(s): CHRISANTHA FERNANDO, RHTTA-LIISA VALIJÄRVI and RICHARD A. GOLDSTEIN
Source: Human Biology, Vol. 82, No. 1 (February 2010), pp. 47-75
Published by: Wayne State University Press
Stable URL: http://www.jstor.org/stable/41466658 .
Accessed: 22/06/2014 02:38

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 [email protected].

Wayne State University Press is collaborating with JSTOR to digitize, preserve and extend access to Human
Biology.

http://www.jstor.org

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 02:38:55 AM

All use subject to JSTOR Terms and Conditions
A Model of the Mechanisms of Language Extinction and
Revitalization Strategies to Save Endangered Languages

CHRISANTHA 3
FERNANDO,'RIITTA-LIISA ANDRICHARD
VALIJÄRVI,4 A.
GOLDSTEIN1

AbstractWhyandhowhavelanguages diedout?Wehavedevised a math-

ematicalmodeltohelpus understand howlanguages go extinct.Weusethe
modeltoaskwhether language extinction canbe prevented inthefuture and
whyitmayhaveoccurred in thepast.A growing number ofmathematical
models oflanguage dynamics havebeendeveloped tostudy theconditions for
language coexistence anddeath, yettheir phenomenological approach com-
promises their toinfluence
ability language revitalizationpolicy.In contrast,
herewemodelthemechanisms underlying language competition andlookat
howthese mechanisms areinfluenced byspecific language inter-
revitalization
ventions,namely, privateinterventions toraisethestatus ofthelanguage and
thuspromote language learningathome,publicinterventions toincrease the
useoftheminority and
language, explicit ofthe
teaching minority language in
schools.Ourmodelreveals thatitis possible topreserve a minority language
butthatcontinued long-terminterventions willlikely be necessary.Weiden-
tifytheparameters thatdetermine which interventionswork bestunder certain
and
linguistic societal circumstances. In this the
way efficacy of interventions
ofvarious types canbeidentified andpredicted. Although therearequalitative
arguments fortheseparameter values(e.g.,theresponsiveness ofchildren to
learninga language as a functionoftheproportion ofconversations heardin
thatlanguage, therelative importance ofconversations heard inthefamily and
elsewhere,andtheamplification ofspoken toheard conversations ofthehigh-
statuslanguage becauseofthemedia), extensive quantitative dataarelacking
in thisfield.Weproposea wayto measure theseparameters, allowing our
model, as well as othersmodels in the to
field, be validated.

Languagesare culturally
transmittedsymboliccommunication systemsthatare
uniqueto humans(Jablonkaand Lamb 2005). Onlyhumanlanguageis produc-
unitscanbe combinedtoform
tive;thatis,inhumanlanguagesmallphonological
'National
Institute
for
Medical TheRidgeway,
Research, Mill London
Hill, NW7 1AA,United
Kingdom.
Collegium
Budapest for
Advanced
(Institute u.2,H-1014
Szenthâromsâg
Study), Budapest,
Hungary.
ofInformatics,
department ofSussex,
University Brighton, BN19QG,
Falmer United
Kingdom.
4School andEast
ofSlavonic Studies,
European University Gower
London,
College London
Street, WC1E
United
6BT, Kingdom.
Human ,February
Biology v.82,no.1,pp.47-75.
2010,
©2010
Copyright State
Wayne Press,
University Detroit, 48201-1309
Michigan
KEYWORDS:LANGUAGE LANGUAGE
REVITALIZATION, LAN-
COMPETITION,
ANDCOMPUTATIONAL
MATHEMATICAL
GUAGEEXTINCTION, MODEL,SOCIOLIN-
SECONDLANGUAGE
GUISTICS, BILINGUALISM.
ACQUISITION,

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 02:38:55 AM

All use subject to JSTOR Terms and Conditions
 ET AL.
48 / FERNANDO

newutterances in unlimited ways(Pinker1994). The originof languagehas not

been solvedyet, nor is it clearwhether therewas one protolanguage or whether
language was developedindependently on several continents (Christiansen and
Kirby2003).Itis wellknown, however, that new arise
languagesgenerally through
geographicisolationwhenformer dialectsbecomemutually unintelligible to the
speakersand whenlanguagesget combined, that is, when creolization occurs
(Thomason2001; ThomasonandKaufman1992).
In thispaper,however, we deal withlanguageextinction and variousmea-
suresthatcould be takento prevent it fromhappening. Thus we presentan an-
thropological studythatcan help us to understand the problemsof language
endangerment andlanguageextinction bothinthepastandin moderntimes.One
can assumethatlanguageshave alwaysbeen goingextinctas theconsequence
of environmental catastrophes (e.g., earthquakes, droughts, or floods),interven-
tionof newtoolsor weapons,development of agriculture, populationmovement
to new territories, or religiousor secularimperialism (Dixon 1998). Yet these
changeshave,accordingto Dixon (1998), historically had littleimpacton the
relativeequilibrium of languagesand theiroverallnumberbecauseconstant ex-
pansionand splitting ofpeoplesand languageshavecompensated fortheloss of
languages.
Now languagesaregoingextinct at an increasing ratelargelyas a resultof
colonization andglobalization, wherethelanguageoftheeconomically powerful
takesover(Mufwene2001,2004). In otherwords,themainreasonsforlanguage
endangerment todayare socioeconomic,political,and cultural(Campbelland
Muntzel1989; Fishman1991; Nettleand Romaine2000). Speakersof minority
languagesadoptthemajority languageso thattheirchildrenwillhavebetter job
prospects orbecause the minority language is simply not promoted in the society.
Some minority groupschoosenotto speaktheirlanguageforfearofpersecution.
Membersof otherminority groupssee theinvadingdominantcultureas more
appealing and modern and abandon theirtraditional cultureand language.These
trendshaveacceleratedwiththeriseofthenation-state andtheone language-one
stateideology(Dorian 1998) and withtheintroduction of theWesterneduca-
tionsystemand economies(Aikhenvald 2002). The globalization of culturethat
accompanies economic integration has led to English competing withnational
languages and endangering minority languages (Grenoble and Whaley 2006).
Preventing theloss of linguistic diversity is therefore a socioeconomicproblem
thatinvolveschangingtheattitudes ofspeakers.
Overthepasttwodecadesconcernhas beengrowingovertheloss of lin-
guisticdiversity (Crystal2000; Dalby 2003; GrenobleandWhaley1998,2006;
Harrison2007; Krauss 1992; Nettleand Romaine2000). It has been estimated
thatof the6,000 languagesspokenin theworld,as muchas 50% will become
extinctin thenearfuture (GrenobleandWhaley2006; Romaine2006b).
Despitesome contrary opinions(e.g., Malik 2000), thereasonsformain-
taininglinguistic diversity arenumerous. Hale (1992) notestheimportance oflin-
guisticdiversity tohumanintellectual lifenotonlyinproviding subjectmatter for

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 02:38:55 AM

All use subject to JSTOR Terms and Conditions
 LanguageExtinction
and Revitalization
Strategies/ 49

linguistsbutalso as formsof artistic expressionand culturalheritage.Different

languagesprovide us with botanical,biological,andgeographic information and
insight into human cognition(Harrison2007; Nettle and Romaine 2000). Most
important, however,linguisticdiversity can be considereda humanrightfrom
thespeakers'pointof view (Hill 2002). Whena languagebecomesendangered,
it loses notonlyspeakersbutalso domains(i.e., thesocial contextswherethe
languageis spoken)and becomesimpoverished and structurally simplified(Dal
Negro2004; UNESCO Ad Hoc ExpertGroupon Endangered Languages2003),
withheavyinfluencefromthelocallydominantlanguage(Clyne2003). Inter-
generational transmission is lost,andyoungerspeakersspeakthemoredominant
language.
Linguists,membersof endangered-language communities, governments,
nongovernmental organizations, andinternational organizations suchas UNESCO
and theEuropeanUnionare activelyworkingto save and stabilizeendangered
languages.This is done,forexample,by developinglinguisticdocumentation,
creatingorthographies, producingdictionaries and language-learning materials,
promoting positiveattitudes towardan endangeredlanguagebothoutsideand
withinthecommunity, planninglinguistic programs, andintroducing andenforc-
inglinguistic policies.In all thisactivity, a cleartheoretical distinctionis made
betweenwhatFishmancalls "reversing languageshift"(now generallyreferred
toas languagerevitalization) and"languagemaintenance" (Fishman1991,2001).
Language revitalizationefforts aim to increase the number of speakersof an en-
dangeredlanguage and to extend the use of the language to differentdomains,
whichrequiresa changein theattitudes of the speakersthemselves;language
maintenance, on theotherhand,refersto thesupportgivento languagesthatare
stillvitalbutthatneed to be protected fromoutsiders'attitudes (Grenobleand
Whaley 2006). In practiceboth language revitalization and languagemaintenance
are neededforthe survivalof a language.This is a complexand emotionally
chargedfield,withdisagreements abouthowbestto intervene to save low-status
languages from extinction.
The dynamicsof humancommunication, cooperation,and competition
has been modeledusingsocial evolutionmodels(Boyd and Richerson2005;
McElreathandBoyd2007). Althoughmodelinginthesocial sciencesis increas-
inglyinfluential (GilbertandTroitzsch2005), therealinfluence ofmathematical
modeling has been severely limited in the fieldof language This
revitalization.
is because modelshave notsoughtto engagewiththeintellectualframework
used by linguists;rather, theyimposean ecologicalor evena chemicalkinetics
metaphor that is not alwayshelpfulto linguists.Also, modelerslack thekind
of statisticaland formaldata thatwouldprovidethemwiththemeasurements
theyrequirefortheirmodelsto be validatedand therefore forthedynamicsof
language revitalization to be properly understood and for themodelsto have
predictivepower. This lack of data itself reflectsthe paucityof mechanistic
mathematical models- modelsthatcan identify theimportant parameters that
needto be measured.

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 02:38:55 AM

All use subject to JSTOR Terms and Conditions
 ET AL.
50 / FERNANDO

Ourpurposein thispaperis, first, to developa moresophisticatedmathe-

maticalmodelof languagecompetition and deathby introducing morerealistic
dynamicsand moreplausibleparameters. Second,by investigatingthemecha-
nismsunderlying languagerevitalization we can identify
efforts, theparameters
thatare crucialforpredictingthesuccessof a languagerevitalizationprogram.
We aim forourmodelto be usefulin languageplanningby showingtheeffects
ofpublicandprivateintervention strategieson an endangeredminority
language.
We discusssomeofthefactors thancan influence thissuccessandthushighlight
themostrelevant statistics
thatneedtobe measured.

Modeling Language Competition, Death, and Survival

The modelingof languagedeathand competition was made popularby
AbramsandStrogatz(2003). In theirmodel,speakerscan speakeitherlanguageX
orlanguageY, withtheinterconversion ofX andY speakersmodeledas a revers-
ible reactionwitha simpleinevitablelong-term behavior.Abramsand Strogatz
foundthatthelanguagewiththelowerstatusdeclinedto0 witha sigmoidaldecay
curve.Because themodelis purelyphenomenological ratherthanmechanistic,
it cannotaddresstheunderlying dynamicsof languageextinction, thatis, why
languageextinction is occurring and,moreimportant, howthedynamics couldbe
changed.Furthermore, Abramsand Strogatzsay thatthemodel"maybe useful
inthedesignandevaluationoflanguage-preservation programs" (2003: 900), yet
because of theirlack of mechanistic considerations, they are unable to suggest
anything moreexplicitthantheobviousadvantageofincreasing thestatusofthe
low-status language.
Abramsand Strogatz'smodelhas also beencriticized forphenomenologi-
cal orientation,its lack of realism,and its neglectof bilingualism and different
social spheres.A numberof different studieshaveattempted to addresssomeof
thesedeficiencies. Forinstance, bilingualism is commonin multilingual societies
andcan havean important influence on languageuse. In contrast to theassump-
tionsofAbramsandStrogatz, ifmorespeakersspeakthehigh-status language,it
does notautomatically followthatthelow-status language is lost,because speak-
erscan becomespeakersin one languagewithout loss ofproficiency in theother.
In an extensionofAbramsand Strogatz's(2003) paper,MinettandWang(2008)
exploredtheeffectof bilingualism (see also Wangand Minett2005). Although
theinclusionofbilingualism is an important addition, MinettandWang'smodel
retainsthe phenomenological notionof attractiveness, whichis a functionof
status,and "peak attractiveness" (a minormodification of theAbrams-Strogatz
model)- parameters thatcannotbe directly measuredand whosevaluescannot
be justified.
In theMinett- Wang model, attractivenessis a zero-sumgameso that
it is impossibleto makeone languagemoreattractive withoutmakingtheother
less so. Minettand Wangalso consideredattractiveness as dependingsolelyon
itspresencein thecommunity, disallowingmany forms of community interven-
tionsuchas schooling.Giventheseconstraints, MinettandWangconcurred with

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 02:38:55 AM

All use subject to JSTOR Terms and Conditions
 LanguageExtinction Strategies/ 5 1
and Revitalization

AbramsandStrogatzthatoneoftwocompeting languageswilleventually acquire

all thespeakers, irrespective oftheinitialconditions ortheexistenceofbilingual-
ism.It is notclearwhether thisresultis due to constraints on theirmodel(such
as thezero-sumnatureoflanguageattractiveness orthelimitedformsofpossible
interventions) or becauseof theparticular parameter valuestheyused.In theab-
sence of a stablemultilingual equilibrium, theyconsideredtheintervention of
simplyswappingthestatusof thetwolanguageswhenever thefrequency of the
lower-status languagedeclinedbelowa certainthreshold (30%) in thepopulation
ofspeakers.How sucha dramatic intervention couldbe achievedis notexplained.
In his paperon minority languagesurvival,Grin(1992) consideredthe
measurableindividualquantity: theamountof timea bilingualspeakerspeaks
one languageor another. This workemphasizedhowindividualmotivations and
decisionsinfluence languagesurvivalandhowthesedecisionscanbe settomaxi-
mizeutility. Grinobservedthepossiblepresenceofstablemultilingual equilibria,
witha criticalthreshold of minority languageuse belowwhichthelower-status
languageis lost. He showedhow interventions can be effective by modifying
speakers' calculations of utilityby, for instance,providing conditions in which
thelanguagecanbe used.Interestingly, Grinalso modeledtheroleofexpectations
of languagedeclineand foundthisto be a significant factorin intergenerational
transmission, a perception thatcan be remediedbyoutspoken support forthemi-
nority languageby authorities.
Castellöet al. (2007) and othershavecomplemented theAbrams-Strogatz
modelbytakingintoconsideration boththeroleofbilingualindividuals andsocial
structure in languagecompetition. In Castellöet al.'s model,thetwolanguages
haveequal status.Neitherbilingualism norsocial structure, whichwas modeled
as a small-worldnetwork (Watts1999), was sufficientto allow coexistence oftwo
languages.Bilingualagents could not form stablecommunities but disintegrated
intoone of themonolingual domains,and in a small-world systemtheyactually
acceleratedlanguagedeath.This relatesto theimportance of social networks in
and
minority endangered language maintenance: Without strong social links with
theoriginallanguagecommunity orwithout contactwiththespeakersofthesame
language,languageshiftis likelier(De Bot and Stoessel2002; Sallabank2007;
Stoessel2002).
Patriarcaand Leppänen(2004) extendedtheAbrams-Strogatz modelby
usingpopulationdynamics and reaction-diffusion equations. Their model pre-
dictsthattwocompeting languagescan coexistifthemainconcentrations oftheir
speakersarein twoseparategeographic areas,withdiffusion in theborderzones
andhigherstatusin one geographicarea.In otherwords,themodelpredictsthe
survivalofan endangered languagewithmatingsegregation. It correctly predicts
thesurvivalofa minority of a
language particular region if that minority language
is spokenas themajority languagein someotherregion,suchas thepersistence
ofSwedishinFinlandandofFinnishin Sweden.Although suchgeographic mod-
els can provideimportant few are
insights, regions completely isolated anymore
and mostendangered minority languagesare spokenwithinnation-states where

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 02:38:55 AM

All use subject to JSTOR Terms and Conditions
52 / FERNANDO
ET AL.

theydo nothavean officialstatus(Romaine2006a). In thesesituations, children

are educatedin themajority language,thelanguageofthemediais themajority
language,andtheminority languageis threatened despitethegeographic center.
Mostmodelsofbilingualism do notmakeanyassumptions aboutthenature
of thetwo languages.Mira and Paredes(2005) showedthata strongtendency
to bilingualism caused,forexample,by languagesimilarity (e.g., Galicianand
Castilian)can resultin a stablebilingualsituation. Staufferand Schultze(2005)
used a quasispecies-style stochasticbit-stringmodelof multipleequallyfitlan-
guagesandexaminedtherequirements forcoexistence, withlanguagemutations
p (childrenspeakinga slightly differentlanguageto theirparents)and language
transfer q resultingfromimitation. Theyobserveda phase transition in which
one languagedominatesin thespace ofp andq parameters. Thisprocessoccurs
on timescalesmuchslowerthanthatresponsible forthekindsof socioeconomic
languagedeclinewe considerhere.
A conceptually similaranalysistoourswasperformed byWickström (2005).
His theoryis based on a separation of theutility of variouslanguagescompared
withthedegreeofthespeakers'emotionalattachments. Wickström demonstrated
thatitis possibleforbilingualism tobe dynamically stable(eitherwithorwithout
monolingual speakersof bothlanguages)so longas neither languagehas much
greaterstatusthantheother.Variouscombinations of socialstatusandemotional
attachment (manifested in therateat whichchildrenof monolingual-bilingual
andbilingual-bilingual partnershipsareraisedbilingually, respectively)canresult
ineitherstableuniversal bilingualism,monolingualism-bilingualism coexistence,
or coexistenceof bothlanguageswithbilingualism. In theselasttwocases,it is
important thatchildren ofmixedmonolingual-bilingual parents havea sufficiently
highprobability of becomingbilingual.Wickström's analysiswas theoretical
ratherthanmechanistic, inasmuchas he did notconsiderthetimeevolutionof
languageuse,nordidhe modelthefactors thatwouldinfluence languagechoice.
As a result,Wickström could notconsiderhow interventions mightchangethe
dynamicsandtheresulting stablesituations.
Ourmodelis motivated bytheneedto createa mechanistic modelcharac-
terizedby parameters thatare measurablein thefield.We includebilingualism
as an integralpartof themodel,replacinggeneralfunctions thatdescribethe
probability of a childbeingmonolingual or bilingualwithmorespecificfunc-
tionsof thechild's familytypeand speakerfrequency thatallow us to simu-
late thedynamicsof the system.We includea factorthatarisesfromexplicit
(language-frequency-independent) languageteachingat school,and we distin-
guishbetweenlanguageheardin thepublicdomainand languageheardin the
privatedomain(language-frequency-dependent effects).We also explicitly con-
siderthekindsofintervention thatmightbe capableofpreserving theendangered
minority languageand explorea rangeof parameters. In addition,we presenta
phasediagramshowinghowstability depends on the parameters ofourmodel.In
agreement withWickström (2005),we also observethata strong tendency ofindi-
vidualsto learnbothlanguagescan resultin stablebilingualism. We also observe

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 02:38:55 AM

All use subject to JSTOR Terms and Conditions
 LanguageExtinction
and Revitalization
Strategies/ 53

monolingualism-bilingualismcoexistenceforthecorrectchoice of parameters,
the of
although possiblerange theseparameters is narrow.We do notobserve
any situationin whichcoexistencedependson theinitialvaluesof thevarious
languages;thereis no "threshold."
However,we do finda situationin whichthe
choice of whichlanguagedominatesin monolingualism dependson theinitial
ratioofhigh-statusandlow-status
languageuse.

Scope of Our Model

Our modelis a social evolutionmodelbased on differential equations.It
includesspeakersof a high-status language,speakersof a low-status language,
and bilingualspeakersof boththeselanguages.For thepurposesof ourmodel,
we definebilingualism as theabilitytofunction confidently intwolanguages,that
is, theabilityto havecommunicative competencein twolanguages.In practice,
bilingualism is notas simpleas this.For example,an individualmaybe able to
read,write,or speakaboutcertainthingsonlyin one of thelanguages,use one
of thelanguagesmore,havea onlypassingknowledgeof one of thelanguages,
or havelearnedone of thelanguagesas an adultand theotheras a child(Butler
and Hakuta2006; Edwards2006). Our bilingualspeakerscoverbothsuccessive
andsimultaneous bilingualspeakers, thatis,thosewholearnedone languagelater
or learnedbothlanguagessimultaneously. Theydo notneedto be idealizedfully
balancedbilingualspeakers.In otherwords,theycan be moreproficient inone of
thelanguages,althoughtheymayfunction well in both.Notethatthisdefinition
differs fromthemodelofWickström (2005: 83), whereitwas explicitly statedthat
monolinguals "may be able to communicate very well in the otherlanguage, but
notwiththeease orcomfort ofa nativespeaker."
Ourmodeltakesintoconsideration notonlyintergenerational transmission
[i.e.,parentsteaching theirlanguage(s) to theirchildren] but also horizontaltrans-
mission(i.e.,languageacquiredoutsidethehomeandlearnedformally inschool).
Thus,unlikepreviousmodels,forexample,Wickström's (2005) model,we have
includedtwodomainsin ourmodel:a publicdomainand a privatedomain.This
enablesus to takeintoconsideration thelinguistic realityin anygivenbilingual
or One
society community: language is used in the privatedomain(e.g.,athome,
withfriends, in personalletters), andtheotheris usedin thepublicdomain(e.g.,
in thecity,at work,in education,in government, in media).This societalbilin-
gualisminvolving a low-status and a high-status language,ora low andhighreg-
ister,is referredto as diglossia(Ferguson1959; Romaine2006b).The use ofthe
endangered low-status languageis oftenrestricted totheprivatedomain,whereas
a high-status is
language spoken in the public domains.
Because our goal is to showtheeffectsof languagerevitalization efforts
on a low-status endangered language, we have included three different typesof
intervention measuresin our model.The threestrategies are (1) increasingthe
perceivedstatusofthelow-status languageso thatbilingualfamilieswillchoose
to teach the low-statuslanguageto theirchildrenand childrenwill be more

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 02:38:55 AM

All use subject to JSTOR Terms and Conditions
 ET AL.
54 / FERNANDO

motivated to use it;(2) increasing theamountofthelow-status languageheardin

society,thusincreasing exposureto thelanguageand facilitating thelearningof
thelanguageas well as raisingitsstatus;and(3) formallanguageteachingofthe
low-status languageto childrenwhowouldotherwise speakonlythehigh-status
language.
In practice, increasing theperceivedstatusofthelow-status languagerefers
to awareness-raising programssuchas providing promotional materialsto par-
entshighlighting thenumerous benefits ofspeakingthelow-status languageorof
beingbilingual.In Walesleafletsgivento bilingualparentshavebeensuccessful
in highlighting thebenefitsof beingbilingualin Welshand English(Edwards
and Pritchard-Newcombe 2005). Oftentheinterest linguistshave shownin and
theworktheyhaveconductedon theindigenous languagehaveled to a linguistic
and culturalrevival(Bradley2002). Local languageactivistscan also achieve
a changein attitudes (Florey2008), as can a changein nationalattitudes and
policies(Grenobleand Whaley2006; Wurm2002). This typeof measureaims
to secureintergenerational mothertonguetransmission, whichFishman(1991)
identifies as thekeytoreversing languageshift.
Increasingtheamountof thelow-status languageheardin societytargets
thepublicsphere,thatis, thehigherdomains.This typeof intervention can in-
clude radioand televisionbroadcasts,newspapersand otherpublications being
printedin thelow-status language,development of specializedterminologies in
fieldssuch as technology or commerceforthatlanguage,standardization, and
orthography andliteracy development (GrenobleandWhaley2006).
The thirdintervention involvesteachingin a formalsetting(such as a
school)thelow-status languageto childrenwhowouldotherwise speakonlythe
high-status language. Previous teaching efforts have concentrated on teachingthe
minority language in schools as a foreign or second to
language semispeakers or
nonnative speakers, or, more frequently, teaching it to bilingualspeakers of the
minority languageas a mother tongue.Accordingto Ö Riagâin(2001), thestable
ratesof bilingualism in Irelandare caused by successfullanguageteachingin
schools,notby intergenerational transmission amongtheoriginalIrish-speaking
community. the
Naturally, fluency ofthese new bilingualspeakersvariesandthere
is boundto be interference fromtheirdominant languagein theformally learned
language. Total immersion programs, in which all teaching is conducted in the
local language,can lead to a highlevelof fluencyand subsequentintergenera-
tionaltransmission. The mostfamousimmersion programsare theMaori lan-
guage nest, Te Köhanga Reo (King2001), and the Hawaiian languagenests,'Aha
PünanaLeo (Warner 2001). Educational are
programs costly, becausetheyrequire
skilledteachersand thedevelopment of teachingmaterials, and theyoftenpres-
entseriouslogisticalproblems(e.g.,Aikhenvald A
2003). perhapsless costlybut
similarintervention is theMaster- Apprentice Programdevelopedin California
in 1992; thisprograminvolvesan older,morefluentspeakerof theendangered
languageinteracting witha youngersemifluent speakerornonspeaker ofthelan-
as in
guage theyengage everyday activities(Hinton2001; Hinton et al. 2002).

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 02:38:55 AM

All use subject to JSTOR Terms and Conditions
 LanguageExtinction Strategies/ 55
and Revitalization

Herewe modelspecifically theteachingofthelow-status languagetothose

childrenwhopreviously spokeonlythehigh-status language,thatis, conversion
tobilingualism fromthecommunity ofhigh-status languagespeakers.We do not
modelsemispeakers and so cannotrenderthefullscope of theinterventions dis-
cussedin theliterature.Giventhislimitation, teachingofthelow-status language
to bilingualspeakerswouldbe includedas interventions of thefirsttwo types
to theextentthatit increasesboththestatusof thelow-status languageand the
exposureto low-status languageconversations. A component of teachingindeed
worksby thiseffect. The reasonwe havea distinct thirdintervention is thatwe
wishtoalso describethosemathematical effects ofteaching thatdo notdependon
thefrequency oflow-status languageconversations heard.
The model is used to studythe conditionsunderwhichthe low-status
languageand/orbilingualism is stable(i.e., has a nonzerosteady-state equilib-
riumfrequency) and how thisstability dependson thenatureand extentof any
intervention.
We havechosenthesethreeparticular measuresbecausetheyrelateto the
use of theendangered language in theprivate and publicspheres,whichhas not
been modeledextensively before.Furthermore, thethreemeasurescoverother,
moredetailedmeasuressuggestedin theliterature (Crystal2000; Fishman1991,
2001; GrenobleandWhaley2006); forexample,literacy, use oftechnology, and
use ofthelanguageingovernment areincludedinintervention type 2. Other mea-
sures,suchas increasingthewealthof thelow-status languagespeakers,could
havebeenhighlighted or chosen,butthiswouldhaveled to a morecomplicated
modelthatis outofthescopeofthispaper.Some degreeofsimplification is nec-
in
essary any mathematical model. Additional revitalizationmeasures and their
to
relationship ourmodel and the factorsthat we have chosen are discussed in the
"ConclusionandDiscussion"section.

The Model
We assumethatthereare threetypesof speakers:monolingualspeakers
of thehigh-status language(//),monolingual speakersof a low-status language
(L), and B
bilingual( ) speakers.Speakers mate witheach otherandhavechildren
whospeakatleastone language.Speakerstendto matewithspeakersofthesame
type,although bilingualspeakerscan matewithanyone,resulting in fivetypesof
families:HH, HB, LL, LB, andBB. Childrenwho growup in one of thesefami-
lies learnto speakone or bothof thelanguagesdependingon thelanguagesof
theparentsandthefrequency ofconversations thatthechildhearsin each ofthe
languages, bothinside and outsidethe home. The probability thata childofHH,
HB, or BB parentslearnsL dependson theamountofL theyareexposedto and
theirsusceptibility to L, parameterized by aL (see Table 1). A similarparameter
the
aH represents susceptibility of childrenfromLL, LB, andBB parentsto learn
H as a functionoftheamountofH intheirenvironment (see Table 1). Differences
in socialstatusmaymakeaH significantly larger thanaL.

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 02:38:55 AM

All use subject to JSTOR Terms and Conditions
 56 / FERNANDO
ET AL.

Table1. Assumptions
oftheModel:ChoiceofParameter
ValuesandtheFormofthe
Equations3
Plausible
Parameter Range Justification DatatoBeAcquired
Initial
fractions
ofspeakers 0-1 Theoutcome ofthemodel Proportions ofspeakers ofL
ofL andH depends ontheinitial andH.
concentration ofspeak-
ers.Inanendangered
language situation,
theinitial number ofL
speakers isoftenlow.
aL: Responsivenessofa 1-1.5 Thisfigure islowbecause Theprobability thata child
child born toHHorHB (seeFigure1) a childborn toHHor borntoHHorHBparents
parents tospeakL asa HBparents isunlikely becomes anL speaker in
function ofL conversa- tospeak L ina typical various environments with
tions heard endangered language L frequency.
different
situation.
aH: Responsivenessofa 1-3 Thisfigure ishigh inour Theprobability thata child
child born toLLorLB (seeFigure1) model, representing the born toLLorLBparents
parents tospeakH asa enhanced of
desirability becomes anH speaker in
function ofH conversa- speaking H. variousenvironments with
tions heard different
H frequency.
Ratio
77: oftheeffectof 1 We are assuming thata Theaverage percentage ofL
family andnonfamily child receives 50%of andH conversations heard
conversations heard
by herlinguistic inputat bythechild both athome
child onthelanguage home and50%outside andoutside thehome,
spoken bythechild thehome. combined with thefraction
ofchildren thatareL,H,
orB.
u:Amplification factor
of 1-3 Although wewould expectTherelative amount ofL
H conversationsspoken H to dominate public andH heard inthepublic
inthecommunity toH sources, wewould sphere,excluding and
conversationsheardfrom expect that thelanguage including public announce-
allpublic sources exposure duetothese ments,media, etc.
sources isnotmuch
greaterthan thatdueto
spoken conversations.
A:Fraction ofallconversa- 0-0.1 Thisfigure isconsidered Longitudinal dataonthe
tions heardbya childas lowinourmodel, butit increased exposure to
a resultofgovernment could behigher inthe theminority language in
intervention presence ofsuccess- thepublic sphere andits
fulliteracy ormedia influenceonnumber of
projects. speakers
mHB : Rateatwhichteaching 0-0.1 Thisfigure isassumed to Longitudinal dataonthesuc-
converts H children
toB belowinourmodel, cessrate fordifferenttypes
relativetothereplacement butitcould behigher ofprograms where the
rate ofthepopulation with successful teaching second language learned is
programs. a minority language.
//,
High-status L,low-status
language; language;B,bilingual.
a. Inthis table
wesummarize thearguments forourchoice ofparameter values
andsuggest furtherexperi-
ments forfield tousetoconfirm
linguists theseparameter values.

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 02:38:55 AM

All use subject to JSTOR Terms and Conditions
 LanguageExtinction
and Revitalization
Strategies/ 57

The amountof each languageheardby thechildhas twocomponents be-

cause oftheinfluence ofthefamilyandtheinfluence ofthecommunity, where77
is a parameter in ourmodelthatrepresents thestrength ofthefamilycontribution
relativeto thecontribution fromthecommunity (see Table 1). This factoralso
includestheeffectiveness of thefamilies'discoursein influencing thelanguage
of thechild.The amountof H and L spokenat homedependson thelanguages
spokenbytheparents.The amountofH andL spokenin societydependson the
number ofH, L, andB speakers.We also considerthat,becauseofa difference in
status,H mightbe givenmoreprominence. As an example,H mightbe overrep-
resented in themassmedia.We introduce another parameter, u, whichrepresents
theextraprominence ofH overL (see Table 1). u amplifies theconversations of
theH typeso thatforthesamenumber ofH conversations spoken,relatively more
areactuallyheardbythechild.
We can now considerthreedifferent strategiesforpublicintervention in
theinterest of maintaining thepresenceof L in thesociety,eitherin theformof
L or B speakers.The threestrategies, introduced in theprevioussection,are (1)
promoting thelearningof L by raisingitsperceivedstatus,thusencouraging its
learning bychildren inHH, HB, andBB families;(2) usinggovernment programs
to increasetheamountofL heardin society;and (3) formally teachingL to chil-
drenwhowouldotherwise speakonlyH. Promoting thelearning ofL inthehome
is represented in ourmodelbyincreasing thevalueofaL so thatchildren exposed
to a givenamountof L wouldhave an increasedmotivation forlearningL. We
modelgovernmental intervention
programs to increasepublicdiscourseinL with
an additionalparameter A,whichrepresents thefraction ofall discourseheardby
thechildas a resultofthisgovernment intervention (see Table 1).
The thirdapproach,directteachingof L to schoolchildren, involves"con-
verting" H children to B children.mHB is the rateat which teachingconverts H
children toB relativeto thereplacement rateofthepopulation(see Table 1).
The population dynamicsarerepresented witha standard population genet-
ics modelofthree-allele one-locusselection, withnonrandom mating(H speakers
cannotmatewithL speakers)and a non-Mendelian methodof determining off-
springtypes based on existingpopulation frequencies, a Lamarckian processthat
is inappropriate forgeneticinheritance butlegitimate forlanguageinheritance
(McElreathandBoyd2007). The dynamicvariablesin themodelaretherelative
populations ofH , L, andB speakers.
The parameters thatdetermine thedynamicsin our modelare shownin
Table 1. We showtheparameters alongwiththeirvalues,justification fortheir
values,andnumerical datathatneedtobe acquiredfromminority andendangered
languagecontexts to confirm theirvalueswherenecessary.
We briefly discussthejustification fortheparameter values.Understanding
aL and at
aH requireslooking Figure 1. The higher these values arecomparedto
1,themorelikelyitis fora childtobecomebilingualas a function ofthepropor-
tionofconversations heardin thelanguagethatis notsharedbybothparents.aH
is largerthanaL because we assumethatchildrendo notneedto hearas many

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 02:38:55 AM

All use subject to JSTOR Terms and Conditions
 ET AL.
58 / FERNANDO

values
(6)and(8)forvarious
1. Equations
Figure a = 0.1,0.5,1.0,2.0,and3.0.
ofa, including

conversations in thehigh-status languagein orderto learnit as theydo to learn

a low-status language.Currently no experimental data showhow theseparam-
etersshouldbe set,butourresultsare robustfora widerangeof a values.The
7]valueof 0.5 assumesthathalfthechild'sconversations areheardin thepublic
domainand halfin theprivatedomain.This is a neutralassumption, and ifem-
piricalestimation of this value is made, our model can be appropriatelyfurther
constrained, u is an amplification factorthatconverts conversations spokeninto
conversations heard.We assumethatforthehigh-status languagethisamplifica-
tionfactorliesbetween1 and3. The resultsofthemodelarerobustwithinatleast
thisrange.A further modifiestheratiosofconversations describedby 77.It is the
ratioofconversations thatthechildis exposedto as a resultofgovernment inter-
ventionversustheconversation frequency heard without government intervention
in thecommunity. mHB is theconversation-frequency-independent component of
second-language learning resulting fromteaching. Both À and mHBare assumed to
be less than10% oftheproportion ofspeakersproducedbyverticaltransmission.
We nowshowhowthelanguagecompetition situation justdescribedis rep-
resentedmathematically. The modelis usedto studytheconditions in whichthe
low-status language and/or bilingualism is stable(i.e., has a nonzerosteady-state
frequency in the population of speakers). ("Steady state" is a technicaltermthat

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 02:38:55 AM

All use subject to JSTOR Terms and Conditions
 LanguageExtinction Strategies/ 59
and Revitalization

meansthefinalvaluesobtainedifthemodelis runfora sufficiently longtimefor

thevaluesto stopchanging.)
As statedearlier,ourmodelassumesthattherearethreetypesof speakers:
monolingual language( H), monolingual
speakersofthehigh-status speakersofa
low-status B The
language(L), andbilingual( ) speakers. proportions of speakers
in thepopulationaregivenbythevariablespwpL,andpB, respectively.
Next we wish to obtainequationsfortheproportion of different couple
as a function
types(i.e.,parents)inthepopulation oftheproportion ofspeakersin
thepopulation. We assumethatspeakersmatewitheach otherandhavechildren
whospeakatleastone language.Speakerstendto matewithspeakersofthesame
type,althoughbilingualspeakerscan matewithanyone.The fraction of
p(XfXy.)
all couplesthataretypeXtXpwhereX e {//,L , B } (i.e., whereX can standforH,
L, orB), is givenby

P(HH) = P2h+PhPl, (1)

P(HB) = 2PHPB, (2)

p(LL) = p2L+pHpL, (3)

p(LB) = 2pLpB, (4)

P(BB) = P2B. (5)

These equationsfulfillthe appropriate conditionsthat(1) thetotalnumberof

speakers of each language in each couplerepresents theoverallfraction of total
+
speakers[e.g., p(HH) 'p(HB) = pH] and (2) in the absence of bilingualism
lp(B) = 0], thefraction of monolingualspeakersspeakingX is thesame as the
fractionofX speakers.
Nextwe wishto obtainequationsfortheprobability thata childwillspeak
H orLot be B, as a function ofthetypeofparents. We assumethatchildren who
growup in one of these families learn to speak one or both of theselanguages
depending on thelanguagesoftheparents(XfXy) andthefrequencies ofconversa-
tionsC(Y I XtXj)thatthechildhearsin eachofthelanguages(Ye {//,L}), which
also dependson thelanguagesspokenbytheparentsandthefrequencies ofthese
languages in thesurrounding society. [C(Y I XX^) is standard probabilitynotation
fora conditional probability(i.e., theprobability of something happening, given
thatsomething else has alreadyhappened).It shouldbe readas theprobability
of conversations thechildhearsspokenin languagetypeY giventhatthechild
is theoffspringofcoupletypeXfX..]We willdescribeequationsto determine the

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 02:38:55 AM

All use subject to JSTOR Terms and Conditions
60 / FERNANDO
ET AL.

frequencies ofconversationslater.We assumethatthechildren can alwaysspeak

to bothparents,so thata childin an HH or HB familycan speakH or be B, but
notonlyspeakL. Considera childofHH orHB parents.She willdefinitely speak
H butwillalso speakL (i.e., willbe B) witha probability
p(B'HX) (X G {H, B })
thatincreasesfrom0 to 1 as thefraction of languagesthatshehearsspokenin L
increasesfrom0 to 1.A flexiblefunction thatdescribestheseassumptionsmathe-
maticallyis

.
P(B'HX) = qllC(L'HX
L--J :) . (6)
1 l + (aL-l)C(L'HX)

aL measurestheeffectiveness of hearinglanguageL in motivating its learning

(i.e.,thereceptiveness
ofthechildtoL); ifaL is small,thechildmustbe exposed
toa significant
amountofL beforesheis likelytolearnit,whereasa largevalueof
aL meansthata childwillbe motivated tolearnL evenifconversationsinthislan-
guagerepresent onlya smallfractionofthosethatshehears.aL represents,among
otherthings,the"status"ofL, wherestatusis usedtomeantheentireconstellation
ofsocietalfactorsthatmotivatethelearning ofa givenlanguage.
ThecorrespondingprobabilitythatthechildwillspeakonlyH is thengiven
by 1 minusthisamountofL, becausethesearetheonlytwoprobabilities; there-
foretheymustadd up to 1.

miMm l~C(L'HX^ m
' ' + (aL-X)C(L'HX)

We havesimilarfunctions
forLX couples:

TO- «"«"I"',
- ■ (8)
1+ (aH ')C(H'LX)

WIÜQ- . (9)
l + (a„-l)C(H'LX)

wherewe do notassumethataH = aL. In particular, differencesin social status

maymakeaH significantly than
larger aL. We will describethe effectsof social
statuslater.
The bestwayto understandthesefunctionsis to visualizethem,as in Fig-
ure1,whichshowstheproportion ofchildrenwhobecomebilingualas a function
oftheproportion
ofconversationsheardinthelanguagethatis notsharedbyboth
parents.
Now considerthemorecomplicatedcase of BB parentalpairs.Here the
childrenmaybe one of threetypes:speakersofH or ofL, or B. We assumethat
parentsmay omitto teachone of the languages.We assumethatthe relative

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 02:38:55 AM

All use subject to JSTOR Terms and Conditions
 Strategies/ 61
and Revitalization
LanguageExtinction

fraction children(i.e., H and B) who also speakL is givenby an

of //-speaking
to
analogous Eq. (6):
expression

p(B'BB) = aLC(L IBB)

p(B IBB) + (//1BB) 1+ (aL - 1)C(L I BB)
The relativenumberof L speakerswho also speakH is givenby an expression
analogousto Eq. (8). Thisresultsin

p(H'BB) = KX~C{UBB' (11)

aLC(L I BB)

C^H '
p(L IBB) - Kl~ BB' (12)
olhC(HIBB)

^BiBm=KJiz£rnm+izmm+1''
a„C(H'BB) aLC(L I BB) j

Figure2 showstheseequations.The probability ofa childbeingoftype//,

L , or ifbornto /?/? parentsdependson theprobability of hearingL conversa-
tionsin theenvironment C(L) and on thevaluesof aH and aL. If receptiveness
is equal to bothlanguageswith{ aH,, aL } = {0.5, 0.5}, thenone obtainsthesolid
curves.If [aH, aL } = {3.0, 1.0} so thattheresponsiveness to H is threetimesthe
responsiveness ofL, one obtains the dashedcurves. The difference can be seenif
one considersthesituation in whichC(L) = 0.5, thatis, where50% oftheconver-
sationsheardarein L. With{ aH, aL] = {0.5, 0.5}, a childbornto BB parentshas
a 40% chanceof speaking//,a 40% chanceof speakingL, and a 20% chanceof
beingB. With{aw aL } = {3.0, 1.0}, atthesamevalueofC(L) thechildhas a 43%
chanceofspeaking//,a 14% chanceofspeakingL, anda 43% chanceofbeingB .
Now we wishto obtainequationsforthefrequency ofconversations heard
by thechild in the environment as a functionof thefrequency of conversations
actuallyspoken.The amountof each languageheardby thechildP(Y I XfXy.) has
twocomponents becauseof theinfluence of thefamilyand theinfluence of the
community:

C(Y) = 7? I XX.) + (l-r,)C_ly(F),

Cfamily(r (14)

where77representsthestrength ofthefamilycontribution relativetothecontribu-

tionfromthecommunity, (1 - 77).This factoralso includestheeffectiveness
of
thefamilies'discoursein influencing thelanguageof thechild.The speechin
thefamily, is an explicitfunction
Cfamily(yiXfXy.), of thelanguagecategoryof the
parents(X^p.

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 02:38:55 AM

All use subject to JSTOR Terms and Conditions
 ET AL.
62 / FERNANDO

2. PlotofEqs.(11)-(13)showing
Figure P(H'BB)(graycurves, tobottom-right),
top-left P(L'BB)
(graycurves,bottom-left andP(B'BB
totop-right), ) (black asa function
curves), ofC(L)
for{aw,aL} = {0.5,0.5}(solid
curves),
{2.0,2.0}(dotted and{3.0,1.0}(dashed
curves),
curves).

We first considerthecommunity distribution

oflanguages.We assumethat
conversations aredyadicandthattheprobabilitiesofa conversation betweentwo
peoplearegivenbyexpressions identicaltothoseofEqs. (l)-(7) (i.e.,thatconver-
sationsbetweenpairsis of equal ratioto matingbetweenpairs).Two interacting
peoplehaveonlyone languageincommon,whichtheyspeak,withtheexception
ofconversations betweentwobilingualspeakers,who,we assume,speakequally
frequentlyin eitherlanguage.Thatis,

(^0 = (Ph PhPl) ([^PhPb}~^iPb

^community ( ^)

witha similarexpressionforCcommunity(L).
We makeanothersmallmodification to theseequationsto considerthat,
in status,H mightbe givenmoreprominence.
because of a difference (As an
example,H mightbe overrepresented in themassmedia.)We introduce another
parameter, theextraprominence
cj,representing ofH overL.

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 02:38:55 AM

All use subject to JSTOR Terms and Conditions
 Strategies/ 63
and Revitalization
LanguageExtinction

(^0 = ( Ph + PhPl "I"^ PhPb+ 2

^community )> 0 ^)
~ß

^community pI+PhPl +2PLPb +ïPI)> ^

term:
whereß is a normalization

^ = l + (o;-l)C_ity(//). (18)

theconversations
lj amplifies ofH typeso thatforthesamenumber ofH conver-
sationsspoken,relativelymoreareactuallyheardbya child.
Forthefamilycontribution, we considerthatHH familiesspeakonlyH at
home,whereasLL familiesspeakonlyL. ComparedwithHH families, we would
expectHB familiesto exposetheirchildren to moreL, especiallywhenthebilin-
gual parentspeaksH as a secondlanguage.The presenceof L in suchfamilies
mightdependon therelativestatusofthetwolanguagesandon theirfrequency in
thesurrounding culture.In thismodelwe makethesimpleapproximation thatHB
familiesspeakH five-sixths ofthetime,representing thatconversations couldin-
volveoneorbothparentsandthatonlya fraction ofthoseconversationsinvolving
onlytheB parentcouldoccurinL. We mightexpectthatthefraction ofH spoken
in LB familiesmightbe largerbecauseof thehigherstatusof H , butwe use the
simpleapproximation thatL will be used five-sixthsof thetime.Again,some-
whatsimplistically, we assumethatBB familiesspeak bothlanguagesequally
frequently.This resultsfromtheassumption thathalfof B speakersare primar-
L
ily speakers and thathalf of B speakers are primarily H speakers.Modifica-
tionsofthesefractions affectthespecificresultsofthemodelbutnotthegeneral
conclusions.
whatwe havedone is to modifya standardpopulationgenet-
Effectively,
ics modelof three-alleleone-locusselectionwithnonrandom mating(bilingual
speakers can mate with everyone) and a non-Mendelian methodof determining
offspringtypes based on existingpopulationfrequencies (McElreath and Boyd
2007). Assumingthatthenumberof childrendoes notdependon thelanguage
spokenbytheparents, equationsfor
we arriveat thefollowingsetof differential
'
thechangeoflanguage(allele) frequencies overtime.pK is therateofchangeof
pKovertime:

- IP* ïoxK = (19)

Pk = i 1W }•
.M
wherethesumis overall typesof familieswith andX-takingvaluesfromthe
rangeK and7 is thebirthrate.To maintainpH+ pB+pL= so thatp can properly
as theproportion
be interpreted ofeach speakertypeinthecommunity, we setthe
oftheformj pK.
in a loss ofpopulation
deathrateequal tothebirthrate,resulting

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 02:38:55 AM

All use subject to JSTOR Terms and Conditions
 ET AL.
64 / FERNANDO

Thistermeffectively imposestheselectionpressureon speakertypes;growth of

onetypeofspeakermeansthedeclineoftheothertwotypesofspeaker.The equi-
libriumconditions obtainedin the"Results"sectionarebynumerical simulation
ofthegivenequations.
To reiterate,
thedynamicvariablesinthemodelaretherelativepopulations
ofH , L, andB speakers,givenbypH,pL,andpB,respectively. The parameters in
themodelare(1) theeffectiveness ofeachlanguageinmotivating speakers,given
as aH andaL; (2) therelativecontributions
offamilyandsocietyto thelanguages
spokenby thechildren, parameterized by 77;(3) thebias in theprominence ofH
comparedto L, givenby u; and (4) thereplication rate7, which,so longas it is
small,willnotaffect
sufficiently thesteady-state concentrations.
Statusis represented
as followsin ourmodel.The factthatH is thehigher-
statuslanguageis represented byaH beinggreater thanaL (i.e.,children aremore
motivated to learnH) andbyu beinggreater than1 (i.e., a greater
proportion of
sentencesspokeninH areheardbychildren thansentencesspokeninL).

Modeling Intervention
We can now considerthreedifferent forpublicintervention
strategies in
theinterestofmaintaining thepresenceofL inthesociety, eitherin theformofL
speakersorB speakers.Thethreestrategies are(1) promotingthespeakingofL by
raisingitsperceivedstatus,(2) usinggovernment to
programs increasetheamount
ofL heardin society,and(3) formalteachingofL to children whospeakH.
How do we describeintervention 1 in ourmodel?Promoting thelearning
ofL in thehomeis represented in ourmodelbyincreasing thevalueofaL so that
childrenexposedto a givenamountof L will have an increasedmotivation for
learningL.
How do we describeintervention 2 in ourmodel?We assumethatgovern-
mentprograms act to increasetheamountof L perceivedby languagelearners,
thusincreasing To modelthiseffect,
Ccommunity(L). we consider

C(Y) = (K) + (1- r?- A)Ccommuni(y

IXX ) + ACgoverament (7), (20)

where,fora governmental program = 1 and

toenhanceL, Cgovemment(L) CgmemmeJH)
= 0. Thatis, we introducea government-mediated termintotheequationthatde-
scribestheproportion of languageconversations heard.À representsthestrength
oftheintervention andis thefraction
ofall discourseheardbythechildthatis due
to thisgovernment intervention.
How do we describeintervention3 inourmodel?Thethirdapproachwould
"convert"H children toB children.Thiscan be modeledas a modification ofEq.
(19) to

Ph = 7 J2p(X.X^P(H 1W -7 P„ - imHBpH, (21)

. <«>

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 02:38:55 AM

All use subject to JSTOR Terms and Conditions
 LanguageExtinction
and Revitalization
Strategies/ 65

Pl = 7 £>(X,X )/U I W -7Pr (22)

. <#>

-7Pfl (23)
. «0

where is therateat whichteachingconverts H childrento B relativeto the

replacement rateofthepopulation. Thiseffect is independentofthefrequency of
languagespeakersin thecommunity.
Intervention 3 is specificallytargetedat nonnative speakersof L. A more
typicalintervention,systematic teachingofL toexisting bilingualandsemispeak-
ersof L, cannotbe explicitly includedbecausewe do notmodeldegreesof lan-
guagefluency. Thisintervention,however, can be modeledas bothan increasein
theamountthesechildren areexposedtoL [intervention 2, increasingCcommunity(L)]
andan increasein theperceivedstatusofL in thesecommunities (intervention1,
modeledas an increasein aL).

Results

Intergenerational Transmission. We firstconsiderthesituationwithoutin-

tervention in thepublicsphere,shownin Figures3A and 3B. Typicaltemporal
dynamicsare shownin Figures3A and 3B, starting froma populationof 50% L
speakersand 50% H speakers,wherethereis a statusdifferential favoringlan-
guageH (i.e., aL = 1, aH = 3, and uj = 2). As can be seen,L goes extinctovera
relativelyshortperiodoftime.The mechanism forthischangeis thetendency of
childrenof L speakersto be B. This allowsthemto matewithotherB speakers
as well as withH speakers,witha highlikelihoodof havingoffspring who can
speakonlyH.
Figures3C and 3D demonstrate theresultof publicintervention at year
100 duringtheperiodof languagedecay.The intervention we modelis a mixed
one thatencourageschildrento be morereceptiveto L at homeby settingaL =
1.5,whereL is heardmorein public(A = 0.1) and wheregovernment programs
encourageteaching(mHB= 0.1), representing thatapproximately 10% of the
//-speaking childrenlearnL. At year300 theintervention is removedand the
variousparameters returnto theirinitialvalues.As can be seen,L is rapidly
eliminated fromthepopulation, indicatingthatitspreservation dependson con-
tinuousintervention.
Figure4 exploreshowthefinding thatL is lost(withoutintervention) de-
pendson theinitialconditions (i.e.,theinitialproportionsofspeakersof//,L, and
B in thecommunity). A ternary plotis used,whereeach pointinsidethetriangle
corresponds to a triplet
of numbers thatgivestheinitialproportions of//,L, and

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 02:38:55 AM

All use subject to JSTOR Terms and Conditions
 ET AL.
66 / FERNANDO

3. Language
Figure dynamics fora low-status
languageL competing witha higher-status
language
H.Thisdifference ofstatus byaH= 3,aL= 1,andu = 2.Theinitial
isrepresented popula-
tionstartswith50%H speakers and50%L speakers. PartsA andC show changesinthe
fraction
ofthepopulation speakingH (grayarea),L (stippledarea),andboth(hatched
area).PartsB andD show therelative
fraction
ofcouples thatareHH(solidgray area),
HB(gray-hatched BB(white-hatched
area), LB(white
areas), stippledandhatchedarea),
andLL (white stippledarea).PartsC andD showtheeffect ofa mixed governmental
intervention
program startingatyear100consistingofa government intervention
toen-
courage more spoken L (A= 0.1),with formal teaching(mHB = 0.1)andencouraging
learningathome (aLincreased to1.5).Wehaveuseda valueof7 = 0.01,resulting ina
biological"generationtime" of100simulation Ifweassume
steps. thata humangenera-
tionisapproximately 25years, thenthismeans that
eachsimulation to
stepcorresponds
approximatelyone-fourth year.Theresult
isthestablemaintenanceofL,primarilyamong
bilingualspeakers,solongas theinterventioncontinues.
When theintervention
endsat
year300,L isquickly lostinthepopulation.

B speakersinthecommunity. We see thatonlyifthecommunity startswitha high

proportionofL does H notinvade,dominate, andeliminateL.
Figure5 demonstratesthat,in general,bilingualism
is unstablewithout in-
terventionforlow valuesof aL. The shadingsshowwhichkindof languageper-
sistsatequilibrium
fordifferentparameter values.Stippling
showsthatL persists,
grayshadingthatH persists, and hatching thatB persists.It is possibleto have

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 02:38:55 AM

All use subject to JSTOR Terms and Conditions
 LanguageExtinction
and Revitalization
Strategies/ 67

4. Ternary
Figure plotshowing howtheresulting
finalpopulationdepends ontheinitial
popula-
tion.
Thelocationinthegraph theinitial
represents ofH,L,andB speakers.
proportion
Consideraninitial
population, bytheblack
represented dot.Therelative of
proportions
L,andB speakers
//, aregivenbythelengthsofthelinesnotatedaspm, andp^, re-
A point
spectively. inthemiddle,forexample, inwhich
thesituation
represents allthree
typesareequally
numerous.Apoint inthebottom right a system
represents with
starting
onlyL speakers.
Theshading oftheregionsurroundingevery point thefinal
represents
Dotted
population. regions theareawhere
represent onlyH speakersexist
atequilibrium,
anddarkregions theareawhere
represent onlyL speakers
exist Inallcases
atequilibrium.
thesteady-state istheexistence
solution ofa singlelanguage, Hunless
generally theinitial
contains
population fewH speakers.
Themodel isthesameasinFigure 3: aH= 3,aL=
1,anduj= 2.

stablebilingualism so longas thepropensity to learna language(i.e., thechild's

is
receptiveness) sufficiently strong;that is, aH and aL are sufficiently large.This
is showninFigure5, whichshowssteady-state population as a function ofaH and
aL foruj = 1 (i.e., equal status;
Figure 5A) and u = 2 (i.e., H status higher thanL
status;Figure5B). The hatchedareasshowwherethereis stablebilingualism for
high values of aH and aL. To theleftof the hatched region, H , L, and B speakers
coexistin a steadystate.

Interventionin the Public Sphere. In theprevioussectionwe examinedthe

dynamics modelinwhichH andL started
ofa atequal frequencybutL eventually

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 02:38:55 AM

All use subject to JSTOR Terms and Conditions
68 / FERNANDO
ET AL.

5. Equilibrium
Figure ofpopulation
concentration when theinitial
conditionsare90%H speakers
and10%L speakersforvariousvalues
ofawand aLfor(A)us= 1and(B) u = 2.Where
thefinal isofL speakers,
population theregion
isstippled.
Where thefinal isof
population
H speakers,
theregion
issolidgray.
Where thefinal
populationisbilingual,
theparameter
regionisshadedwith
lines.
Thelackofsymmetry inpartA isduetothedifference
inini-
tialpopulation.
Large ofaHandaLresult
values instable Evena moderate
bilingualism.
change inu increases
therequired
valueofaLnecessaryforL todominate orforstable
bilingualism.

wentextinct. Whatmanipulations to thesystemcouldsaveL whenL has already

beendecayingso thatitsproportion is low (10%)?
Dynamicsofa mixedintervention inFigures3C and3D showthatan inter-
ventionstrategy can promotelanguagestability. Whattypesofintervention work
best?Figure6 showstheresultsof exploring theparameter space possiblein-
of
terventions. In general,increasing thereceptiveness ofchildren toL (i.e.,changes
in aL) are notsufficient unlessdrastic.Moderatevaluesof A and mHB can be ef-
fective,especiallywhenused together alongwithan increasein aL. In general,
an increasein A is abouttwiceas effective as a similarly
sizedincreasein mHB in
increasing theproportion ofL speakersin a community (see Figure6). Once the
parameter settingsof themodelhavebeenconfirmed, governments coulddecide
to investin alteringA,mHB , or aL, to achievemaximum effect.

Conclusion and Discussion

Our resultsare relevant
to thefieldof languagepolicyand languagerevi-
talizationbecause theygive mathematical validationto thesupportive
policies
whosegoal is languagerevitalizationand maintenance.The mainfeaturesofthe
previouspolicyrecommendations are guaranteeing transmis-
intergenerational
sion,increasingthe statusof theendangeredlanguage,changingthe attitudes
of the speakers,improving theeconomicsituationof thepeople,teachingthe

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 02:38:55 AM

All use subject to JSTOR Terms and Conditions
 LanguageExtinction
and Revitalization
Strategies/ 69

6. Effect
Figure ofgovernment intervention
byincreasingmHB when
andA,including aL is in-
creased.
Darkness ofshading
represents
steady-state ofL andB speakers
proportion in
(A)aH= 3,aL= 1,lj = 2,andstarting
thepopulation. is90%H and10%L.
population
(B)Additional
effect aLto1.5.Notethat
ofincreasing strong canresult
intervention inH
extinct
going (i.e.,theentire speaks
population L).Ingeneral, Aisabout
increasing twice
aseffective
asincreasingmHBbythesameamount.

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 02:38:55 AM

All use subject to JSTOR Terms and Conditions
70 / FERNANDO
ET AL.

endangered languagein schools,increasing itsuse in thepublicdomain,andim-

provinglegislation(Cantoni1996; Crystal2000; Fishman1991; Grenobleand
Whaley2006; Nettleand Romaine2000; UNESCO Ad Hoc ExpertGroupon
Endangered Languages2003). Ourresultsalso emphasizetheinterconnectedness
of variousrevitalization measuresand showwhichmeasuresare mosteffective.
The followingdiscussionrelatesourresultsto existingsupportive policiesand
languageplanningsolutions.
First,ourmainfinding, consistent withpreviousmodels,is thatlanguages
aremutually exclusiveunlesschildren arehighlyresponsive to learning bothlan-
guages(large,roughly equal valuesofaL andaH). Existingevidencesuggeststhat
without interventionthisis unlikely tobe thecase,becauseminority languagesare
in factdyingout.It is an empirical questionwhether thehighvaluesofaL andaH
requiredforlanguagecoexistencecan be achieved.Some effective policiesand
planningmeasuresforincreasing aL includenotonlygivingthelow-status lan-
guagean officiallegal status(Romaine2002) butalso takingactivemeasuresto
makethelanguageeconomically usefulandattractive in thelabormarket, thatis,
givingitmarket value(Grin1999),andconducting languageawarenesscampaigns
thathighlight andexplainthebenefits ofbeingbilingual(EdwardsandPritchard-
Newcombe2005). Greater representation ofhigher-status languagesinthemedia,
resultingin u > 1,increasesthevalueof aL necessary formaintenance ofL. This
suggeststhatactivelyreducing lucan assistinlanguagerevitalization efforts.
Second,ourmodelshowsthatincreasing theamountofthelow-status lan-
guage in thepublicsphereor providing formallanguageinstruction can work;
however,interventions thatincreasetheproportion of the low-statuslanguage
heardin thepublicsphereare particularly effective and synergistic withinter-
ventionsthatincreaseresponsiveness aL or provide formal teaching (mHB)(see
Figures3 and6). Assumingthattheparameter valuesinFigure6 canbe supported
withquantitative data,we findthatcontinuous intervention thatincreasestheuse
in publicdomainsis approximately twiceas effective as interventions thatuse
formalteaching(Figure6). Teaching(mHB)can achievean initialconcentration
of speakersofthelow-status language,andthisconcentration can be maintained
withotherinterventions.
The observedsynergy betweeninterventions highlights theinterconnect-
ednessof revitalization policiesand measures.Strubell(2001) and Walshand
McLeod (2008) also suggested thatincreasing publicservicesavailableinthelow-
statuslanguagemakesit moreattractive, whichincreasesits attractiveness and
leadsto thelikelihoodof itbeingstudiedandused,andthisin turnincreasesthe
needforpublicservicesandso forth. Notethattheimplications ofourresultsdif-
ferfromFishman's(1991) policyrecommendation of securingintergenerational
transmission and achievingstablediglossia(i.e., increasingthelanguagebeing
spokenin low-status domains)beforebringing it to publicspheresor schools.It
is difficult
to achievethesuggested highinitialpopulation bytargeting theprivate
domainon its own,as ourfirstresultshows.A moreintegrated approachis re-
quiredinsteadofthestagesthatFishmansuggests.

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 02:38:55 AM

All use subject to JSTOR Terms and Conditions
 LanguageExtinction Strategies/ 71
and Revitalization

Third,ourmodelshowsthat,as in Figure4, thelower-status languagecan

an
persistagainst externally stable higher-status language ifthe initial proportion
of speakersof thelow-status languageis extremely high,as in Quebec,where
80% of thepopulationalreadyspokeFrenchwhenlanguagerevitalization was
initiated(Bourhis2001), or in England,whenlower-status English was able to
survivetheNormanconquest.In bothcases current bilingualism is due to con-
tactand populationmixingwithexternalpopulations. The policyrecommenda-
tionthatfollowsthisresultis thatlanguagemaintenance shouldbe started when
a low-status languagehas a high initialpopulation in a certain area. This differs
fromthegeneralpracticein whichcommunities andpolicymakersaregenerally
notconcerned aboutsecuringthefuture oflargelanguages.
Fourth,ourmodelshowsthatifthestatusof thetwolanguagescannotbe
made high,thencontinuousgovernment intervention is neededto preservethe
low-statuslanguage.When government supportmeasuresare abandoned,the
low-status languagedies out unlessone has managedto dramatically increase
its numberof speakersso thatthelow-status languageis able to dominatethe
high-status language(see Figure4). A robustfinding (overa wide rangeof re-
alisticparameter values) of our modelis therefore thatthelow-status language
can be stable(in coexistencewiththehigh-status language)onlyifintervention
is continuous (Figures3 and6). Continuousintervention (e.g., aL = 1.5,A = 0.1,
=
mHB 0.1) can resultin stablebilingualism evenifthelow-status languagehas a
negligible number ofspeakerstobeginwith.In practicethismeansthatabandon-
inga revitalizationprojecttooearlywouldreverseanygainsmade,whichis what
GrenobleandWhaley(2006) pointedout.
ourmodelshowsthatsegregation
Fifth, can supportlanguagemaintenance
together withintervention inthepublicsphere(cf.Patriarca andLeppänen2004).
Ourmodelinitially assumedthatall speakersofthelanguagescan matewitheach
other,so longas theycan communicate witheach other,whichleads to thelow-
statuslanguagegoingextinctas speakersof thelow-status languagematewith
bilingualspeakers and their bilingual children mate with speakers of thehigh-
statuslanguage.Increasing theamountofsegregated matingin ourmodelslowed
downtheloss ofthelow-status language(givenno intervention) butdidnotaffect
theultimate outcome.It did,however, increasethesteady-state concentration of
thelow-status language and of bilingualism when government intervention was
increased.Thussegregation is synergistic withgovernment intervention butis not
a long-term solutionon itsown.
Matingsegregation could in practicereferto a situation in whichcultural
segregation existssympatrically (as is the case with the successfully revitalized
Hebrewor withtheAmish,who have managedto maintaintheirlanguage)or
whereregionalism is strong(i.e.,wherespeakersofendangered languageschoose
to stayin theirtraditional area). These are instancesof self-imposed boundary
maintenance (Paulston1994), or whatBourhis(2001) calls a separatist orien-
tation,whichis ratherrare.Policymakersshouldinsteadtryto encouragethe
speakersof an endangered languageto stayin a local areathrough development

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 02:38:55 AM

All use subject to JSTOR Terms and Conditions
 ET AL.
72 / FERNANDO

programs thatwouldsecurean incomeinthatparticular place,without thespeak-

ers havingto moveto urbancenters[or to thevicinityof rubberplantations or
mines,as has happened,forexample,in theAmazon(Aikhenvald 2003)]. Simi-
larly,Scots Gaelic speakerswereforcedto switchoverto Englishbecausethey
losttheirlocal livelihoodas fishers andhadto seekotheremployment elsewhere,
whichledtothedisintegration oftheirsociety(Dorian1981).Thisfurther empha-
sizes theneedto developlanguagepoliciestogether witheconomicandregional
policies(NettleandRomaine2000).
The remaining elementsof modelvalidationare to justifythechoice of
parameter values for77and u. We have assumedthat77is 0.5; thatis, 50% of
languagelearningandlinguistic socializationtakesplace at homeand50% takes
place outsidethehome.Thisvaluecan be changedaccordingtothetypeofsitua-
tionin questionbasedon theempiricaldataacquired.The valueofujis an impor-
tantparameter forunderstanding theroleofmediaandhowgovernment discourse
dominates publicdiscourse,in thiscase,in theuse oflanguage.
One shouldbearin mindthatnotall languagerevitalization projectshave
thesame aims.The aim is notalwaysto achievefullfluency by all community
members andtouse theendangered languageinall domains.Few programs reach
thatgoal. The revitalization of Cornishhas been successfulin teachingpeople
some Cornish,notexpectingfluency, and thesame appliesto Hawaiian,which
is spokenmainlyin schoolsbutnotin theprivatedomain(GrenobleandWhaley
2006). So in thesecases therehas in essencebeen an increaseof semispeakers
usingthelanguagein limiteddomains.Efforts in cases wherea languagehas few
speakersshouldperhapsbe focusedon documentation workand culturalaware-
ness programs, as it is rarethatformalteachingwill lead to intergenerational
transmission. The smalllanguagescould insteadsurviveas markers of identity,
typesofculturalandsymbolictokensthatareusedin ceremonies orwhengiving
namesor greeting, as is thecase withthereclaimedAustralian languageKaurna
(Amery 2000). Modeling thesealternativegoals remains for future work.The
modelingof multilingual systemsalso remainsforfurther work.In practice,it
is importantto setrealisticandpracticalgoals in languagerevitalization projects
(Dauenhauerand Dauenhauer1998) and notto expectpolicychangesto give
instantresultswithoutlong-term implementation and a multifaceted approach,
whichis supported byourmodel.

Received
15April2009;revision
accepted 2 September
forpublication 2009.

Literature Cited
D. M.,andS. H.Strogatz.
Abrams, 2003.Modeling
thedynamics
oflanguage
death.
Nature
424:900.
A.Y.2002.Traditional
Aikhenvald, andlanguage
multilingualism InLanguage
endangerment. En-
andLanguage
dangerment Maintenance
,D. Bradley
andM.Bradley,
eds.London:
Routledge
Curzon,24-33.
A.Y.2003.Teaching
Aikhenvald, anendangered
Tariana, from
language northwest
Amazonia.
Intl.
J.
Sociol.
Lang.161:125-139.

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 02:38:55 AM

All use subject to JSTOR Terms and Conditions
 LanguageExtinction
and Revitalization
Strategies/ 73

Amery,R.2000.Warrabarna Kaurna! Reclaiming anAustralian Language. Lisse, Netherlands:Swets

andZeitlinger.
R.Y.2001.Reversing
Bourhis, language shift inQuebec. InCanThreatened Languages BeSaved?
Reversing Language Shift : A list-Century
, Revisited Perspective, J.Fishman, ed.Clevedon,
U.K.:MultilingualMatters, 101-141.
R.,andP.Richerson.
Boyd, 2005.TheOrigin andEvolution ofCultures. Oxford, U.K.:Oxford Uni-
versityPress.
D. 2002.Language
Bradley, attitudes:
Thekeyfactor inlanguage maintenance. InLanguage Endan-
germent andLanguage Maintenance , D. Bradley andM.Bradley, eds.London: Routledge
Curzon, 1-10.
Y.G.,andK.Hakuta.
Butler, 2006.Bilingualism andsecond languageacquisition. InTheHandbook of
Bilingualism,T.K.Bhatia andW.Ritchie, eds.Oxford, U.K.:Blackwell, 114-144.
L.,andM.C.Muntzel.
Campbell, 1989. Thestructural consequences oflanguage death. InInvestigat-
ingObsolescence: Studies inLanguage Contraction andDeath , N.Dorian, ed.Cambridge
U.K.:Cambridge University Press,181-196.
G.,ed.1996.
Cantoni, Stabilizing
Indigenous Languages. Flagstaff:Northern Arizona University.
X.,L. Loureiro-Porto,
Castellö, V.M.Egufluz etal.2007.Thefateofbilingualism ina model of
language InAdvancing
competition. SocialSimulation : TheFirst World Conference, S. Taka-
hashi,D. Sallach,
andJ.Rouchier, eds.Tokyo: Springer,83-94.
M.H.,andS. Kirby.
Christiansen, 2003.Language Evolution.NewYork: Oxford University Press.
M.2003.Dynamics
Clyne, ofLanguage Contacts: English andImmigrant Languages. Cambridge,
U.K.:Cambridge University Press.
D. 2000.Language
Crystal, Death. Cambridge, U.K.:Cambridge UniversityPress.
A.2003.Language
Dalby, inDanger: TheLossofLinguistic DiversityandtheThreat toOurFuture.
NewYork: Columbia University Press.
DalNegro,S. 2004.Language contact anddying languages. Rev.Franc. Linguist.Appl. 9(2):47-58.
Dauenhauer,N.M.,andR. Dauenhauer. 1998.Technical, emotional, andideological issuesinre-
versing languageshift:Examples from southeast Alaska. InEndangered Languages: Current
IssuesandFuture Prospects, L. A.Grenoble andL. J.Whaley, eds.Cambridge, U.K.:Cam-
bridge Press,
University 57-98.
DeBot,K.,andS. Stoessel.2002.Introduction: Language change andsocial networks. Intl.J.Sociol.
Lang.153:1-7.
R. M.W.1998.TheRiseandFallofLanguages.
Dixon, Cambridge, U.K.:Cambridge University
Press.
N.C. 1981.
Dorian, Language Death: TheLifeCycle ofa Scottish GaelicDialect. Philadelphia: Uni-
versityofPennsylvania Press.
N.C. 1998.
Dorian, Western language ideologies andsmall-language prospects. InEndangered Lan-
guages: Current
Issues andFuture Prospects, L. Grenoble andL.J.Whaley, eds.Cambridge,
U.K.:Cambridge University Press,3-21.
J.2006.Foundations
Edwards, ofbilingualism. InTheHandbook ofBilingualism, T.K.Bhatia andW.
Ritchie,eds.Oxford, U.K.:Blackwell, 7-31.
V.,andL.Pritchard-Newcombe.
Edwards, 2005.Language transmissioninthefamily inWales. Lang.
Probl.Lang. Plann.29:135-150.
C. 1959.
Ferguson, Word
Diglossia. 15:325-340.
J.A.1991.
Fishman, ReversingLanguage Shift: TheoreticalandEmpirical Foundations ofAssistance
toThreatened Languages. Clevedon, U.K.:Multilingual Matters.
J.A.,ed.2001.CanThreatened
Fishman, Languages BeSaved? Reversing Language Shift,Revisited:
A21stCentury Perspective.Clevedon, U.K.:Multilingual Matters.
M.2008.Language
Florey, activismandthe"newlinguistics": Expanding opportunities fordocu-
menting endangeredlanguages inIndonesia. InLanguage Documentation andDescription, P.
Austin,ed.London: SOAS,v.5,120-135.
N.,andK.G.Troitzsch.
Gilbert, 2005.Simulation fortheSocialScientist. Buckingham, U.K.:Open
University Press.

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 02:38:55 AM

All use subject to JSTOR Terms and Conditions
74 / FERNANDO
ET AL.

Grenoble, L. A.,andL. J.Whaley, eds.1998.Endangered Languages: Current IssuesandFuture

Prospects.Cambridge, U.K.:Cambridge UniversityPress.
Grenoble, L.A.,andL.J.Whaley. 2006.Saving Languages: AnIntroduction toLanguage Revitaliza-
tion. Cambridge, U.K.:Cambridge University Press.
Grin,F.1992. Towards a threshold theory ofminority language survival.Kyklos 45:69-97.
Grin,F.1999. Market forces,language spread, andlinguisticdiversity.InLanguage: ARightanda Re-
source - Approaching Linguistic Human Rights,M.Kontra, R.Philipson, T.Skuttnabb-Kangas
etal.,eds.Budapest: Central European UniversityPress, 169-186.
Hale,K. 1992.Language endangerment andthehuman valueoflinguistic diversity.
Language
68:35-42.
Harrison,D. K.2007.When Languages Die:TheExtinction oftheWorld's Languages andtheErosion
ofHuman Knowledge. Oxford, U.K.:Oxford UniversityPress.
Hill,J.H.2002.Expert rhetorics inadvocacy for endangered languages: Whoislistening, andwhat do
they hear?J.Ling. Anthropol. 12:119-133.
Hinton, L.2001.Themaster-apprentice learningprogram. InTheGreen BookofLanguage Revitaliza-
tion inPractice,L. Hinton andK.Hale,eds.SanDiego: Academic Press, 217-226.
Hinton, L.,M.Vera, andN.Steele. 2002.HowtoKeepYour Language Alive. Berkeley,CA:Heyday
Books.
Jablonka, E.,andM.J.Lamb. 2005.Evolution inFour Dimensions: Genetic , Epigenetic,Behavioral,
andSymbolic Variation intheHistory ofLife. Cambridge, MA:MITPress.
King, J.2001.TeKöHanga Reo:MäOrilanguage revitalization.
InTheGreen BookofLanguage
RevitalizationinPractice ,L. Hinton andK.Haie,eds.SanDiego: Academic 119-128.
Press,
Krauss, M.1992. Theworld's languages incrises. Language 68:4-10.
Malik, K.2000.Letthem die.Prospect 57:16-17.
McElreath, R.,andR.Boyd. 2007.Mathematical Models ofSocialEvolution: A Guide forthePer-
plexed. Chicago: University ofChicago Press.
Minett,J.,andW.S.-Y.Wang. 2008.Modeling endangered languages: Theeffects ofbilingualismand
socialstructure.Lingua 118:19-45.
Mira,J.,andA.Paredes. 2005.Interlinguistic similarityandlanguage death dynamics. Europhys.Lett.
69:1031-1034.
Mufwene, S. S. 2001.TheEcology ofLanguage Evolution.
Cambridge, U.K.:Cambridge University
Press.
Mufwene, S. S. 2004.Language birth anddeath. Annu. Rev.Anthropol. 33:201-222.
D.,andS.Romaine.
Nettle, 2000.Vanishing Voices: TheExtinction oftheWorld's Languages. Oxford,
U.K.:Oxford University Press.
Ö Riagâin, P.2001.Irish language production andreproduction 1981-1996. InCanThreatened Lan-
guages BeSaved? Reversing Language Shift, - A21stCentury
Revisited , J.Fish-
Perspective
man, ed.Clevedon, U.K.:Multilingual Matters, 195-215.
M.,andT.Leppänen.
Patriarca, 2004.Modeling languagecompetition. PhysicaA 338:296-299.
Paulston,C. B. 1994.Linguistic Minorities inMultilingual Settings:Implications onLanguage Poli-
cies.Philadelphia:John Benjamins.
Pinker,S. 1994.TheLanguage Instinct: HowtheMind CreatesLanguage. NewYork: HarperCollins
Romaine, S. 2002.Theimpact oflanguage policy onendangered languages. Intl.J.Multicult.
Soc.
4(2):194-2 12.
Romaine, S. 2006a.Thebilingual andmultilingual community. InTheHandbook ofBilingualism,
T.K.Bhatia andW.C.Ritchie, eds.Oxford, U.K.:Blackwell, 385^05.
Romaine, S. 2006b.Planning forthesurvival oflinguistic Lang.
diversity. Policy 5:441-473.
Sallabank, J.2007.Endangered language maintenance andsocialnetworks, http://www.hrelp.org/
events/conference2007/abstracts/8.pdf (accessed January 7,2009).
D.,andC.Schultze.
Stauffer, 2005.Microscopic andmacroscopic simulation ofcompetitionbetween
languages.Phys. LifeRev. 2:89-116.
Stoessel,S. 2002.Investigatingtheroleofsocialnetworks inlanguage maintenance andshift.
Intl.
J.
Sociol. Lang.153:93-131.

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 02:38:55 AM

All use subject to JSTOR Terms and Conditions
 LanguageExtinction StrategiesI 15
and Revitalization

M.2001.Catalan
Strubeil, a decade InCanThreatened
later. LanguagesBeSaved? ReversingLan-
guage Shift A21stCentury
, Revisited: ,J.A.Fishman,
Perspective ed.Clevedon,U.K.:Multi-
lingual Matters,260-283.
Thomason,S. G.2001.Language Contact:AnIntroduction. DC:Georgetown
Washington, University
Press.
Thomason,S. G.,andT.Kaufman. 1992.Language Contact, andGenetic
Creolization, Linguistics.
Berkeley: ofCalifornia
University Press.
UNESCO AdHocExpert Group onEndangered 2003.Language
Languages. andendanger-
vitality
ment. Paper attheInternational
presented ExpertMeetingonUNESCOProgramme Safe-
guarding ofEndangered Languages,Paris,March 10-12.http://www.unesco.org/culture/ich/
doc/src/00120-EN.pdf(accessed April19,2008).
J.,andW.McLeod.
Walsh, 2008.Anovercoat wrappedaroundaninvisible
man? Languagelegislation
andlanguage inIreland
revitalization andScotland.
Lang. 7:21-46.
Policy
W.S.-Y.,
Wang, andJ.W.Minett. 2005.Theinvasionoflanguage:Emergence,
change,anddeath.Tr.
Ecol.Evol.20:263-269.
S. 2001.Themovement
Warner, torevitalize
Hawaiian
language InTheGreen
andculture. Bookof
Language inPractice
Revitalization andK.Hale,eds.SanDiego:
,L. Hinton Academic Press,
133-144.
D.J.1999.
Watts, SmallWorlds:TheDynamics ofNetworksBetween andRandomness.
Order Prince-
ton, NJ:PrincetonUniversityPress.
B.-A.2005.Canbilingualism
Wickström, bedynamically Asimple
stable? model choice.
oflanguage
Rationality andSociety 115.
17(1):81-
Wurm,S.A.2002.Strategiesforlanguage maintenance InLanguage
andrevival. Endangerment and
Language Maintenance andM.Bradley,
, D. Bradley eds.London:
RoutledgeCurzon,11-23.

This content downloaded from 185.2.32.49 on Sun, 22 Jun 2014 02:38:55 AM

All use subject to JSTOR Terms and Conditions