Chapter12
LexicalCategories
and the
Evolution
of NumberMarking
MarianneMithun
Number has traditionally been considered a property of entities, a
characteristicof objects and people Accordingly, it has usually beenassumed
to be an inflectional categoryof nouns, When number distinctionsappear
on other words, such as verbs or adjectives, they are often automatically
classified as agreement markers, indicators of the syntactic relationship
betweenthosewordsand associatednouns.rather than inherentfeaturesof
the words thernselves,Th€y are thus frequently describedin terms of rules
that copy featuresof nouns onto associatedlexical items.For somelanguages,
such an.analysis seemsadequate For many others, however,copying rules
fail to capturethe essenceof number marking. In theselanguages,the number
markersthat appear on different kinds of lexical items operateindependently
of each other and perform subtly different functions. The functional differences result in formal differences in morphological number-marking
systems.In what follows, the nature of thesefunctional and formal differences
is examinedand the diachronicprocesses
behind them explored.
Frajzyngier(1985)and Durie (1986)haveshownthat nominal and verbal
number markers operate independently in many languages.This is the case
with most of the severalhundred languagesindigenous to North America.
The problemsintroducedby an agreementanalysisfor theselanguagesbecome
apparentas soon asspontaneousspeechis examined.In the majority of North
American languages, verbs can constitute grammatical predications in
themselves.Separatenominal constituentsare not only unnecessaryfor grammaticality, they are frequently absent.Still, evenwhen no nouns appear,verbs
can bear number markers. Note the plural prefix in the verb below from
InesefloChumash,a California languageof the Chumashfamily.
Th@ptical Motpholog
Copyright @ 1988by Aedenic p!€$, Itrc
All righ[ of Eprcduction in oy forn .*rrd.
Mithun
Marianne
212
0)
InesefloChumash
s-iy-art-
kum
3pl iterative dance
'They are dancing.' (Applegate' 1972' p. 458)
Thereis no nominal sourcefor the number specification,nothing for a copying
rule to copy.
In order to maintain a strict agreementanalysis, one might posit underlying nominal constituents as part of every sentence,allow copying rules to
duplicate thefu number features on associatedverbs, then permit optional
nominal deletion. This analysis would run into a second set of problems,
however.When separatenominals are actually present,they do not necessarily
agreein number with associatedverbs.
@
Ineseffo Chumash
ha ku
kum
s-iy-axi
the Person
3pl iterative dance
(Applegate, 1972,p, 458)
people
dancing.'
are
'The
In many of thesesituations,plural forms of the nounssimply do not exist'
In fact, in the majority of North American languages,nouns referring to
nonhumansare not marked for number at all. Yet evenwhen plurals do edst,
th€y are not necessarilyused.The InesefloChumashnoun 'tl 'persorf has
a cor]rfierp'lrt kuhkuz 'peoplfq' but it was not chosen for the above context'
Such mismatchesare usually not the result of sloppinesson the part of
speakers,although this possibility has sometimesbeensuggestedby frustrated
erammarians.
1. Number Marking with Ditferent Lexical Categories
1.1 Nouns
In sornelanguagesin North America, among them Thos,Kiowa, Zuni, and
the Algonquian laneuages,all nouns are inflected for number,just asin IndoEuropean. In the vast majority of North Anerican languages,however,only
certain nouns have plural forms. In most of these, only nouns referring to
human beings have plurals, or only some nouns referring to humans, often
kin terms. (Multiple animals that are consid€red'sentientbeings,'suchas pets
or charactersin legends,are also often referredto by plural nouns.)The plurals
that do exist are used only on some occasions,not everytime multiple particiDants are discussed.
TheE olutionof NumberMarking
2lt
1.2 Adjectives
Many North American languageslack a special lexical class of adjectives.
Modification is accomplishedby statile verbs.Where adjectivesdo constitute
a separatecategory, they often have plural forms, but use of these forms is
usually not obligatory, independently of whether human beings,animals, or
objectsare described.
1.3 Verbs
Verbs typically include a variety of number markers. Number is most frequently marked on verbs by bound pronouns, stem alternations, and various
derivational processes.
1.3.1 fuund ptonouns In the majority of North American languages,pronouns referring to the core arguments of a predication are bound to verbs
or auxiliaries as affixes or clitics. They are usually obligatory whether separate nominal constituentsaxepresentor not, as in (l) and (2). Languagesvary
in their repertoires of bound pronouns. Bound pronouns may representall
persons,or only first and second.Th€y may distinguishnumber systematically
throughout, or only for certain human referents,or, surprisingly often, not
at all.
l-3.2 Stemalumation In many North American languages,verb st€msalternate accordingto the number of participants involved. The set of alternating
stemsconsistsof a limited number of common verbs, in somelanguagesonly
two or threg in others up to severaldozen. They usually include intransitives
such as 'sit,' 'lie,' 'stand,' ?o,' 'walk,' 'run,' 'fly,' 'die,' and transitive such as
'take,' 'pick up,' 'carry,' 'throw,' 'kill,' Note the completelack of phonological
resemblancebetweenthe singular and plural verbs for 'sit' and 'kill' in the
languages(3) and (4), for example
'(oneto) sit'
?em
p. 52)
(Gibson,
1973,
qafii
(Sapir,1930,p. 242)
p. 276) qlao
(Swanton,.l9ll,
(Newman,
p. 55) rimo
1965,
(Kinkaid,1975,p. 48)
ttwiK
(3)
hit/dwell'
Shuswap
SouthemPaiute
Haida
Zuni
Upper Chehalis
(4)
'kill'
(Gibson,1973,p. 52)
Shuswap
SouthemPaiute (Sapir, 1930,p. 242)
(Swanton,l9ll, p. T16)
Haida
(Newman,1965,p, 55)
7nr11
UpperChehalis (Kinkaid,1975,p. 263)
'kill (one)'
P{i
W.a
tid
2ayna
XilA
'Groupto) sit'
Mq
yuvwi
L!tl
tina
ld\il
'kill (sweral)'
il(
qtT
Adl
llt!
-ai-
214
Mafianne Mithun
For intransitive verbs,the selectionof a stemreflects the number of subj€cts
involved,whether they are agents(walkers'runners' fly€rs) or patients(fallers'
sitters,corpses).For transitive verbs,it reflectsthe number of objectsinvolved:
how many people are killed, how many objects are picked up, thrown, given,
dropped, and so on. The pattern remainsthe sameregardlessof whether other
parti of the grammar operateon an active,accusative,or ergativebasis' This
is no accident. The subjects of intransitives and patients of transitives share
an important role: They are the participants most directly affected by an
action. The primary function of stemalternation is not to enumerateentities,
but to quantify the effect of actions, states,and events.
This stem alternation has sometimesbeenreferred to as suppletion. In the
strictest sensg suppletion refers to allormorphic alternation conditioned by
a systematicinflectional distinction. A prototypical exampleof stemsuppletion in English, wherewery verb is obligatorily inflected for tense,is the alternation between8o and wen'. The relationship betweenverbs like kick nd
s/dA by contrast, is not consideredsuppletive,since we do not systematically
spicify the instrument involved in everyaction. Kic,t and s/cp simply happen
to include different instruments as parts of th€ir basic meanings.The North
American verbs that alternate according to the number of entities affected
are related in a sirnilar way. The implied plurality of effect is a featur€ of
their basicmeaning.Walking alone is classifiedlexically as a different activity
from walking in a group; speaking is different from conversing;murdering
an individual is different from massacringa village The pairs of verbs are
related semantically but not inflectionally'
Reactionsof speakersconfirm the lexical independenceof thesealternating
stems.An excellentspeakerof Central Pomq a California Pomoan language'
was asked for the plural of the verb yow '(one) goes.' She srDplied hlaan
'Geveralare)walking.' On another occasion,when askedfor the singular form
of hlaan, sheprovided waan '(one is) walking.' Such chains are not unusual,
evenfrom the best speakers.
Not all verbs implying multiple effects wen havespecific singular counterparts, or vice versa.Note the asymmetryin the setsof verbsusedwith singular'
dual, and plural subjectsfor 'sit' and 'dwell' in Koasati,a Muskogeanlanguage
of lousiana:
(5)
Koasati
Singular
cokkblin
d:tan
Dual
cikki:lun
dswan
Plural
i:san
l:san
'sit'
'dwell' (Kimball, 1985,p.273)
The transitive verbs in (6) from Klamath, an Oregon language, constitute
another common type of set, Different handling verbs are used depending
upon the nature of the substancehandled: round, flat, alive, or multiplg but
not 'multiple round' or 'multiple flat.'
TheEvolutionof NumberMa*ing
(6)
Klamath
Poy
n'oy
ksoy
szewan|
'to give a round object'
'to give a flat object'
'to give a live object'
'to giveplunl objects(round, flat, live, etc)' (Barker, 1964,
p. 176)
The lexically plural verbs are not unlike English congregaE,dkryrce, gather,
and scatter.These English verbs also imply multiple subjects when intransitive and multiple patients when transitivg but it is not immediately obvious
what their singularcounterpartsshould be. Here, toq the number distinction is semanticrather than grammatical. Pants(grammaticallyplural, semantically singular) could not scatter, but a crowd (grammatically singular,
semanticallyplural) could gather and disperse (SeeDurie, 1986,for a detailed
discussionof similar issues.)
Although verb stemsimplying multiple effects may appear to agreewith
plural nouns or pronouns, a match is not obligatory. kxically plural verbs
may denote events involving groups, not simply multiple participants. In
English,if eachsolitary hiker who passesa benchreststherefor a moment,
they havenot necessarilycongregated;if a murderer kills threevictims during
his lifetimg they havenot necessarilybeen massacred.Verbswith plural subjects or objects may contrast with lexically plural verbs' In Yurok, a California Algic languaggthe verbnep'eat' hasa counterpart?e']gah'havea meal
together.'It also has rplural form neph. which could be usedif sweralpeople
ate individually. Th e verbcwinl<ep'speak' has a semantic cotll:'terparttohkow
'talk together, converse,'but it also has a p|rulalcwinltepoh '(several)speak'
(Robins, 1958,p, 42).In his grammar of Sarcee,an Athabaskanlanguage
of Alberta, Cook remarksthat "like the unmarkedstems,the plural stems
may occur with either a singularor a plural subject,With a plural subject,
-ddl means'three or more move,' and with a singular subject it means 'one
movesin companywith two or mord (1984 p. 59).Kinkaid (1975,p. 49) reports
similar systemsin Salishlanguages.
1.3.3 Derivational Prccesses Most North American languagesshow derivain their verbal morphology that appearto reflect number.
tional processes
verbs
below from Karok, a language of Northern California.
Compare the
(7)
Karok
qivr{r.htih
0ivru'hti'h-va
ikyi.m-kurih
ikyimku'rih-vs
2dkunv-ai
zdkunvan-va
'(one object) to be floating'
'(severalobjects)to be floating'
'(one) to fall in'
'(several)to fall in'
'(one) to go hunting'
'(several)to go hunting' (Bright, 1957,p.92,93)
Ma anne Mithun
216
The suffix -va appearsto indicate a plural subject, at least with intransitive
verbs. When it appearswith transitive verbs, it can indicate a plural patient.
(8)
Karok
pasndp-i\(rih)
pasnapi'3i.h-va
'to glue down (one)'
'to glue down (several)'
'to be looking into (one object)'
'to be looking into (severalobjects)'
(Brisht, 1957,P. 92, 93)
itrdo.mnihtih
itmmni'htih-va
Actually, the suffix need not pluralize any particular entities at all.
(9)
Karok
uknah
taknd.h-va
'to
'to
'to
'to
'to
'to
hop'
play hopscotch'
sed
sed
weavearound (once)'
v[k-pa?
weavearound and around' (Bright, 1957' p.92')
vikpd.|-vo
The primary function of -v4 is to multiply actions, It may imply that the
actions are distributed over multiple agents or multiple patients, but this is
a secondaryeffect. The actionscan just as well be distributedover time or
sDace:
mah
md.h-va
00)
Karok
ikre.mydhii@h)
ikre.myahi'iri h-vo
?drykunih
?apaku'nih-vs
'to start to blow'
'to blow off and on'
'(earth) to slide down'
'to be a landslidd (Bright, 1957,p. 92)
Multiple event or distributive suffixes do not appear every time multiple
agentsor patients are involved. The first verb in (11)involves a single cook,
as shownby the singular pronominal prefix tu- 'helshe'' The secondinvolves
severalcooks, as shown by the plural pronorninal prefrx kun-'they.' In both
cases,however,the cooking is considered a single activity, so no multipleevent suffix appears.
0l)
Karok
2 -mni|
hn-imnis
'he cooks'
'they cook'
The Karok multiple eventsuffix -v4,like thosein other languages,is derivational. not inflectional, It creates lexical items with specific meanings.
taknnftvd to play hopscotcf ('jumd-vd) couldjust aswell mean
Et]rmrologLcaliy,
'to hop up and down repeatedly,'or 'to hop around in different directions,'
but it has beenlexicalizedwith the sp€cific meaning it was derived to express.
TheEvolutionof NumberMa*W
211
Although multiple-eventor distributive markersareby far the most common
verbal quantifiers in North American languages, other kinds of number
markers also appearon verbsin many languages.They typically specify such
featuresas distributive causation,collectivecausation,collectiveagency,multiple displacement, and iteration of various sorts. Coos contains special suffixes for multiple causativepassivesand multiple reflexives, phonologically
unrelated to their singular counterparts @rachtenberg, 1922,p. 358). The
various verbal number markers do not necessarilyco-occur, although they
can. They quantify different aspectsof the eventpredicated. Central Pomo,
for exarnple,exhibits extensiveverb stem alternation accordingto the number
of participants affected,and it alsocontainsa multiple-eventsuffix, a multipledisplacement suffix, and a collective-agentsuffix, a multiple-displacement
suffix, and a colletive-agent suffix, among others. The number markers are
in boldfacein the commandsin (12).
(12)
Central Pomo
znee-
la-
m
throw one object down sg imp
'Throw it down!'
m
lamUthrow multiple objects down sg imp
'Throw them down!'
tamIamEathrow multiple objects down multiple event collective agency
mel '
pl imp
'Throw them down!' (FrancesJack, Personal Communication)
In the last command, every morpheme except the directional suffix 'down'
quantifies some aspectof the action, but none requires the pr$ence of any
other grammatically.
A similar exuberanceof verbal number markers can be seenin Southern
Paiute, a Uto-Aztecan language of Utah described by Sapir (1930), Some
distribut€ actions over timg functioning essentiallyas aspectrnarkers. They
indicate suchthings as continuous repetitive action ('patter') or durative iteration (?naw'), Southern Paiute also exhibits verb stem alternation according
to number, as well asvarious distributive and multiple-agencymarkers. They
can occur singly or in combination, as in (13).
(13)
SouthernPaiute
qsvi.
nvqi
tA\t'uPiva
reduceto smallpiecesdurativeiterative transitive pl agent remotepast
'They chippedit into small pieces.'(Sapir, 1930,p. 670)
218
Ma 4nne Mithun
Verbal number marking is pervasivein North America, but it doesnot represent simple agreementwith nouns. It operatesindependently, modifying
the verbs themselves.
of Number
2. The Grammaticization
The numbermarking typical of Indo-Europeannouns enumeratesentities,
while that of North American verbs usually quantifies aspectsof events.But
exactlyhow doesthis come about? Do number markersbind morphologically
with noun or verb stems only when their semantic contributions are
appropriate? Or do their meanings shift after they have become bound?
The synchronicmorphological systemsof someNorth American languages
provide glimpsesof the path of developmentof number marking' In several
languages,phonologically sirnilar number markers appear with words from
different lexical classes.Their resemblancessuggesta common diachronic
source and presumably a single ancestral function.
2.1 Reduplication
The most common form of number marking over multiple lexical categories
is reduplication. In some North American languages,such as those in the
Algonquian and Pomoan families, only verbs are reduplicated. In many
languages,however,the samereduplicative processesthat mark number on
verbs also appear on nouns and even adjectives'
2.1.1 Thefunctions of reduplication in different lexical categories The Tsim'
shian texts recordedby Boas on the Nass River in Northern British Columbia provideexcellentdocumentationof the function of reduplicationin that
language Verbs, nouns, and adjectivescan all be reduplicated according to
the samepatterns.(Most of the pattemsare somewhatcomplex,but th€y apply
to all lexical classes.)In his grammatical sketch, Boas (lgUb) simply listed
the reduplicated forms as the plurals of the plain forms. An examination of
the t€xts,however,indicatesthat reduplicationdoesnot producesimpleplumls.
Reduplication appearsonly a fraction of the time that multiple entities are
discussed.
Verbs can bear reduplicative prefixes when multiple intransitive subjects
are involved as in (14), or multiple transitive objects, as in (15).
(14)
NassTlimshian
NLk E ad'd'dTk'sleL wi-hc'ldEm qc'wun.
gull
then redup came many
(Boas,
p. 103.13)
gulls
1902,
many
came'
'Now
TheEwlutionof NumberMarkw
O5)
219
NassTbimshian
NLk 'c q'ox.qbyd'ant.
Then redup he clubbedthem
'Then he clubbedthem'' (Boas, 1902'p. 70.9)
Yetverbsare not reduplicatedwery time multiple subjectsor objects areidentified. The verb hwant 'sit' is not reduplicatedin (16).
06)
NassTbimshian
Ga.a'aL g.at hwant aL g' c'nx'g'e.
they saw man sitting at inland
'The men who weresitting under the trees sarvwhat he was doing"
(Boas, 1902,p. 111.12)
The fact that severalmen were present is clear from the choice of the verb
hwant'(serrcral) sil'; a different verb db is used if one person sits alone The
verb 'severalsit' does havea reduplicated form. It follows the same pattern
as hwiflhuwi' p 'house,/houses,'and hwa/huu)d' 'name/names'' The
reduplicated forrn appears in contexts like that in 07).
07)
Nass Tbimshian
txanctwL k'opE$'a'66.
NLk"c huwa'nL
the little birds
redup they sat down all
then
And all the birds sat down.' (Boas,1902,p. 124.15)
NassTbimshianreduplicated verbs distribute an eventover multiple times or
locations, including multipl€ participants.
The function of reduplication on nouns is interesting. The simple noun
gan 'stick' refers to single stick in (18).
08)
NassTsimshian
gan
NLk.A gaL
he took a stick
then
'Then he took a stick.' (Boas, 1902,p. 68'3)
The reduplicated form ganga'n refers to multiple sticks in (19)'
09)
Nass Tsimshian
K'E hwil sag4it-hd'p:aaL tbn l E-hisyat tst aL gonga'n,
then all together they who all over hit him with redup stick
'Then he was attacked and beaten with sticks,
aL nfg.it g.a'alL g'at.
Eam'ba'8ait'bEbEsba'tskL gutga'n
theywerelifted rcdupstick
he saw a penon by themselves
and not
althoughhe did not seea singleperson.The sticksmovedof themselves
220
MariameMithun
t'an itisya' tsL tEpl'a'nt' tEm'qc' st, qa'an'A'ft' asEsa'et.
his body, his head, his hands, his feet'
which hit
hitting his body, his head, his hands, and his feet.'
(Boas, 1902,p. 62.12-13)
Reduplicated nouns are not simple plurals, however.The noun for sricl<is
unreduplicatedin (20), eventhough we know that multiple stickswereinvolved
from the choice of the verb dOq'to take (several).'(The verb gd means 'to
take (one).')
(20)
Nass Tsimshian
HUX d6'qdeL gan
again they took stick
Again they took sticks.'(Boas, 1902,p. 97.12)
Like reduplicated verbs, reduplicated nouns function as distributives,
emphasizingtemporal,locativg or conceptualdistribution:the separateness
of the entities th€y identify.
Reduplicatedverbs and nouns often appear to 'agree.'In (16)' neither the
verb 'sit' nor the noun 'person'is reduplicated.In (21)' both the verb 'load'
and its object 'canoes' are reduplicated.
(21\
NassTsimshian
mmal.
huX mix.mdx.L
again redup they loaded r€dup canoe
Again they loaded the canoes''(Boas, 1902,p. 108'12)
Thesematchesreflect semanticco-occurrencesrather than grammaticalagreement. Distributed entities are often involved in distributed actions. The multiple canoesof (21)requiredmultiple loading. In (16),by contrast,a singlegroup
of peoplewasseatedtogetherat one time in one place,so neither the verb
nor the noun was reduplicated.
Although reduplication is highly productive in NassTsimshian,Boas'stexts
indicate that it is not applied uniformly to all verbs and nouns. It is a derivational process,usedonly when neededto createnew lexicalitems.Distdbution is a more important aspectof someactionsand entitiesthan of others,
and it is thesethat can be describedwith reduplication' Many verbsand nouns
lack reduplicatedforms, so distribution is simply not distinguishedfor them
in any context.In (14),for examplg the reduplicatedverb ad'd'dik'skl \ame
indicates distributed activity, but the noun for 'gull' remains unspecified for
distribution.
Nass Tsimshianadjectivescan also be reduplicated.The qualities they
specify are distributed individually over each item described, rather than
ascribed to the set as a unit. Reduplication on adjectives operates
independentlyofthe distributedstatusof the nounsmodified' The noun for
221
of Numbet
Marking
Ihe Evolution
for
'salmon,' for example,has no distributive form, so il is alwaysunmarked
this featua Nevertheless,the quality of fatnessis distributed by a reduplicated
adjective over the individual fish in (22).
(22')
Nass Tbimshian
DEm moct'Elt'c' lx. han
aL K.san.
fut
all redup fat the salmon at Ske€na
'The salmon of Skeenariver shall always be fat.'
(Boas, 1902,p. 20.15)
A similar pattern can be seenin Southern Paiute. Here, as in other UtoAztecanlanguages,both verbsand nounscan showpattial reduplicationfor
number. The rich textual material collected by Sapir, along with his grammatical sketch, show that reduplicative prefixes on verbs function as prototypical verbal number markers: They characterizeevents,not things. They
distribute actions over time, indicating iterative or continuous action, or over
space,indicating multiple locationsor participants.
(23)
Southern Paiute verbs
ivi'to drink'
i'i'pi'
'drinks repeatedly,sips'
nah'ipi:va'
'fire was burning'
nan a'bipiya'
'there werefires burning'
pn ava'severalgo homd
pampttnA'qai'
'(they) go home in parties'
b?wa'Ai'
'patchesond
tctc'q.wa:hi''patchesseveral'(Sapir, 1930,pp,258-261)
As in NassTsimshian,Southem Paiutenominal reduplicationis not a simple
inflectional plural marker. Sapir points out that reduplicated nouns are
distributive, "not plurals, though sometimes,particularly in the case of
animatenouns, practicallyequivalentto such" (1930,p. 257).
(24\
SouthernPaiute nouns
qa'nl
'housd
qaqqt'nl
'houses'
pt'
'trail'
pcvc'r
'trails'
piqwa'wife'
pivi'4wa.mi
'their (visible) wives' (Sapir, 1930, pp. 257-258)
Distributive nouns axeusually plunl, but not all plural nouns are distributive
Both nouns for 'horsd in (25) can be usedto refer to multiple horses,for
example,but only the secondis distributive.
222
(25)
Mafianne Mithun
SouthernPaiute
puqqutqwiraqwA'our(inclusive)hors€s(owned
collectively)'
bur horses(one or more owned
pumpu'4qu4wilm4wA
individually by each one of us)'
(Sapir, 1930,p. 258)
Similar examplescomefrom Maidu, a Califomia Penutianlanguage'Dixon
(1911)remarks in his grammatical sketchthat reduplication is quite frequent
in verbs, where it servesa distributive function'
Q6)
Maidu
witasvtitasiinoitsoia
'he went about picking here and there
(Dixon' l9ll' P. 689)
The samereduplicativeprocesescan be applied to nouns,adding a distributive
sense.Dixon lists four reduplicated nouns.
(27)
Maidu
scu'scuto
hobo'bota
ya' mdnmanto
tsa'tsato
'each,everyriver' (se'wi'river')
'every house, or camp' (hbbo' 'hortsd)
'every mountain'
'werY tred (ts4 tred)
He notes.however,that few nouns seemto havereduplicatedforms, and those
that do are rarely used.Thesedistributives "appear not to be usedin ordinary
conversationto any extent, and are rare in the texts. The aboveare practically
all the forms that havebeennoted" (19u' p. 689' 708).
The samesituation wasreported by Hoijer for Tbnkawa,a languageisolate
formerly spokenin central Texas'
repeated
Reduplicaiionaffectsverbthemesfor the mostpart. . . . It symboliz€s
Examples:
intense
action'
vigorous
or
rarely,
or
plural
subject,
action,
lintran$itive]
totop-'to c\t rcpeatedly'(top' 'to cut'), wawana-'severalfall forward' (wa'na'
'to fatl for*ud'). . . .In nounthemes,rcduplicatedformsoccurrately:'nantoton
'a nnge of mountain'(no.to, 'mountain'),lca'k-an'women' (tr'-a'n'woman'),
20 ?on 'blood veins'(?on 'blood') (1946,p. 297).
Beeler describeda similar systemin Barbareio Chumash. Reduplication
repeated,continuing action" (ms: 3l). "Nouns may show
'brdinarily expresses
initial reduplication accompaniedby glottalization and falling tone on the
final syllable;this formation, though it hasbeencalled a plural, appearsrather
to havedistributivemeaning" (ms: l7).
Tbkelma, a languageisolate of Oregon' exhibits the samecharacteristics'
Sapir (1922)reported that
TheEvolutionof NumberMa*ing
223
Reduplication is used in Takelma as a grammatical processwith surpdsing ftequency(p. 57).
Frcquentatives, continuatives, and usitatives are formed from simpler verb
forms in great part by various methodsof repetition of all or part of the phonetic
material of the stem. . . .The frequertative idea rnay have referenceto the repetition of the act itself (iterative or usitative) or to the plurality of th€ transitive
object or intransitive subject affected (distributive); any shary characterization
of the manner of the frequentative action in each caseis, however, doubdess
artificial apart from the context (p. 127).
Reduplicatednouns are not frequert in Takelma, particularly when otre considersthe great importance of reduplication as a grammatical devic€in the verb
(p. nQ.
A few adjectives form their plural or frequentative by reduplication. . . .That
theseplurals are really frcquentative or dist butiYein folce is illustrated by such
forms as da' kbloits./i/tf'il' 'red-cheeked,'which has referencenot necessarily
to a plurality of pe$ons affected,but to the frequencyof occurrenceof the quality
predicated (p. 264).
This pattem is common across North America. Reduplication of verbs
usually serves a prototypical verbal function of distribution. In the same
language, reduplication of nouns, when present, usually shares the distributive
function, rather than simply pluralizing entities. Nominal reduplication is
typically rarer than verbal reduplication, appearing only when the entities
it quantifies are scattered (trails, rivers, veins, houses, objects individually
possessedby different owners) or especially distinctive (people). Reduplication can also be extended to adjectives, still with the same basic function,
distdbuting the quality expressed over time, space, or individuals, rather than
over a static group as a whole.
2.1.2 Thefunctional shift of reduplicationover time There is evidencethat,
under certain circumstances,the function of reduplication can change over
timg according to nature of the host lexical category.Kwakiutl, a Wakashan
languageof British Columbia, exhibits typical distributive reduplication in
verbsand nouns. Boasnotesin his gmmmar that the reduplicatedverbal stem
purelytheideaof distlibution,of an actiondonenowandthen. . . .The
conveys
rather
of a noun€xpresses
Reduplication
ideaof pluralityis not clearlydeveloped.
the occu[enceof an objecthereandtherq or of differentkindsof a particular
object,thanplurality.It is thercforeruthera distributivethana trueplural (1947,
p. 206).
Compare the use of the simple noun for '\volf in the first sentencein (28)
with that of the reduplicated distributive noun in the second, for example.
224
(28)
M$ianne Mithun
Kwakiutl
ealdnE'ma
a. la' Eae qlc-nEma Ia enEma'dzaqwagEmottlaleda
'Then. it is said, many wolves howled all at the same time'
b, Ie sEinbEndxaena'la lEewa' gatnuLeSEmo'tedaqle'nEme
eeeal0genE'ma"
(singly)'
'Then the whole day and night wolves were howling
At the sametime, Boasdetectedan incipient shift in the function of nominal
reduplication:
a purelypluralsignificanceIn many
that this formis graduallyassuming
It seems
casesin which it is thus appliedin my texts,the older genemtioncriticizesits
useasinaccutate'Only in thecaseof humanbeingsis reduplicationappliedboth
p. nq'
as a plurat and a distributive(1947
'
He suggeststhat the shift may be due to the influenceof English:
Thereare a greatmanyexamplesof the useof reduplicatedplural forms for
inanimateobjects.Thepresentusageis certain,but weareunderthe impression
that the frequentuseof plural formsis dueto the influenceof English'There
in which the absoluteform is usedand we haveheardthe
a" ro ."ny
"""",
reduplicatedforms so often criticizedas unidiomaticthat we believethe old
grarnmarwouldrcquiretheabsoluteform for all pluralsin whichthesingleobjects
(1947p.293).
arenot individualized
'
2.2 Phonologically Constant Derivational Affixes
Thereare also many languagesin which verbs,nouns,and sometimesadjectives bear similar, phonologicallyconstantnumber affixes'
2.2.1 The functions of constant derivational affiscesin different lexical catesories Ai noted eariier, Central Pomo exhibits an elaboratearray of verbal
ievices reflecting number. Among them is a suffix -Ia- that might at fint
appear to indicate multiple subjects. [All of the Central Porno examplesare
from Frances Jack (personal communication)'l
Q9)
Central Pomo
ran mad{tmaE
ya nndtmat'lam
'I woke uP'
'we woke uP'
On transitiveverbs,-{z- appearsto correlatewith the number of patients'
(30)
Central
tsiitkdm
ts'iitldal<am
'fold it'
'fold lots of things'
225
TheEvolutionof NumbetMLrking
The suffix -tra-is actually a multiple event marker. Not all verbs involving
multiple participantscontain the suffix' In the secondsentencein (31)' the
prcrtottn m{tuluyal 'them: refers to a plural patient, but there is no multiple
event suffix, since a single went is predicated.
(31)
Central Pomo
a. 2aa mtlulu mandol'
pay semelfactive
I
him
paid
him.'
'I
b. taa m utuyal manftat'
pay semelfactive
I
them
(The
paid
work crew receiveda singlecheckjointly')
them.'
'I
The suffix appears only when severalseparateactions are predicated.
(32')
Central Pomo
zao mttuluyal mandataayfraw
pay multiple event multiple displacement aspect
I
them
'I paid them.' (Each worker was paid individually.)
Both verbs in (33) could describethe actions of severalworkers.Only the
second, which predicates multiple activities, contains the suffix -lr-'
(33)
Central Pomo
b€enmaw
'(group) carrying something together'
befui'
'(several)each carrying something individually'
Another Central Pomo suffix, -y-, indicates multiple displacement' It frequently co-occurs with the multiple event suffix.
3$
Central Pomo
sapdalu liqan
sapinlu liyqan
sapdalu liyq'Pma
'(one) putting on a shod
'(one) putting on a pair of shoes'
'(several)putting on their shoes'
The multiple event suffix and the multiple displacement suffix both function as prototypical verbal modifiers. They quantify aspectsof events.
Most Central Pomo nouns haveno plural forms. 'Plurals' can be elicited
for certain nouns referring to people,as in many languages,and for the noun
l6q' 'lhing.' The basic form of the plural suffix on nouns is -Izy (The dental
stop disappearsfollowing another consonant.)
(35)
CentralPomo
mdaln
mnalalay
Ioo t'dalzkre
too i'dall|{elay
'woman'
'women'
'my sibling'
'my siblings'
qanemat'
qantmat'ay
16q'
16qhy
'relativd
'relatives'
'thing'
'things'
226
Marianne
Mithun
The nominal suffix -Iry bears a striking resemblanceto the multiple event
suffix -ts-,possiblyin combinationwith the multiple displacementsuffix y-.
As in other North American languages,'plural' nouns are not used every
time multiple human beingsare discussed.In fact, they are more often absent
than present.When peopleact jointly, the suffixesnormally do not appear.
Note that the nounsarethe samein the singularand collectiveplural sentences
in (36).
(36) Central Pomo
tsoE' walda.
a. y6ohloht
south from person go sg continuativesg
'Therds a personcoming from the south.'
y6ohW
EaaE' hladda.
b.
south from person go pl continuative pl
'Therds a group of peoplecoming from the south.'
The 'plural' nounstend to appearwhenpeopleact individually,at different
times, in different directions,at different locations,and so on.
(37) Central Pomo
quadtezmaw.
a. Cdaihyya
person pl topic fight reciprocal multiple agent aspect
'Therds a bunch of people fighting.'
b. MEen 2e ihai;'awa
mta yhLIa.t'.
it is person pl topic stuff do multiple eventirnperfectivepl
so
'That's how peopledo things.'
-14l-hasretaineda distributive function in the relatively rare contextsin which
it appearson nouns. The fact that it appearswith the noun for 'thing' as
well as nouns referring to personsis consistentwith this role A term for 'all
sortsof things' is useful in a languagewithout inflectionalnumbermarking
on nouns.
-Iay- appearswith adjectivesas well, where it also servesas a distributive
Like the Tsimshianreduplicativeprefixes,it appearssomewhatmore frequently
with adjectivesthan with nouns, becausethe qualities specifiedby adjectives
are usually distributed individually over members of a set, rather than
attributed to the set as a whole.
Maidu contains a similar number suffix that appearswith both verbs and
nouns.Dixon notes,"this suffix, of generaland very frequentuse.. .is used
in sorne casesto indicate iteratioq in others, reciprocal action; at times it
seemsto point to a plural object" (19U, p. 705).
(39)
Maidu
moLtdn
yo'k-a-fin
'to drink repeatedly' (mon 'to drink')
'to strike repeatedly with fist'
TheEvolutionof NumbetMa*ing
yapai'-to-to4om
(Yo'k-on 'to strike)
p. ?05)
'talking to eachother' (Dixon, 1911,
He found that -to was "also us€das a nominal suffix in connection with the
reduplicated distributives." Recall that although reduplicated verbs are pervasive in Maidu, reduplicated nouns are rare'
(39)
Maidu
tsA'-Ba-to
P. 705)
'tr€es'(Dixon,1911,
In Maidu, as in Central Pomq the suffix retainsits distributive fuhction even
when applied to nouns'
Coos o<hibitsa similar pattem. Frachtenbery Q922,p' 327)describedseveral
verbal distributive suffixes: "All the suffixes expressingdistribution havethe
element,t- in common, which consequentlymay be regardedas the original
suffix conve,yingthe idea of distributive plurality; the nore so, asin the following instancesn- actually denotes distribution,"
(,10) Coos
a. dii
lE dji'ntt
b. cin€tt' k'E
tsEl:nctt$
nkine
'it camd
'they came (singl9'
'you stand!'
'side by side they two were standing"
(Frachtenberg, L922, P' 327')
A similar suffix, also labeled a distributive by Frachtenberg, appearswith
nouns, although much more rarely. riru-is suffixed to nouns of relationship
only and exprissesa degreeof mutual kinship. It is etymologically related
to the verbal distributives-ne', -iinl (1922'p.371).
(41) Coos
a. sla'48
Ax sb'tcini
b. hd'Liitc
i-n hdLtci'ni
tousin'
'they two I'ere mutual cousins'
'elder brother'
'we are brothers mutually'
(Frachtenberg, l92Z' P' 371)
In the limited caseswherethe distributive suffix appearswith nouns, it retains
its primary function.
Karok also showscross-categorynumbermarking. Recallthe verbal multiple
event suffix -va illustrated in exanples (?) through (10)' The suffix appears
with other kinds of lexical items as well, still with the multiple event or
distributivesense.
228
Mafianne Mithun
(42) Karok
a. oxyam
axyard-vah
b. i9A.kmram
iqe.kwrdm-vah
'full'
'all full (of various vessels)'
'one night'
'night after night' (Bright, 1957,p. 82)
Distribution is more often pertinent to €ventsthan to entities,and correspondingly more likely to be marked on verbs than on nouns and adjectives.As
the Central Pomo, Maidu, Coos, and Karok data show,however,distributive
markers can fuse with adjectivesand nouns as well, while still retaining their
original function.
2.2.2 Thefunctional shift ofderivational afiixes over time There is evidence
that the functions of distributive affixes can change after they havebecorne
morphologicallybound, under certainconditions.The Iroquoian languages
contain severaldistributive verbal suffixes. As in other languages,the suffixes function to distribute actionsover time, space,or participants.This effect
can be seenin the verbsin (43) from Cayuga,a Northern Iroquoian language
of Ontario. [All of the Cayuga examplesare frorn Reginald Henry (personal
communication).1
(43)
Caluga
ehsye:thoz
'you will plant'
ghsyethwahsg:') 'you will plant a lot of different things'
Most Iroquoian nouns referring to human beingscontain pronominal prefixes
indicating the gender and number of their referents,but number is not normally specifiedotherwise in nouns. Severalsuffixes, however,clearly cognate
with the verbal distributives, do appear on nominals.
(44) Cayuga
a. kanyo:2
lanyotshi:tqh
b. mehsgnydztha2
enahsenyazt hozI h0:, I h
c. akitkwrwetaz
sketlcvrtetfitshgt
'wild animal'
'game'
bne builds houseswith it, tool'
'housebuilding tools'
'my suitcase,my handbag'
'my baggagd
The suffixes do not appear every time multiple entities are referred to; th€y
specifya vari€ty of assortmentof kind. The term translated'tools' would
not be used for severalidentical hammers.(The words for 'gamd and 'tools'
aremorphologically verbs,but the suffix modifies them aslexicalizednominals
rather than etymologicalverbs.The term for 'tools' doesnot mean bne builds
various houses\Mithit.')
Distributive suffixes do appear systematicallywith certain terms referring
to more than two people They alwaysappear with plural terms for thildren'
or bld people,'for example,although not with duals.
TheEvolutionoJNumberMarking
(45)
Cayuga
thild, girl'
ekd:|ah
Iuelcstnh,:eoh thildrenl
229
bld man'
hakihtsih
kaelehtslhshge'oldpeopld
The suffix also appears with certain kin terms wheneverthey are plural'
(46)
Cayuga
lukhehtsizlhshg'
'thev are my older siblings, older siblings'
'my cousins, we all are related'
akwaten1hkshgt
kakhdhsotshgt
'my grandparents'
I<nkheyatre'tsh0:toh'mygrandchildren'
kaqnthdwgkshq?'theirchildren(offspring)'
with thesenominals, the force of the suffix has moved from distribution to
plurality. This is not surprising; human beingsare often consideredinherently
individualistic and differentiated.Still, the suffix has not becomea full-fledged
inflectional plural marker. It does not appear systematically with all nouns
referring to Lu-"ot ot evento kinsmen. It is lexically governed' The basic
nouns for 'men' and 'women,' for example, do not contain it.
(47)
Cayuga
ok,:kweh
kd:kokweh
'woman'
'women'
hokweh
honh:hleh
'man'
'men'
Although thesepatterns are normally maintained by all Iroquoian speakers
in spontaneousspeech,the influenceof English can be seenon occasion,as
teacherswork at normalizing grammatical patterns for language curricula'
As they search for paradigms on English models, younger, more Englishdominant speakers som€times take advantage of the distributive as an
equivalentof the Engtishinflectional plural. They still do not usedistributives
as general plurals in spontaneousdiscourse,however,
Another example of a shift in the function of a derivational affix comes
from Chinook, a Chinookan languageformerly spoken in southern coastal
Washington.When Boasrecordedthe languagein 1890-1891,there wereonly
two remainingspeakers.In his grammaticalsketch,he describeda verbalsuffix
-Em: "distribution at distinct times. . . followed usually by -r (customary)"
(l91la, p. 596). Th€ texts he recordedamply show that the suffix functioned
as a verbal distributive.
(48)
Chinook
a-tc-L'k4atE'qo'im-x
a-Lg-io-Wa'M-Em'x
a-L-x'a''x'um'x
'he always stood on them severally'
'he hides it everywherd
'they always did here and therd
(Boas, 1894,1911a,p. 596)
Both verbs and nouns in Chinook contain pronominal prefixes
distinguishing person, gender,and number' In the verbs in (48), the prefixes
230
Ma ame Mithun
following the aorist d- refer to ergativeand absolutivearguments.Pronominal
prefixes similar in form to the absolutives distinguish masculine, feminine,
neuter (indefinite), dual, or plural nouns. Gender is no longer biological,
although Boas detected enough of a pattern to conclude that it once was.
Compare the prefixes below: i- masculine singular, o- feminine singular, and
f- plural.
(49)
Chinook
r'l@la
*alauks
iAoa-inc'nc
t-qoa-inc'nc
'man'
'men'
'beaver'
'beavers'
bld man'
i-qleyo'qxut
o-qleyo'qxut'oldwoman'
qleya'qfikc
bld peopld
a-hd'tlau
'virgin'
t-hatlaund'na'virgins'
d'kci
t-l{ci
'finger'
'fingers'
In addition to their pronominal prefixes, some plural nouns contain suffixes. Nouns for human beings bear a suffix, -ikc or'ukc, as in the words
for 'men' and bld peopld above.A particularly interesting suffix on many
plural nouns is -ma. Boas consideredthis nominal suffix to be cognate with
the verbal distributive 'Em'x: "the frequent plural-suffix -r 4 (Kathlamet mdx) seemsto havebeen originally a distributive element' This appearsparticularty clearly in the wordse.fiEmae'sometimes' (et't 'ondi -rnadistributive;
4 adverbial);luna' mtEnw 'botlf (knna'm 'both, togelher'; -ma distributive)"
(Boas, 1911a,p. 608).
The textual material from Chinook showsthat the distributive force of the
suffix had been generalizedtoward pluralization by the time Boas recorded
it. Nouns that have-rza forms usually appear in the texts with the -rza when'
ever multiple entities arc mentioned. These are only a subset of the nouns
in the language,howwer, and they tend to refer to entities especially likely
to occur distributively, such asislands,towns, houses,things, warriors, rocks,
days,or elk.
The functions of number markers can thus shift after they have become
bound to nouns,toward a distinction that is more generallypertinent to nouns:
pluralization. It is easyto seehow the shift might come about. Distributive
markers on nouns tend to recur with certain semantic classes:those that are
highly differentiated. Since human beings are often eonsideredinherently
individuated, highly distinct by their very nature, nouns referring to them
might easilybear distributive markersnearly everytime multiple human beings
are involved. The reinterpretation ofthe distributive marker as a plural under
suchconditionsis a short step.It alsoservesa useful function: Speakersand
hearersare usually more interestedin the number of peopleinvolved in events
than in the number of objects.Data from Kwakiutl, Cayuga,and Chinook
TheEvolutionof NumberMatking
231
indicatethat the shift may continue one step further under specialconditions. Under heavy influence from a language like English, wliere plural is
an inflectional category of nouns, number marking may be reinterpreted as
a more general category and extended to more contexts.
3. The Predominanceof Verbal Number Markers in North America
North American number-markingsystemsstill remain strikincly different from
thosefound in many other parts ofthe world. ThroughoutNorth America'
productive,and elaborate
verbal number rnarkersare generallymore perv-asive,
than nominal number markers. The nominal number markers that do exist
are usually verbal in origin and in character,signalingdistribution rather than
simple plurality. These tendencies are probably not unrelated to other
characteristicsof the languagesthemselves.A surprising proportion of North
American languagesshare certain structural traits, although they are not
demonstrably related genetically. As a whole, they are predominantly verb
centered.Verbsare typically elaboratemorphologically, containing bound pronouns referring to core arguments, as well as numerous locative, temporal,
and other adverbial distinctions, and often even incorporated noun stems'
Verbs can usually function as self-contained sentencesin themselves'When
they contain third person referential pronouns, separatenominal constituents
are not syntactically bound to verbs as they are in English' They function
more as appositivesto the bound pronouns, supplying identification where
necessary.It is not surprisingthat theseappositional nominal adjuncts should
fail to be inflected systematically for number'
4. Inflection, Derivation, and the Lexicon
Number distinctions in North America are typically not only verbal, they are
usually lexical and derivational rather than inflectional. This is no accident'
As Bybee(1985)points out, inflectional distinctions arc obligatorily specified
on all membersof a lexical classor subclass.In English, number is marked
inflectionally on all count nouns, and t€nseon all finite verbs. We know that
every singular count noun has a plural counterpart, and every presenttense
verb has a past tensg although we may not know exactly what the forms are'
Inflectional categoriesmust therefore be sufficiently general in meaning to
be applicableto all m€mbersof a lexical classor subclass.
The distinction between singular and plural number may be equally
applicableto all nouns,but it is not equallypertinentto all verbs'.Forsome
events,number can be centrah One bird flitting by can be conceivedof as
a different event from a flock flying by in formation; picking up one object
Ma anneMithun
can be classified as a different motion from gathering up a lot of objects.
Such distinctions can be sufficiently important to be encodedlexically, and
in North America, eventsof this tlpe are frequently expressedby meansof
different verb stems.Such classificationscan of coursebe culture specific,
but the similarities acrossNorth America are striking' Languagesdiffer more
in the number of verb roots that encode number than in th€ types of verbs
that do. Most commonare verbsof position ('sit,' 'stand,''lid), of handling
('put,' 'take, 'pick up,' 'drop,' 'throw'), of basicmotion ('go,' 'walk,' 'run,'
'fly,' 'swim'), and killing and dfng. Also typical are ceflain social activities,
such as conversing,eating, and crying (mourning).
In many cases,number may not be a sufficiently central aspectof an event
to be encodedin a distinct verb root, but it may be pertinent enough to
motivate the creation of a speciallexical itern. Hitting a man once and giving
him a beatingmight be classifiedas subtypesof the samebasic activity, so
with verbsbasedon the sameroot. If the repeated
they might be expressed
hitting correspondsto a recognizable,conceptually unitary event, speakers
might exploit their derivational morphology to createa word for this concept.
Similarly, doing laundry may be simply categorizedas washing, or a special
term may be created, perhaps with multiple event suffix, for this special
thesedevicesvary in productivity from
activity. Like all derivational processes,
languageto language
Verbal number distinctionsrarely becomeinflectional, becausethe pertinence of number varies so greatly from one type of event to another' As
Bybeepoints out, oncethe most important number distinctions are expressed
lexically, and other significant number distinctions expressedderivationally,
thereis little motivation to extendnumber specificationto the remainingverbs
in a language.
Basically verbal number distinctions are sometimesextendedto nouns, but
they seldom spread to all nouns, becauseof their specializedfunction'
Distributivemarkersretain a distributivemeaning,servingto emphasizethe
of the entities referredto by nouns. In somecases,
distribution or separateness
they eventually appear with all nouns referring to nultiple human beings,
since people can be consideredinherently individualistic and differentiated.
At this point, they can function essentiallyas plural markers,but still, only
within the range of nouns referring to people This is not an unstable state;
the number of human beingsinvolvedin an eventis generallymore interesting
to speakersthan the number of inanimateobjects.Evenlanguageswith no
inflectional or derivationalnumber marking at all on nouns often contain
separatenoun stemsfor 'person:and 'peoplej 'man' and 'men,' 'woman' and
'women.' thild' and thildren.'
5. Conclusion
Number is thus not necessarilya nominal categoryor evenan inflectional
one. In many North American languages,it is primarily a verbal category:
TheEvolutionof NumbetMa*ing
233
Number markers appear predominantly on verbs and quantify aspectsof
events.On occasion,basically verbal number markers are extendedto nouns
and adjectiveswhereappropriate,but they retain their basic functions, usually
marking distribution.
Oncedistributives are extendedto nouns, their functions can shift' Because
people can be viewed as inherently differentiated, distributives often come
to appear with all nouns referring to multiple human beings. At this point,
the markersmay be reinterpretedas plurals, but only in this restricteddomain.
The next step, the extension of the plual function to all nouns, apparently
occurs only under the heavy influence of another language with a general
inllectional plural, such as English.
The functional difference between nominal and verbal number markers
r€sults in a formal difference elurality is a sufficiently general feature of
entitiesto be applicableto all count nouns. It is thus often inflectional. verbal
number distinctions, by contrast, vary widely in their importance and
applicability to events.They rarely become inflectional. Instead, they are
encoded lexically, derivationallg or not at all.
Acknolvlegments
I am grateful to Mrs, Fraoces Jack, of Hopla[d, Califomia, for sharing her expertise on h€r
languagq Central Pomo, a4d to Mr. Reginald HeNy, of Six Nations, Ontario, for sharing his
expertise on his languagq ca''uga.
Reterences
AppleSate R.B. (1972).Ineefro Chum6h gamrrl4l. Ph.D dissertatiotr, Univetsity of Califomia,
Berkeley,
Barker, M.A.R. (1964).Klamath grammar. U"i€rsity of Calfomia Publicationsin Linguistics 32,
Beeler,M.S. (1987)."Topics in BarbarcfroChumashcrarnmar." UtdveNity of Califomia, B€rkeley.
Boas, F. (1894).Chinook texts.Bureau of Amefican Ethnolog/ hr etin 20. GovernmentPrinting
Oifice, Washington D.C.
Boss, F. (1902).'Ifimshiantexts. hneau of Ameri.4n hhnologt htll*in n. Goverffnent Printing
Officq Washington D.C.
Boas,F. (lgua). Chinook. Handbook ofNorth Americad Irnguages, PaIt l- h)rcau ofAme can
Ethnolog) Rulletin 4, 559-6?8. Government Prit$ing Officg washington D.C.
Boas,F. (lgllb). 'Ifimshian. Handbook of North Am€dcar laJigJuaes',Padtl. hrcau ofAnefican
Ethnolog Rulletin 4,283-422. covernmettt Printing Office, Washi4ton D.c.
Boas, F. (1947).Kwakiutl crammar. Tl,ftsqctions of the Amedcsn Philosophical Society, New
Se/iet 37. Part 3.
Bdght, w. (1957).The Karok language Unive ity of Cafumia htblic4tions in Linguistics 13
Btb.r, J. 0985). 'Morphology. A Study of the RelationbetweenMeaning and Form." Beojamins,
Aog€rdam.
Cooli, E--D. (1984). '1q.Sarc€€Grammar," Univ. of British Columbia Press, Vancouver.
IXa, R.B O9ll). Maidu, Handbook of North American lanSuages,Part L hMau of AnE can
FLa!-h) ain 4,679-?34. covernment Printing Officq washinglon D.c.
234
Matianne Mithun
c,:tegory.Prcceedingsof the Twlfth
Durie, M. (1986).The grammaticizationof [umber as
^rcrbal
355-310.
Annurl Meeting of the krkzley Linguistics Society
Frachtenberg,L.J. (1922).Coos. Handbook of Nofih American Langrages, P^rt 2. Bweau of
Amefican Ethnalogy Ruletin q, 297430. coverfinent Printing office, washington D.c.
FEjzyngier, z. (1985).Ergativity, numb€r, and a$eetne[t. Prcceedingsof the Eleventh Annual
Meeting of the Be*cley Linguistics fuciety 96.106'
cibsod, J.A. (1973).Sluswap gamrfutical structue Ph.D dissertation, University of Hawaii.
Hoijer, H. (19,16).Tonkawa. "Linguistic Structures of Native Americaj' pp. 289-311.WennerGren Foundation for Anthopological Research,New York.
Kimball, c.D. 0985).,{ des ptive gammat ofKoat4ti Ph.D. dissertation, Iulane Universitv.
KiI}aid,M.D-(1975).PhMliutioninUppetCheMb.Paperreadatthe1975SalishConference,
Ne$rnan, S. (1965).Zuni Gn maL Univetsity of New Mexico h.tblications in Anthropolog/ (14\. '
Robins, R,H. (1958).The Yrrfok language University of Califomia htblicatons in Unguttics 15
Sapir, E. (1922).The Tbkeloa lanerrageof southw€stemOregon. Handbook of American Indian
lAnfjt ges,Paft 2. Rueau of Ametican Ethnowy Bullain 4, 1-296. Government Printiog
Office, Washitrgton D,C.
Sapir, E. (1930).SouthernPaiute, a Shoshoneanlaoguagelixts of the Kaibab Paiutesand Uintab
of the Ameican Acadeny of Atts and Sciences
Utes. Southem Paiute Dicti onary.MW
65, l, 2, aDd 3.
swanton, J.R. (l9lr). Haida. Handbook of American Indian Languages, Part l. Bureau of
Aneican Ethnolog/ Bulletin 4, m5-282. Governm€nt P nting Office, washiogto4 D.c.
watkins, L.J. (1984). 'A Crardmar ol Kiowa." University of Nebraska Press, Lincoln.
i
| "'i
I
I
I