in K i s w ah ili Cla s s e s
A s s ib i A . A m i du
(N orw e g ian U niv ers ity of S c ienc e & T ec hn olog y , T ron dh eim )
A m idu , A . A s s ibi . 200 2 . T he P arad ox of N um be r and N on - n um b e r in K i s w ah ili Cl a s s e s . T he L ing uis tic A ss ociation of K orea J ournal, 10 (4), 149- 178. Within the Bantu ling uistic w orld, g enetic clas s sy stem s st and as centr al pillar s of their com munication s y stem s . In this study , w e arg ue th at , in clas ses , s ocial or g anization and n atur al kind sy stem s , ther e ar e an om alies th at do not confir m entir ely the st andar d division of num ber as sing ular v er su s plur al pair s of affix es in B antu . W e sug g est th at a pr inciple of n on - number is res pons ible for such a conclusion . T o achieve our g oal, the study w ill ex am ine the r elation ship betw een Kisw ahili g r amm atical number , the s ocial org anization of countability , and the Bantu view of n atur al kinds . W e dr aw attention t o contr adiction s inv olving n on - num ber in the sy s tem s and sugg est th at clas s affix es are n ot inherent num ber affixes . T he an om aly is par tly r es olv ed in Bantu clas ses only w hen n on - number is des cribed as a neutr al or central categ ory . We conclude th at Bantu clas ses ar e prim arily clas s ifier s of n atur al kind s of object , and only secondar ily deter miner s of number .
K e y w ord s : Kisw ahili, Bantu , clas ses , num ber , g ender
1 . In t ro du c t i o n
Bantu is a clas s language family , but it is not alone in expr essing it s univ erse of r eference through genetic class or classifier m arkers . Lakoff (1986) and Dix on (1968, 1982) hav e w ritt en about the Dyirbal class or classifier sy st em and it s s em antic im plication s . Wilkins (2000) and Aikhenv ald (2000) hav e recently com e out w ith descriptions of classifier/ class sy st em s laying emphasis on their gramm atical an d sem antic cat egories .
1)Both class and classifier languages ar e gener ally
1) Allan (2001, p . 116) h as rig htly obser ved in rev iew ing A ikhenv ald (2000)
referred t o as either clas s or classifier syst em s . S om e linguist s , how ev er , refer t o Bant u languages as class sy st em s as opposed t o classifier sy st em s (Craig , 1986; Amidu , 1997; Aikhenv ald, 2000). In this study , I use t he t erm s int erchan geably . I shall restrict myself, how ever , t o Kisw ahili an d Bantu evidence.
T he st udy argues that t here is a high count able or pr obability r atio in natural kind sy st em s such that obj ect s m ay be paired as one v ersus m ore than one, and their linguistic t erm s m ay also be paired as singular ver sus plural in a predict able m anner . But there is n o direct correspondence betw een class and number in Bantu . I begin , how ever , by claiming that the concept ual organization behind Bantu classes and their univ erse of r eference, as seen through Kisw ahili, em anat es from tw o thin gs : 1. T he social organization of the family am on g Bantu people, 2. T he Bantu univ ers e of nat ural kinds of obj ect s . I view the Bantu family simply in Durkheimian t erm s (see §§ 2.- 3. below ), and so I w ill not go int o an ant hropological description of w hat or how it looks like. Within these t w o param et ers , w e discov er that t here are lim it ations t o a one t o one m apping of social count abilit y and natural kinds in a natural language like Kisw ahili Bantu , particularly w ith regar d t o a ) number and b ) sem antic assignm ent rules . T his study w ill only look at problem s of number description s in Bantu .
2 . A t h e o ry o f s o c i al t h in k in g a n d o rg a n i z a t i o n
S ocial ant hropologist s hav e long draw n att ent ion t o the fact t hat hum an bein gs are organised int o communities , n ot so much for the sake of the individual but for t he surviv al of the group or fam ily as a w hole (Lienhardt 1964, 1966).
2)According t o Lienhar dt (1966), Durkheim w as
th at "A m azonian lang uag es are g iven dispr oportion ate pr om inence in the book ."
2) I w ish t o th ank Ar ne Kjell F oldvik of the Depar tment of Ling uistics , Norw eg ian Univer sity of Science and T echn olog y , T r ondheim , for giv ing aw ay his book on s ocial or g anization . T he b ook h as contr ibuted im m ens ely t o the hypothesis in this study and in my recent s tudies of clas s es . I als o th ank w arm ly Ali H . M ar oug a of T r ondheim and Abdulaziz Y . Lodhi of Upps ala
the first t o point out that a st udy that proceeds from the individual t o the m ajority is likely t o miss out the fact t hat t he in dividual is him self a pr oduct of t he society . He or she is , as a result , incapable of representing the t ot al goals and salience of that society . A st udy of hum an society should, therefore, be based not on t he individual, but rather on t he 'collectiv e r epresent at ions ' , that is , on the fact that
"Different societies exhibit different p atterns of thought , different 'collect iv e r epresent at ions ' ..." (Lienhardt , 1966, pp. 31). Lienhardt explains t he doctrine of div ersit y in a lucid m ann er as follow s : "In general, the F rench s ociologist s of Durkheim ' s school est ablished convincingly that social tradition m oulds the individual conscience m ore fully than ev en the m ost self- conscious m embers of a society usually recognize."
T he concept of difference an d diver sity in pat t erns of thought and hence of 'collectiv e repres ent ations ' m ay appear t o rule out a univ ersalist explicat ion of the em ergence of t oday ' s Bantu classes . An int erpr et er m ay also as sum e t hat differ ence and div ersity m ean that there are no unifying threads that link language syst em s t ogether . T hese conclusions are, how ev er , not justifiable in Bant u or in linguistics generally as typological studies show (Greenber g , 1966, H eine and Reh , 1984, Croft , 1990). W e stres s , how ev er , that Bant u classes can best be studied from a position of an underst anding of w hat Krapf (1850) called the 'Nilotic Idiom ' (NI). T he NI is a collect iv e idiom that repr esent s t he African ' s conceptualization of his or her univ erse of reference. An int erpr et er w ho does not underst an d t he ' collectiv e represent ations ' in the NI cann ot truly do just ice t o the w ay the African perceiv es the w orld and how this is reflect ed in and through his linguistic strings . T he linguist ic strin gs ar e them s elv es deriv ed from his ant ecedent usage.
It should be n ot ed t hat the affirm at ion of difference and div ersity of thought is a non - racial st at em ent , t hough pedlars of racial doctrin es could s eize upon it for their ow n goals (Lucy , 1992, ch . 1, on relativity ).
Univ er sity for ver ifying s ome of m y dat a .
3 . S o c i al o rg a n i z at i o n in c l a s s o rg an i z a t i o n : t h e c o n c e pt o f g ram m at i c a l n u m b e r
F ollow ing from the Durkheimian School, w e might say t hat , essentially , hum an societies consist of individuals an d a collectiv e represent ation of in dividuals . T he collectiv e represent ation is the family , a v ery import ant and indispensable landm ark of hum an social organization and exist ence. T he Bantu family is , t herefore, a collect iv e represent ation of individuals . In Amidu (1997, ch . 9), t he relationship bet w een individuals and their families w as r educed t o a sy st em of ont ological opposition (I) for individual obj ect v ersus (F ) for family of individual object s .
3)T he opposition reflect s the fact that the family organization of Bantu m an and w om an as a collective is reflect ed also in the linguistic organizat ion of their speech . T his can be seen in the fact that t he genetic classification of t heir gr amm ar att empt s , albeit oft en unsuccessfully , t o correlat e som e linguistic t erm s w ith (I) and other lin guistic t erm s w ith (F ). W e also discov er that som e it em s are neither (I) nor (F ) obj ect s . T he gramm at ical represent ations of the family or collect iv e represent ation , on the one hand, an d the individual represent ation , on the other hand, are kn ow n lin guistically as the opposition of singular ver sus plural. T he opposition (I) v ersus (F ), therefor e, r eveals a direct impact of social organization on language organization , in m y view . T his does not imply that one w ill alw ay s find a one t o one opposition or correlation betw een language and fam ily or collectiv e represent ation in the society or the univ ers e of r eference.
3 .1 . Count ability an d c las s c las s ific ati on in B antu
If w e look at the dat a (1)- (16) below , w e discov er that the opposition
3) T he ter m ' indiv idu al ' should be inter preted br oadly t o include all objects and pos sible objects w ith independent exis tence as opposed to the collectiv e exis tence of their fam ilies of object . In this sen se, ' indiv idu al ' is n ot r estricted t o hum an being s as object s .
(I) v ersus (F ) in social organization underpins much , but n ot all, of Kisw ahili class organization , t oo.
(1) a . Class 1 MU 1
b . m - s ichana y u- le m - ref u a- m eolewa m - yu - m - a-
Cl. 1 she- girl Cl. 1 she- that Cl. 1 she- t all Cl. 1 SM she- RE CENT PA ST - m arry - PA SS .- MOD .
'T he t all girl is m arried.' (2) a . Class 2 WA
b . wa- lev i wa- le wa- w ili wa- naim ba w a - w a- w a- w a-
Cl. 2 they - drunkard Cl. 2 they - that Cl. 2 t hey - tw o Cl. 2 SM they - PRESENT - sing - MOD.
'T hos e tw o drunks ar e singing .' (3) a . Class 3 MU2
b. m - to w - etu m - k ubwa u- m ef urika m - w - m - u -
Cl. 3 it - riv er Cl. 3 it - our Cl. 3 it - big Cl. 3 SM it - RE CENT PA ST - ov erflow - ST AT IVE - MOD.
' Our big riv er has ov erflow ed (it s banks ).' (4) a. Class 4 MI and Class 1 MU 1
b. mi- iba hi- i my - embamba i- li- m - choma m-p ishi mi- i- my- i- | - m- m-
Cl. 4 they - t orn Cl. 4 this - t hey Cl. 4 they - t hin Cl. 4 SM they - PA ST - Cl. 1 OM she- prick- MOD. Cl. 1 she- cook
' T hese slender thorns pricked t he cook .' (5) a . Class 5 JI
b. j i- cho l- ak e m oj a li- naum a ji- l- Ø - li- Cl. 5 it - ey e Cl. 5 it - his Cl. 5 it - one Cl. 5 SM it - PRESENT - pain - MOD .
' One of his ey es sm art s w it h pain ' , Lit . his one ey e sm art s w ith pain .'
(6) a . Class 6 MA 1
b . m a- g om bano y - ao y - a j uz i y a- m ekw isha m a - y - y a- y a- Cl. 6 they - quarrel Cl. 6 they - their Cl. 6 they - of Cl. 0
ø- day - befor e- y est erday Cl. 6 SM they - RECENT
PA ST - ST RE SS AF X.- finish - MOD.
'T heir quarrel/ s of the day befor e yest erday has/ hav e been buried.'
(7) a. Clas s 7 KI
b. ch- ak ula hi- k i ch- ote k i- m ehabirika ch - - ki ch - ki- Cl. 7 it - food Cl. 7 this - it Cl. 7 it - all Cl. 7 SM it - RECENT PA ST - spoil- ST AT IVE - MOD.
' All this food has gone bad.' (8) a. Clas s 8 VI
b. v i-j ik o vy - ang u vy - ote v i- m eibwa vi- vy - vy - vi- Cl. 8 they - spoon Cl. 8 they - my Cl. 8 they - all Cl. 8 SM they - RECENT PAST - st eal- PA SS .- MOD .
' All my spoons hav e been st olen .'
4)(9) a. Class 9 NI1
b . k eng ele y - etu m - pya i- nalia Ø - y - m - I- Cl. 9 it - bell Cl. 9 it - our Cl. 9 it - n ew Cl. 9 SM it - PRESENT - cry - MOD.
' Our new bell is ringing .' (10) a. Clas s 10 NI2
b . tend e z - enu tam u z i- m euz wa Ø - z- Ø - zi-
Cl. 10 they - dat e Cl. 10 they - your Cl. 10 they - sw eet Cl. 10 SM they - RECENT PAST - sell- PA SS .- MOD .
' Your sw eet dat es hav e been sold.' (11) a. Clas s 11 U 1
b . u-f a u- le m - k ubwa u- taz ibwa u - u - m - u -
Cl. 11 it - crack Cl. 11 it - that Cl. 11 it - big Cl. 11 SM it - F UT URE - st op up- PA SS .- MOD .
4) N ote th at - ibw a ' be st olen ' is a pas siv e of the basic predicate ver b - iba ' s teal ' . It is u sed in the N orthern dialect s of Kis w ahili, such as Kim vita, but it is not u sed in a S outhern dialect like Kiung uja, accor ding t o Ali H . M ar oug a of T r ondheim , and A bdulaziz Y . Lodhi of Upps ala Univer sity , both nativ e s peaker s of Kisw ahili fr om Zanzibar . In the Kiunguja dialect of Zanzibar , the applic ativ e pas sive form - ibiw a ' be st olen fr om ' is r ather u sed . A s a r esult, the pas sive of the sim ple pr edicate v erb - iba is , str ictly speaking , often unu sed in Kiung uja .
' T hat large crack w ill be st opped up.'
(12) a. Clas s 12 KA (dorm ant and no longer used activ ely ) b . ka- toto ka- le ka- dog o ka- nalia ka- ka- ka - ka-
Cl. 12 it - child Cl. 12 it - that Cl. 12 it - sm all Cl. 12 SM it - PRESENT - cry - MOD.
' T hat sm all child is crying .' (13) a. Clas s 14 U2 an d Class 2 WA
b . u- shirika u- takatif u hu- u u- ta- wa- im arisha wa- k ulim a u - u -
u - u - | - w a- w a -
Cl. 14 it - communion Cl. 14 it - holy Cl. 14 t his - it Cl. 14 SM it - F UT URE - Cl. 2 OM they - be firm - CAUS .- MOD. Cl. 2 they - farm er
' T his Holy Communion w ill fortify the farm ers .' (14) a. Clas s 15 KU and Class 2 WA
b . k u- iba k w - ak e k u- le k u- li- wa- aibisha wa- g eni ku - kw - ku - ku - | - w a- w a-
Cl. 15 it/ they - thievin g Cl. 15 it/ they - his Cl. 15 it/ they - that Cl. 15
it/ they - PA ST - Cl. 2 OM they - sham e- CAUS .- MOD. Cl. 2 they - guest
' His thieving/ s abashed the guest s , Lit . that/ those his thieving/ s
embarr assed the guest s .'
(15) a. Class 16/ 25 MA 2, also
traditional Cl. 16 PA - , and Proclass 2/ 1 1st Per s .
b . m a- hali p a- le p a- k ubwa p a- na- tu- tisha s is i m a - pa- pa- pa- | - tu -
Cl. 16 it/ they - place Cl. 16 it/ they - that Cl. 16 it/ t hey - big Cl.
16 SM
it/ they - PRE SENT - ProCl 2/ 1 OM w e- fright - CAUS .- MOD.
ProCl. 2/ 1
w e that/ t hose large place/ s fright en/ s us .'
(16) a. Class 17/ 26 NI3 , also
traditional Class es 16. PA - , 17. KU - , 18. MU -
b . ny um ba- ni ha- p a p - ote p a- m ep akwa rang i - ni pa- p- pa-
Cl. 17/ 26 house- there Cl. 17/ 26 t his - there Cl. 17/ 26 ther e- all Cl.
17/ 26 SM there- RECENT PA ST - paint - PAS S .- MOD. Cl. 9 it - paint
' T his entir e house has been paint ed.' Lit . In and out of this entire
hous e has been paint ed paint .
c . ny um ba- ni hu- k u k - ote k u- m ep akwa rang i- ni ku - k - ku - Cl. 17/ 26 house- there Cl. 17/ 26 t his - there Cl. 17/ 26 ther e- all Cl.
17/ 26 SM there- RECENT PA ST - paint - PAS S .- MOD. Cl. 9 it - paint
' T his entir e house has been paint ed.' Lit . In and out of this entire
hous e has been paint ed paint .
d. ny um ba- ni hu- m u m - ote m - m ep akwa rang i- ni mu - m - m - Cl. 17/ 26 house- in there Cl. 17/ 26 this - in there Cl. 17/ 26 in there- all Cl. 17/ 26 SM in there- RECENT
PA ST - paint - PA SS .- MOD.
Cl. 9 it - paint
T he int erior of the entire house has been paint ed.' Lit . In this entire house has been paint ed paint .
Each predicat ion - sent ence (Pn - S ) repr esent s a class syst em or part of a class syst em .
5)W e discov er , therefore, that there ar e basically 16
5) F or each predic ation - s entence ty pe, the clas s ifier/ s of the clas s , e .g . MU 1 and/ or MI, etc ., underly ing the string is/ are the actu al g ener at or/ s of the string con struction , and hence of the predication - sentence . Im mediately below every deriv ed Pn - S in the dat a abov e, therefor e, w e find it s clas s sig nificant con stituent syllablic unit s (S CSU s ), e .g . w a - w a - w a - tr ig ger ed by clas sifier WA of clas s 2, etc ., w hich determ ine syntax em ic function s or str ing con stituent
or m or e distinct predication - sent ence t ypes in St andard Kisw ahili, ex actly as illustrat ed by (1- 16) s up ra. In a dialect like Kisiu , w e find 15 class es (Eastm an and T opan , 1966). T he dat a are arran ged r oughly on the basis of n atural kinds , such as +hum an , +t ree/ riv er , +body part/ action , +thing , +im plem ent , +abstr act entity , +location , etc., in such a w ay that , w here possible, they show the opposition individual obj ect (I) v ersus family of individual obj ect (F ). T his corresponds roughly t o singular ver sus plural opposit ions in gr amm ars . N ot e that the dat a ar e not paired classes of the sam e obj ect x or y or z . Rather , they are paired as x v ersus y nat ural kinds . Alt ernat iv ely , they are unpaired natural kind p , q, or n . T hus , m s ichana (girl) in (1b ) is an in dividual natural kin d of hum an object correspondin g t o a gramm atical singular noun . It contrast s w ith walevi (drunkards ) in (2b ), w hich is a fam ily of individual nat ural kinds of hum an object correspon ding t o a gram m atical plural noun . T r adit ionally , m ost of the dat a are re- organized int o set s of pairs of the sam e natural kind of obj ect x , e.g . x 1/ x 2, y3/ y4, z5/ z6, q7/ q8, g9/ g 10, n 11/ n 10, w hich are then view ed as sin gular v ersus plural pairs of Pn - Ss . In the traditional syst em , singular regularly correlat es w it h (I) of obj ect x and plural r egularly correlat es w ith (F ) of obj ect x in the real w orld. F or ex ample, datum (1) m ay be paired w ith (17) and datum (3) w ith (18) below .
(17) a. Clas s 2 WA
function s . T he trig g ering effect w hich g ener ates S CSU s is know n as the clas s pr ojection pr inciple (CPP ) (A midu, 1997). T he CPP als o tells us h ow clas ses ar e org anized and then m apped unt o syntactic structur es t o deriv e Pn - S s . T he CPP has been discu s sed in Am idu (1997). T he data ab ove als o c on sist of a and b, ± c,
±d com ponent s . T he a com ponent st ates the clas s m arker/ s and clas s num ber/ s 1, 2, 3, etc . under descr iption, and the b, ±c, ±d com ponents g ive the output pr edication - sentence or - sentences der ived under the CPP or CPP s of the clas s m ar ker/ s . T he glos s follow s the meth od u sed in A midu (1997, 2001a, 2001b ). W e see in the abov e illu str ation s th at clas s description s are m or e c omplex than the mer e itemization of w ord s tr ucture m or pholog y and the com pilation of a tax on om y of w ord clu ster s c alled n oun clas ses , s om ething r ather fas hionable in Bantu g r am m atical studies since it s foundation (Creider , 1975, Denny and Cr eider , 1986, Demuth, 2000).
b. wa- sichana wa- le wa- ref u wa- m eolewa wa- wa- wa- wa- Cl. 2 they - girl Cl. 2 they - that Cl. 2 they - t all Cl. 2 SM they - RECENT PAST - m arry - PAS S .- MOD.
' T he t all girls are m arried.' (18) a . Clas s 4 MI
b . m i- to y - etu m i- k ubwa i- m ef urika mi- y - mi- i- Cl. 4 they - riv er Cl. 4 t hey - our Cl. 4 they - big Cl. 4 SM they - RECENT PAST - overflow - ST AT IVE - MOD . ' Our big riv ers hav e overflow ed (their banks ).'
T he result is the paired strings of Pn - Ss foun d in traditional gramm ar books . But not e that this inform ation about natur al pairs of the sam e obj ect x or y , et c., e.g . girl/ girls , or riv er/ riv ers , is not w hat class classificat ion is about . T he inform ation is , t herefore, not required in a t able of classes , at least in my view . T his is because classes in them selv es are not obligat orily paired syst em s of the sam e obj ect in Bantu , in m y opinion , hence girl/ drunkards is just as good an indication of so- called number as girl/ girls .
T he justification for our claim is as follow s . In both the r eal w orld and in language, w e find individual obj ect s w ithout corresponding 'collect iv e repr esent ations ' or family of obj ect . F or ex am ple, m ag om bano 'quarrel/ s ' in (6) has no class 5 *g om bano 'quarrel ' in Kisw ahili. We also find 'collectiv e represent ations ' or family of obj ect w ithout it s corresponding individual obj ect . F or ex ample, k uiba ' thievin g ' in (14) contr ast s w it h not hing . T his fact is captured by our dat a (1)- (16) than by tr adit ional descript ions and classifications . T his seem s t o suggest that each class is independent m orphologically and synt actically of every other class , and not all classes can be paired, as this study w ill further illustrat e below (Am idu , 1997, for further discussions ).
S o far , w e hav e s een t hat som e Kisw ahili classes m ay be paired int o
singular v ersus plural, correspon ding t o pair s of natural kind obj ect s as
(I) v ersus (F ). W e hav e also seen t hat this pat t ern does not imply that
all clas ses ar e reducible t o a binary num ber sy st em . It follow s that
class es represent natur al kin ds of obj ect , w hether or not the nat ural
kinds corr espond w ith gramm atical number singular v ersus plural. In addition , a s o- called singular noun m ay correspond w it h a family of obj ect s in it s function and a so- called plural noun m ay correspond w ith an individual obj ect in it s function (com pare (15) w ith (6) on on e reading ). Last ly , the s am e so- called singular or plur al class m ay correspond w ith in dividual and family obj ect functions in the gramm ar thus defeatin g the classification of the class int o pair s of sin gular ver sus plur al num ber . See (6), (14), (15), (16) abov e. All these variations suggest that class is different from num ber in Kisw ahili. My init ial hypothesis , therefor e, is that the pairing of classes in Bantu by linguist s and Bantuist s lies out side t he fram ew ork of class classification and organization and belongs t o number clas sification and organization . T he t w o organizat ions , i.e. natural kind an d number , belong t o different syst em s of the gram m ar . W e blur the levels in clas s sy st em s w hen w e fuse them int o the sam e affix es . T his problem is furt her discussed in § 3.2 below .
Con sider also the follow ing aspect s of the dat a. F or ex ample, the
predication - sent ence type (16b )- (16d) abov e in the class 17/ 26 NI3 is
kn ow n in traditional class clas sifications as locat iv e classes 16- 18
follow ing general Bantu practice. T he general pract ice does not reflect
Kisw ahili Bant u string construction s and patt erns . Consequently ,
follow ing discussions and recomm endations in Amidu (1994, 1997), the
traditional syst em of numbering is changed t o reflect the evidence in
Kisw ahili. In view of this , inst ead of classes 16- 18, I pr efer 17/ 26 NI3,
because KU > NI, PA > NI, and MU > NI in Kisw ahili n oun t o noun
deriv at ions . A similar m odification applies t o clas s 16 t o giv e 16/ 25
MA2 or PA in our classification . T he class 16/ 25 MA2 or PA is
justified by the fact that place nouns in the clas s can be m odified by
adj ectives , w hereas locative nouns w ith affix {ni } cannot be so m odified
in Kisw ahili. In addition , the prefix {m a } or {pa } of the clas s 16/ 25
implies , in cont ex t s , number as either singular or plural, or both ,
som ething the locat iv e affix {ni} does not im ply in Kisw ahili. T he class
16 or 16/ 25 is a class form ed by only one borr ow ed foreign w or d,
m ahali (place/ s ) and a couple of gr amm aticalized Bantu w ords , e.g .
p eup e ' open space, clearin g , square in a t ow n ' (Johnson , 1939, p. 87).
M ahali is oft en bantuized as p ahali. Amidu (1980, 1997), follow ing Asht on (1947), an d other scholars , claim s ther e is just one noun w ord in t he class 16 or 16/ 25. T his conclusion is no longer defen sible giv en w or ds such as p eup e ' open space ' abov e. In short , there is m ore t han one noun in clas s 16 PA or 16/ 25 MA2 or PA of Kisw ahili Bant u . F or simplicit y , the student should select either {pa } or {m a } as t he class m arker (Am idu , 1997 for det ails of t he 16/ 25 and 17/ 26 approaches ).
3 .2 . A lim it ation of c ount ability an d num b er pairs
What about the claim that s om e it em s ar e neither (I) nor (F ) obj ect s?
It seem s univ ers ally accept ed that classifier/ class v alues hav e som e
primitiv e m eanings . It is , therefore, believ ed that the m eanings
underlyin g the clas sifiers of a class language reflect t he w ay the
speakers of the comm unity inv olv ed perceive and or ganize the w orld
around them , including possible w orlds , and w orlds of
aut o- comm unicat ion or thinkin g (Amidu , 1980, Davids on , 1975). My
illustrations , how ev er , predict that the Bant u m an and w om an did not
achiev e complet e correlation betw een their family organization an d t heir
speech organization . T his is due t o the fact that t he social or ganization
of society depended and depends , crucially , on the prior recognition of
natural kinds of obj ect s in the univ erse of refer ence. Without such a
recognition , there w ould be not hing t o organize int o (I) v ersus (F ),
nothing t o count as one v ersus m ore t han one, an d n o r eason for class
or language sy st em s w hich recognize gramm atical num ber as sin gular
ver sus plur al, the language correlat es of (I) ver sus (F ). An int erestin g
aspect of Bant u social or ganizations is , therefore, the discov ery that it
seem s relativ ely easy t o pair s et s of ont ological kinds natur ally as (I)
ver sus (F ), e.g . riv er/ river s , or thorn/ thorns . But it is , in practice, not
easy t o pair all ont ological kin ds as (I) v ersus (F ). F or ex ample, m ass
obj ect s like mud, spit tle, crow d, etc., are not easily paired as (I) of x
ver sus F of x (Amidu , 1997). Let us call the failure of pairing of
natural kinds as (I) v ersus (F ) in the organization al syst em the pr oblem
of n eutral or cen tral m onads . Nam ely , there ar e obj ect s that are, strictly speaking , n either (I) n or (F ) obj ect s . T hese are oft en our uncount able obj ect s of gramm ars . Do these obj ect s hav e gramm atical number , t oo?
If w e turn t o Kisw ahili Bantu , w e discover that the problem s of social or ganization just described abov e carries ov er int o class organization . F or ex ample, all the pr edication - sent ences (1)- (18) abov e of the Kisw ahili Bantu gramm ar are not sin gular v ersus plural pairs of const ructions reflecting pair ed class es of natural object s like x 1/ x 2, e.g . dat a (1)/ (17), or ran dom obj ect s like x/ y , e.g . dat a (1)/ (2). In deed, dat a (13)- (16), for ex ample, ar e not pair ed int o class es of sin gular ver sus plural inflect ions . F or these dat a, t he issue of number affix is m eanin gless . Why is this so? Dat a such as (13)- (16) rev eal that the Bantu people recognized t he central cat egory of obj ect s m ention ed abov e. T hey also recognized that , w it hin the dichot omy singular v ersus plural, t here are class w ords w it hout number affix es . Such affix es represent a non - number cat egory . T o resolve this anom aly of non - number w ithin a number syst em , the absence of number as either singular or plural has been called cent ral num ber (Amidu , 1997). It is abbr eviat ed as (Ce.).
F ollow ing from our analysis , w e discov er t hat the Bantu classes ,
especially w hat Mut aka and T am anji (2000) call ' Narrow Bant u ' , consist
of i) singular clas ses , ii) plur al classes , and iii) cent ral class es . In t his
respect , Ce. m eans a class w ith a non - count able and non - number or
number neut ral class affix . But how can this be, if classes are paired
singular v ersus plural affix es in Bantu? And yet , t his is ex actly w hat
w e fin d in dat a (1)- (18). T hat is , som e classes easily pair for natural or
random number , e.g . (1)/ (17), (3)/ (18), (1)/ (2), w hile others nev er do,
e.g . (13)- (16). T his m eans t hat there is a negation of gramm atical
number in the class sy st em s of Bantu , especially w hen num ber is m ade
cot erm inous w ith natural kin ds . In this respect , the evidence contradict s ,
in a fun dam ent al w ay , gram m atical ass ertions t o t he effect that the
Bantu clas ses ar e exclusiv ely a paired sy st em of singular v ersus plural
affix es . I return t o t he issue again in § 4. below .
4 . Cl a s s e s a s re f le c t i o n s o f n at u ra l k in d s a n d s o c i a l o rg an i z at i o n in B a n t u a n d K i s w a h ili
What is the nat ure of the n atural kinds in Bantu? T he dat a (1)- (18) reflect the Bant u conceptual view of the univ ers e, especially t he core ont ological kinds of obj ect that m ake up the w orld. I w ish t o illustrat e here, using tw o descriptions , how Bantuist s , as int erpret ers , hav e order ed and arranged the sem antic cat egorizations of nat ural kinds found am ong Bantu speakers . Here is the first description . Demuth , F araclas and Marchese (1986, p. 455) s elect Sesotho as "an ex am ple of a full noun class/ concordial agreem ent syst em ... — a typical Bantu language." T hey then present 15 clas ses of this language, and explain that "In such a syst em each noun is prefix ed w ith one of a pair of CV - noun class m arkers , one used for the singular form , the other for the plural." T hey go on t o st at e that :
T he m ore conserv at iv e Bantu lan guages typically hav e 5 or 6 pr oduct iv e singular/ plural n oun class or gender pairs , plus a few classes w ith no alt ern ation . While productive sem antic correspondences have been lost for m ost of t hese gen der distinctions , classes 1/ 2 (m o / ba abov e) and 2a (bo- the kinship class ' ) are generally know n as t he hum an clas ses . Class es 9/ 10, in the larger Niger - Congo cont ext , have been called the ' larger anim al ' classes . Most Bant u languages and m any other languages in Niger - Congo also hav e a 'm as s noun ' or ' liquid ' (14 bo) class w hich generally exhibit s no sin gular/ plural pairing (Demuth et al., 1986, pp. 455- 456).
In her recent w ork , Dem uth (2000, p. 272) list s 23 "Various
N ig er- K ordof an ian noun class sy s tem s " (Maho 1999, p. 247- 248, for
similar classification s ). T he sem antic and number classifications of the
Bantu classes giv en by the scholars abov e are s aid t o reflect the Bantu
univ erse of r eference and the Bantu m ode of ordering the univ erse of
natural kinds an d properties of these. T he scholars also ass ert that the
organization of the classes reflect s gram m atical number , i.e. sin gular ver sus plural.
T he second description com es from A sht on . It is specifically about Kisw ahili. A sht on (1936, 1937) dev eloped w hat she t erm ed the ' Idea Appr oach t o Sw ahili,' w hich she then applied t o the w rit ing of her gramm ar book of 1944/ 1947. T he underlying ideas behind Kisw ahili class es provided by A sht on are giv en on v arious pages of her gr amm ar book . T he reader m ay refer t o the dat a (1- 18) abov e t o confirm som e of her claim s . W e summ arize Asht on ' s claim s below . W e give t he page numbers aft er each extr act .
M - WA - classes : "..Living Clas ses ...cont ain the n am es of hum an beings ." (Asht on , 1947, p. 29). E .g . mp ishi ' cook ' .
M - MI- classes : "..nam es of livin g things but not hum an ,....", e.g . all trees , plant s etc. "som ethin g that spreads or ext ends "
(A sht on , 1947, p. 23). E .g . mj i ' t ow n ' , m ti 't ree ' , m lim a 'm ount ain ' , etc.
JI- MA - classes : "nam es of things w hich occur in quantities , but w hich m ay be thought of singly as w ell,...." (Asht on , 1947, p.
65). E .g . j ino 't ooth ' .
KI- VI- classes : "...oft en spoken of as the thing classes , for m any of the nouns are t he nam es of inanim at e t hings as opposed t o anim at e or sentient beings ." (A sht on , 1947, p. 14). E .g . k iti 'chair ' , k ik o 'pipe ' .
N - N - class es : "...w ords w hich are the n am es of comm on obj ect s and of anim als ." (Asht on , 1947, p. 82). E .g . f im bo 'w hip ' , s im ba 'lion ' .
U - classes : "T her e ar e tw o U - Clas ses ...".
a . U - (< BU - ) Class . T his cont ains "W ords w hich adm it of no singular or plural concept , such as abstract nouns den otin g qualities or st at es ." (A sht on , 1947, p. 104). E .g . uz uri 'beauty ' , utu ' m anhood ' .
b . U - (< LU - ) Class . "All w ords , how ev er , refer t o concr et e
obj ect s , w ith a furt her im plication of length or m ass ."
(A sht on , 1947, p. 105). E .g . uk uta 'w all ' , uf ag io ' broom , brush ' , uj i ' gruel ' .
KU - clas s : "T hat of a v erbal noun , as such it form s a clas s of nouns know n as the KU - class ...T hes e v erbal noun s cannot be thought of in t erm s of singular and plural. T hey expr ess the act of doing , of becom ing or the st at e of being ...."
(A sht on , 1947, p. 123). E .g . k us om a 'reading/ s , k uondoka ' departure/ s ' .
MAHALI class :"Reference t o a definit e place." (A sht on , 1947, p. 125).
E .g . m ahali 'place/ s ' .
ADVERBIAL classes (Asht on , 1947, p. 126)
KU - class : "ku - Indefinit e place, direction ". E .g . huk u ' there ' . PA - class : "pa- Definit e place, position ". E .g . hap a 'here '
MU - class : "mu - Ar ea, "alongness ", "w ithinness "". E .g . hum u 'in here/ t here '
Not e that A sht on ' s "adverbial place class es " are also called locativ e class es (Amidu , 1980, 1997, see als o §§ 3.- 3.1. supra). W e refer the reader also t o discus sion s by S acleux (1909), Haddon (1955), and Corbett (1991) on the t opic of t he m eanings of class affix es . Our dat a (1)- (18) alm ost ex actly m atch A sht on ' s classificat ion . T he ' ideas ' underlyin g A sht on ' s classes reflect the core Kisw ahili cat egorizations of the w orld and hence the Kisw ahili and Bantu view of n atural kinds of obj ect in the w orld, and, con sequent ly , the collective v ersus in dividual represent ation s of the natur al kinds of obj ect in t he w orld, includin g possible w orlds .
W e m entioned earlier on that in t heir social organization of the family
or collectiv e unit s of object s , the Bantu m an and w om an soon realized
that there w as a m iddle ont ological cat egory of central m onad or entity
in respect of count ability . We hav e att empt ed t o show that the
gramm at ical organizat ion of classes also reflect s this centr al cat egory . In
the class sy st em , the central cat egory m ay be said t o inv olv e number
neutralization in s om e affix es . W e hav e suggest ed that number
neutralization in affix es m ay part ly be resolved linguistically by
recognizing a central number . Essent ially , therefore, the cent ral number and it s affix es in class syst em s display w hat is kn ow n as syncretism in languages . Within centr al num ber , the distinction singular v ersus plural is neutralized. T his leav es the distinction singular v ersus plural t o be m ade s em antically according t o the cont extual function s of lexical w ords or phrases or Pn - Ss them selves . Or else, no number distinction s are m ade at all, and w e get class es w ithout number m orphem es an d m arkers . Let us look at A sht on ' s list for confirm ation of our analysis .
If w e look at t he above list of cor e conceptual cat egorizations present ed by Asht on (1947), w hat strikes us is that t he classes are paired initially as reflect ing an opposition singular v ersus plural affix es , nam ely M - WA , M - MI, JI- MA , KI- VI, N - N . T hen the opposition ceases . All the classes U - , KU - , MAHALI, PA - , KU - , MU , are not paired. Recall also that the locativ e classes of Kisw ahili act ually display the follow ing m orphemic patt erns of conv ersion : PA - > NI3, KU - > NI3, MU - > NI3, i.e. they are r epresent ed by ex actly t he sam e affix NI3 and the s am e noun ny um bani 'in the hous e ' in (16a )- (16d). Consequently , the Bant u noun m arkers PA - , KU - , MU - , are n eutralized and replaced by a comm on affix - NI3 in Kisw ahili Bantu and display no opposition singular v ersus plural, even sem ant ically . W e cannot call this class (or class es ) a gramm atical number class w hen it s affix es lack inflect ional and sem antic number m orphem es . T he pr efix es PA - , KU - , MU - only surface in m odifying w ords , e.g . k ule 'far ov er there ' , p ale 'just ov er there ' , m le ' in there ' , of the n ouns , as the dat a in (16b ), (16c), (16d) dem on strat e (Amidu , 1980, 1997). T he unpaired class es are evidence of number neutralization or non - number .
5 . A c rit i c i s m o f n a t u ra l k in d s a n d n u m b e r d e s c ript i o n s in B a n t u
What ar e the pairing anom alies in the t radition al descriptions of
Bantu and linguistic gramm arians? T raditional classification s giv e the
impression that number and natur al kinds are the s am e kinds of
m orphem es . T hat is , class affix es describe number as singular or plur al
and hav e gender m eanings . In practice, gram m arians are unable t o defend this appr oach , and appeal t o a principle of fusion (Dix on , 1986, p.
106, Amidu 1997, p. 130). F or ex ample, A sht on ' s (1936, 1937, 1947) ' idea approach ' st at es that "E ach class is associat ed w ith one or m ore underlyin g ideas ". In spit e of this , she clas sifies her classes alon g a number - gen der scale, t hus fusing num ber and gender t oget her . A sht on ' s classification seem s t o m ake the lex ical concept ual feat ures of n atural kinds , such as +hum an , +tree/ riv er , +body part/ action , +thing , +implem ent , +abst ract entity , +locat ion , etc., subordin at e t o the number features , singular/ plural, of her clas ses . In our view , it s eem s bett er t o use a tw o- lev el scale: 1) a lex ical conceptual scale of natural kinds and b) a number property or feature scale. T he m otiv at ion for this is that number is like an adj ectival predicat e in Bantu , w hile underlying ideas are m ostly tied in w ith noun s (Am idu , 1997). In principle, therefore, number and underlying ideas (genders ) are not concept s of the sam e cat egory in the clas s syst em s , ev en though both can fus e t ogether (Amidu , 1997). A tw o- lev el syst em prev ent s number , a 'long series ' feature, from underlying all t he classes . T his in turn prev ent s affix es of w or ds that refer t o underlyin g ideas or natural kin ds , and w hich are really ' short series ' elem ent s oft en restrict ed t o one or tw o classes only , from n ecessarily expressing number as either singular or plural. If my assertion is m otiv at ed, then w e should also be able t o find in Bantu and Kisw ahili classes w ords w hich ex press number by m eans of noun m odifiers and not by m eans of class pr efix es . Such ex am ples w ould illustrat e furt her t hat number is not an inherent feature of class affix es .
We further justify the separat ion of number from n atural kinds below .
F or ex ample, observ e that A sht on ' s class 15 KU , ex em plified by a w ord
such as k us om a 't o read, reading ' , is pair ed w ith nothing in her
Kisw ahili dat a above (see below for her t able cont aining k uchez a
'play ' ). In addition , on page 123 of her w ork , Asht on (1947) ex plicitly
st at es that the verbal noun w ords in class 15 KU "cannot be thought of
in t erm s of singular and plural". F urtherm ore, on page 104 of her book ,
Asht on also claim s that the class U - (< BU ), also giv en as clas s 14 bo
by Demuth et al (1986), does not oft en show sin gular/ plural pairin g in
Kisw ahili. T his claim about class 14 is repeat ed by Demuth et al. (1986, p. 456) as affecting "m ost Bantu languages and m any ot her languages in Niger - Congo". Let us assum e that Demuth et al. (1986) are correct . If so, their initial general claim t o the effect that "In such a syst em each noun is prefix ed w ith one of a pair of CV - noun class m arkers , one used for the singular form , t he other for the plur al." is simply unt enable and s elf- evidently false (Dem uth et al.).
It is clear from the abov e analysis that there is a dr aw back in t he
w orks of Bant uist s . N am ely , t hey excuse som e clas ses from t he
principle of pairing classes int o gender class es defined also as singular
ver sus plur al pairs of affix es , ev en as they insist that clas s affix es ar e
number prefix es used t o distinguish singular fr om plural in Bantu . W e
see in the abov e decriptions , ther efore, quit e clearly the par adox of
number in Kisw ahili an d Bantu class descriptions . T he evidence, in my
view , show s that , in principle, the tw o classes U - an d KU - are, in
m ost Bantu languages an d in m any Niger - Congo lan guages , incapable
of being subj ect t o the principle of arrangin g class es int o paired
singular/ plural affix es . W e discov er from the abov e descript ions ,
therefor e, that gender or n atural kind must be defined for the tw o
class es U - (< BU ) and KU - as s om ething w hich is independent of the
syst em number . T he evidence also m ean s that not all clas ses st an d for
singular or plur al inflection . T his leads t o on e conclusion only . Nam ely ,
som e classes are central class es . Unfortunat ely , Bantu gr amm arians do
not handle the evidence about central classes w ell in w riting t heir
gramm ar s . F or ex ample, in or der t o resolve t he anom aly w hich has
surfaced abov e, som e linguist s , inter alia, H eine (1982), Corbet t (1991)
claim direct ly or indir ectly t hat the class KU - has no plur al. In short ,
they resolv e the problem of either neut ral number or non - number
simply by imposing number on the class KU - (Amidu , 1997 for
discussions ). T her e is no cross - linguist ic evidence for adopt ing such a
solut ion , at least , for Kisw ahili (Amidu , 1997, on Hein e, 1982). W e
illustrat e the lack of inflect ional number in the class 15 KU affix of
Kisw ahili w ith the follow ing datum fr om George Orw ell (1967).
(19) K uj a na k uondoka k wak e k uliwahof isha m no wany am a.
Cl. 15 it/ they - coming Cl. 0 ø- COP - be with Cl. 15 it/ they - leave Cl.
15 it/ t hey - his Cl. 15 SM it/ they - PA ST - Cl. 2 OM
they - fear - CAUS - MOD. Cl. 0 ø- v ery much Cl. 2 t hey - anim al 'His com ing/ s and going/ s fright ened the anim als v ery m uch .'
T he datum (19) has a coordinat e NP k uj a na k uondoka. T he agreem ent in the dem on strativ e k wak e 'his ' for both coor dinat e nouns is {ku } and the predicat e v erb k uliwahof isha 'it/ t hey fright ened them ' also has the s am e affix {ku }. T he construct ion is perfectly good Kisw ahili.
But w hat is int erestin g is that the conj oined elem ent s k uj a 'coming ' and k uondoka 'leaving ' are t ranslationally am biguous . K uj a m any m ean one inst ance of comin g or m any inst ances of com ing . Likew ise, k uondoka m ay m ean one inst ance of leaving or departure or s ever al inst ances of this . In Kisw ahili, w e can also hav e t he follow ing const ruction types .
(20) K uj a k wak e k uliwahof isha m no wany am a.
Cl. 15 it/ t hey - coming Cl. 15 it/ they - his Cl. 15 SM it/ they - PAST - Cl. 2
OM they - fear - CAUS - MOD. Cl. 0 ø- v ery much Cl. 2 they - anim al
'His com ing/ s fright en ed the anim als very much .' (21) K uondoka kwak e k uliwahof isha m no wanyam a.
Cl. 15 it/ t hey - leav e Cl. 15 it/ they - his Cl. 15 SM it/ they - PAST - Cl. 2
OM they - fear - CAUS - MOD. Cl. 0 ø- v ery much Cl. 2 they - anim al
'His going/ s (or departure/ s ) fright ened the anim als v ery much .'
T he subj ect of (20) is k uj a an d t he subj ect of (21) is k uondoka. T he
NPs k uj a an d k uondoka ar e the v ery elem ent s that form the coordin at e
NP subject of (19). In (20), the possessive and t he predicat e v erb of
k uj a are the v ery sam e kwak e and k uliwahof isha w e saw in (19). T he
concord affix es are {ku } for each m odifier . If w e t urn t o (21), w e discov er again that the possessiv e and the pr edicat e v erb of k uondoka are the v ery sam e kwak e and k uliwahof isha w e s aw in (19), and the concord affix is {ku } for each m odifier . In addit ion , n ot e how the fact that k uj a or k uondoka is used on it s ow n as a subj ect argum ent NP does not m ake it aut om atically a sin gular noun . In fact , the dat a (20)- (21) are, in this cas e, translat ionally just as am biguous as (19).
Our study of Kisw ahili patt erns show s t hat the patt erning in class 15 KU effect iv ely rev eals that there is not hing inherently singular or plural about it s affix es and it s class . Disambiguation depen ds on the m odifier s of the head noun , i.e. according t o how they restrict the m eaning of the head noun in a construct ion t o a singular int erpret at ion or plural int erpr et ation . T he distinction singular v ersus plural is not a propert y of the clas s 15 KU it self or it s affix es , at least in Kisw ahili. T he class 15 KU is a clear ex ample of non - num ber , i.e. t he absence of number defin ed as singular ver sus plural in Kisw ahili. T o sav e the number theory in Kisw ahili and Bantu w e need t o r ecognize a middl
e cat egory of number , hence central number .
Table 1: Ashton's arrangement of Kiswahili classes (1947)
singular pronouns Plural pronouns
m-tu a person wa-tu persons, people
m-ti a tree mi-ti trees
ki-ti a thing vi-ti things
j i-cho an eye ma-cho eyes
nj ia a path n-j ia paths
ulimi a tongue n-dimi tongues
ku-cheza to play, playing - -
Mahali a place Mahali places
*W ith tw o exception s , the prefix in the plur al Clas s differ s fr om th at of the s ingular Clas s . Each clas s is as s ociated w ith one or m or e under lying ideas .
A sht on ' s t able and claim abov e suggest that prefix es n eed not differ
at all t o form singular/ plur al classes . And so, she classifies class 15 KU
as sin gular in number , a m ethod repeat ed directly by H eine (1982). And
yet , on page 123 of her book , as seen above, she claim s emphatically that the class 15 KU "cannot be thought of in t erm s of singular and plural". T he t able reinforces the contradict ions not ed abov e. N ot e, how ev er , that A sht on (1947) does not include the class 15 KU am ong the tw o ex ceptions m ention ed in her t able. In cont rast t o A sht on (1947) and H eine (1982), w e hav e n ot ed that Corbet t (1991) an d ot hers ass ert that clas s 15 KU has no plural, an assumpt ion that indir ectly implies that it has a singular pr efix . T he evidence provided abov e show s that Corbett ' s (1991) description lacks gramm atical m otiv ation for Bantu . Furtherm ore, in Bantu lan guages , w here classes such as clas s 15 KU are said t o have plur al prefix es , one discov ers , unhappily , t hat the so- called plur al prefix es do not form inherent or natur al inflect ional pairs w ith the so- called singular prefix es .
6)T he t radition al appr oach t o number does not also account for the fact that m orphem ic m eanings of number as singular or plural m ay be cont ex t dependent on m odifiers of noun phrases , or on the usage of a w or d or a phrase or predication cont aining t he sam e affix . F or ex ample, in k u- p iga k w - ing i 't oo much beating ' or 't oo m any beatings ' , number is a property of t he root - ing i 'm any , much ' in the m odifying w ord kw ing i. Num ber is not in the class affix {ku } t hat defines the class (Amidu , 1997, pp. 361- 370).
A sht on (1947) also st at es that in Kimvit a or Mombasa Kisw ahili, one m ay hear m ahali as "p ahali p l. m wahali". T his w ould seem t o be the additional justification for her classification of M ahali class as inv olving number singular/ plural. Her justificat ion is how ev er w eak in St andard Kisw ahili (cf. dat a (22)- (23) inf ra). Ev en if w e t ake int o account Kimvit a or Mom basa Kisw ahili, w e discov er that , as far as the St andard Gr amm ar goes , A sht on ' s t able of classes above impos es number sem antically , but n ot m orphologically , on her M ahali class . T he
6) T he meth od of pair ing unn atur al clas ses as number clas ses confirm s the adv ant age of the pairing s in (1)- (16) ov er the tr aditional ones (M aho, 1999;
H addon , 1955). T hes e s o- called number pair s are r andom or pseudo- pair s becau se, in Kis w ahili, one can bor r ow affix es fr om other clas ses either as allom orph s of exis ting affixes or for use as inflection al m ar ker s in other clas s es . T his principle is c alled allonom inal con cord m ark ing (Am idu , 1997).