Alexandre Kimenyi
   


THE PERFECTIVE ASPECT AND CONSONANT MUTATION IN BANTU LANGUAGES

Alexandre Kimenyi
California State University at Sacramento

I. Introduction

Many Bantu dictionaries list both the imperfective and the perfective aspect of the verb stem. This practice is due to the "unpredictability" of the phonetic realization of the final consonant of the verb stem. According to this practice, there are no rules. Unfortunately, most Bantuists espouse the same view and because of this, eventhough this consonant mutation is universally known,as the "modified form" of the verb stem, there has been very little attempt to study what this phenomenon really is. Even the most important question such as what the underlying perfective morpheme is has never attracted any attention because it is universally believed to be -ile. This lack of interest is due to the prevalent theory of that time: most of the dictionaries and grammars were written during the structuralist period, when linguists were interested in paradigms and thus the fundamental question of generative grammar about the underlying representation was not asked. When looked at this modified form very closely, however, there seems to be a certain pattern. Data from different languages (Kikongo, Silozi, Herero, Yao, Luganda, Runyankore-Rukiga, Kikuyu, Kinyarwanda, Lingala, Zulu) show that certain rules which account for these various phonetic realizations can be formulated, although without elegance and precision for the time being. Some expections exist ofcourse and it is these ones that should be listed in the dictionary.

II. Cross-Bantu Description

2.1. Kikongo perfective

In Kikongo, the perfective aspect has different allomorphs whose phonetic realization depends on the number of syllables that the stem has and whether the stem ends with a nasal or a liquid.

2.1.1.Monosyllabic verbs:

The perfective aspect of monosyllabic stem is -idi as seen in the following examples.

dia/diidi 'eat'; vua/vuudi 'to leak'; fua/fuidi 'to ooze'; zaia/zaidi 'to know'; yia/yiidi 'to steal'; kaia/kaidi 'to share'

2.1.2 Bisyllabic verbs ending with a liquid:

Bisyllabic verb stems which end with a liquid take -idi also but the stem liquid becomes d too, obviously a case of regressive assimilation.

lula/ludidi 'to taste bitter'; vala/vadidi 'to scrape'; vula/vudidi 'to undress'; yila/yididi 'to boil'; dila/dididi 'to lament'; tula/tudidi 'to go to bed'; lala/ladidi 'to melt'; kala/kadidi 'to deny'.

There are few exceptions such as kala/kedi 'to live'; lela/lelele 'to raise'.

2.1.3. Polysyllabic stems ending with a liquid

Polysyllabic verb stems which end with a liquid have it changed to d and the final vowel is i.

balula/baludi 'to put/push/pull'; bambula/bambudi 'to shoot a marble'; bubula/bubudi 'to oppress/persecute'; bundula/bundudi 'to unfold'; dongula/dongudi 'to extract from the body'; kayila/kayidi 'to give'

There are two possible analyses for this phonetic realization. The first account is that which sees this as a change of the liquid to d. The second alternative is the deletion analysis: the last syllable deletes when the perfective aspect -(i)di is added. The only problem with the last analysis is that the vowel noticed with the monosyllabic stems is not accounted for.

2.1.4. Stems ending with nasals

Monosyllabic and bisyllabic stems which end with a nasal consonant have -ini or
-ine added:

tuma/tumine 'to send'; kona/konine 'to snore'; wuma/wumini 'to dry'; nua/nuini 'to drink'; bama/bamine to incriminate'; zima/zimini 'to exterminate'; china/chinini 'to be scared'

Polysyllabic stems ending with an alveolar nasal consonant behave like their liquid counterparts discussed above. Nothing seems to have been added. It is the same nasal which is found. Again the last vowel is either -e or -i. The same type of analyses proposed for polysyllabuc stems ending with the liquid l apply here. The phonetic output is due to either the non-addition of the perfective morpheme or the final syllable deletion in front of the perfective aspect marker.

kanina/kanini 'to ask for permission'; bundumuna/bundumuni 'to opem'; danuna/danuni 'to break'; bamina/bamine 'to incriminate each other';

Since in Kikongo, the liquid of the applicative morpheme becomes a nasal when the verb stem ends with a nasal, verbs with the applicative extension and unextended verb stems with a nasal ending are subjected to neutralization when they both have a perfective aspect as seen in the examples below:

bama/bamine 'to scold/to find fault'
bamina/bamine 'to find fault with'

Verb stems with a last nasal consonant which have extensions, containing oral consonants, have a nasal consonant in the perfective aspect also.

bangamuka/bangamukini 'to turn onself over'
chamuka/chamukini 'to jump'; vumuka/vumukini 'to sigh';

The only consonant mutation that takes place in Kikongo because of the perfective aspect is the liquid of the polysyllabic stems. The change of t to ch as seen in the following examples is due to a general rule of Kikongo which palatalizes t in front of the high front vowel:

buta/buchidi 'get dark'; fita/fichidi 'pay'

The following data show that the rule is general and affects all ts in words of all categories:

china 'to fear' <-tina; nchima 'heart'<mtima; nchi 'tree'<mti

Kikongo behaves like other southwestern Bantu languages such as Herero, Ciluba, Silozi. In most of the cases, the liquid found in the applicative suffix and the perfective aspect is realized as d. This is true also for the class 5 prefix which is realized as di- instead of ri-. The applicative suffix and the perfective aspect become nasals also if the verb stem ends with a nasal. In Herero for instance, the perfective allomorphs are -ire/-ere or -ine/-ene depending on whether the last consonant is oral or nasal whereas the vowel alternation is due to vowel harmony which is universal in Bantu languages.
Here are some Herero examples:

pata/patere 'shut'
veta/vetere 'throw'
rota/rotere 'dream'
pita/pitire 'go out'
tuma/tumine 'send'
sina/sinine 'smother'
kama/kamene 'squeeze/milk'
hena/henene 'run away'
tona/tonene 'beat'


2.2 Perfective in Silozi

Silozi, like Kikongo, doesn't have many cases of consonant mutation, either, when the perfective aspect is added. This morpheme is clearly -ile as examples that follow show.

2.2.1. Monosyllabic verb stems

nwa/nwile 'to drink'; fa/file 'to give'; shwa/shwile 'to die'; ca/cile 'to burn'; wa/wile to fall'

2.2.2. Stems ending with -ya

loya/loile 'to poison'; toya/toile 'to hate'; siya/siile 'to leave'; zamaya/zamaile 'to walk'

2.2.3. Stems ending with nasals

For verb stems ending with a nasal, it looks as if the perfective aspect is just the vowel -i which is added.

bona/boni 'see'; katana/katani 'meet'; latana/latani 'love one another'
kanama/kanami 'lie on the back'; patama/patami 'lie on stomach'

As the following exceptions indicate, however, it seems as if the whole suffix -ile is added but the liquid and the final vowel delete after:

bina/binile 'dance'; tama/tamile

2.2.4. Alveolar fricatives (s,z) and the palatal nasal ny

The perfective aspect of verb stems which end with the fricative s, z and the palatal nasal ny is realized as -ize:

basa/basize 'decorate'; besa/besize 'roast'; eza/ezize 'do'; buza/buzize 'ask'; anya/anyize 'suck'; kopanya/kopanyize 'mix'; likanya/likanyize 'compare'

Polysyllabic stems ending with -z-

When a polysyllabic verb stem ends with -z-, there seems to be a deletion of this z and the addition of -lize to the stem.

babaza/babalize 'praise'; buzeza/buzelize 'ask for'.

This phenomenon will come to light when we start looking at Kinyarwanda. This is a case of reanalysis. The rule which adds the perfective aspect to the verb stem treats these s and z as "mutated" consonants, thus as a combination of r and the high front vowel of the perfective aspest and removes the high front vowel to its right between the liquid and the final vowel, thus changing the liquid to z in polysyllabic verbs whereas in bisyllabic stems, the vowel of the aspect marker keeps its position but is also copied in the prefinal vowel position which explains mutation in both positions.

2.2.5. Stems ending with the liquid l

Verb stems which end with a liquid, have either -ile added or some changes take place:

nyala/nyezi/nyalile 'marry'; zwala/zwezi/zwalile 'give birth'

In the above examples, the modified form has both vowel coalescence (a+i>e) and consonant mutation. These two changes are common in all Bantu languages which have the perfective aspect.

2.3. Perfective aspect in Yao

In the majority of verb stems, the perfective marker is realized as -ile without consonant mutation except vowel modification.

Monosyllabic stems:

-cha/chele 'to dawn/to clear (rain); -gwa/gwile 'to fall'; -ja/jile 'to go'; -lya/lile 'to eat'; ng'wa/ng'wele 'to drink'; -nya/nyele 'to defecate'; -pa/pele 'to give'; -pya/pile 'to be burned'; -swa/swele 'to set' (sun); -ta/tele 'to name'; -twa/twele 'to pound'; -wa/wele 'to become'.

The majority of bisyllabic verb stems take -ile without modifying the verb:

-puta/putile 'strike'; -luma/lumile 'sting'; -jimba/jimbile 'sing';-lemba/lembile 'write'; kusuma/sumile 'buy'; -mata/matile 'smear'; -sisa/sisile 'to hide'; -sita/sitile 'to iron'; -tema/temile 'to cut down'

Velars (k, g) and alveolar liquid take -ile but the last consonant changes. The liquid and g change to s, k to ch and the prenasalized velar ng to nj:

-lwala/lwasile 'to be sick'; -mela/mesile 'to grow' (plants); -nola/nosile 'to sharpen'; -pola/posile 'to cool/to heal'; -sala/sasile 'to say'; -wola/wosile 'to rot'; -pela/pesile 'suppose'; -mala/masile 'to finish'; -myola/myosile 'to shave';

laga/lasile 'suffer'; koga/josile 'bathe'; -loga/losile 'to bewitch'; -leka/lesile 'stop'; -tega/tesile 'to set a trap'.

-chinga/chinjile 'to herd cattle'; -kalanga/kalanjile 'to fry'; -langa/lanjile 'to take a leave'; -linga/linjile 'to try/measure'; -twanga/twanjile 'to strike with a fist'

-saka/sachile 'to want'; -seka/sechile 'to smile'; -janika/janiche 'to spread';
-koleka/koleche 'to hang up'; -peleka/peleche 'convey/conduct'; -wandika/wandiche 'be near/easy'; jika/jiche 'come',jimuka/jimwiche 'get up';


Polysyllabic stems with a round vowel on the second syllable (o, u) undergo gliding, o becoming we and u becoming wi:

-chochola/chochwele 'to clear the bush'; -dodoma/dodweme 'to hesitate'; -goloka/golweche 'to be straight'; -jogopa/jogwepe 'to frighten'; -koloma/kolweme 'to snore'; -pomola/pomwele 'to husk' (maize); -songona/songwene 'to wisper'; -tokota/tokwete 'to boil'; -kongola/kongwele 'to lend/to borrow'

-chuluka/chulwiche 'to be many'; guluka/gulwiche 'to fly'; -jumula/jumwile 'to be dry'; -kumbusya/kumbwisye 'to remind'; -sangusa/sangwise 'to shake';-tuluka/tulwiche 'to descend'; -nandupa/nandwipe 'to be few/small'.

It is important to note that verb stems which end with the voiceless velar k not only do they undergo vowel coalescence but consonant mutation as well. The effects of the high front vowel are seen in the stem vowel already. The main question is to explain how the same vowel can also at the same time force progressive assimilation making the next consonant palatal.

The low vowel a undergoes vowel coalescence

-jimbala/jimbele 'to be fat'; -gawanya/gawenye 'to divide'; -sakala/sakele 'to be bad'; -simana/simene 'to meet'; -jinama/jineme 'stoop'; -tukana/tukene 'to swear/to use abusive language'; -wangala/wangele 'to ressemble'; -watama/wateme 'to be flat'.

There is no change with polysyllabic stems with front vowels:

-chalila/chalile 'to persist'; -jasima/jasime 'to lend/to borrow'; -jinjila/jinjile 'to enter'; -lepela/lepele 'to fall'; -nonyela/nonyele 'to love/to like';-n'gambila/n'gambile 'to swim'

Deletion of the perfective aspect

Some verbs, like in Silozi, behave as if the liquid and the final vowel are deleted. It is again mostly verb stems ending with a nasal:

tama/temi 'sit'; jima/jimi 'refuse'; wona/weni 'to see'.

There are not enough data to know what is really going on. These examples that we have have undergone vowel coalescence. It seems therefore as if it is the polysyllabic aspect marker assignment that has taken place instead of deletion.

2.4. Perfective aspect in Kinyarwanda

In Kinyarwanda the perfective aspect is realized as -ye. Consonant mutation caused by the perfective aspect takes the following factors into consideration: number of syllables, place of articulation, vowel quantity and whether the last consonant belongs to the stem or to a suffix:

Monosyllabic stems have the final vowel lengthened, bilabials are palatalized, palatal consonants remain unchanged, velar and alveolar consonants are fricativized, nasal suffixes become palatal, whereas liquids of suffixes or regular stems with long vowels are deleted.

p, b, m >py, by, my.
d, r >z; t >s; n>nny; s, z>sh, j
k>ts, g>z

Verb stems which end with a liquid preceded by a a long front vowel (i or e) have the default applicative morpheme -ir- inserted. The long vowel in front of the liquid can be realized as long or short:

-riindiira<>-riindiiriye/-riindiriye 'wait for'
-siinziira<>-siinsiiriye/-siinziriye 'sleep'
-teembeera<>-teembeereye/-teembereye 'walk'
-voomeera<>-voomeereye/-voomereye 'water'

Monosyllabic verb stems which end with a mid back vowel have two alternative forms. Either the last vowel lengthens or they behave like Silozi verbs in that there are both gliding and the change of the vowel to e.

-ko- kwá/kóoye/kwéeye 'pay dowry'
-nyó - nywá/nyóoye/nywéeye 'drink'

A full treatment of consonant mutation in Kinyarwanda caused by the addition of the perfective aspect morpheme can be found in Kimenyi (1979 and forthcoming).


2.5. Luganda

Luganda in many ways behaves like Kinyarwanda. The perfective aspect is realized as either -ye or -dde. The allomorph -dde occurs with monosyllabic stems and stems which end with the liquid in polysyllabic stems or after a long vowel.

2.5.1.Monosyllabic stems

-ba/-badde 'be'; -nywa/-nywedde 'drink'; -lya/-lidde 'eat'; -wa/-wadde 'give'; -fa/-fudde 'die'; ggya/ggidde; gwa/gudde ' ; lwa/ludde 'spend time/stay'; -ta/-tadde 'release'; va/vudde 'come/go from'

2.5.2.Polysyllabic stems:


2.5.2.1.Stems ending with liquids

-wumula>-wumudde 'rest'; -ambala/-ambadde 'dress'; -sigala/-sigadde 'remain'; -zaalila/-zaalidde 'to give birth at'; -wulira/-wulidde 'hear'; -tunula/-tunudde ' look/be awake'; -tukula/-tukudde 'be white/clean'.

2.5.2.2.Long vowels:

-tuula/-tudde 'sit down'; -seela/-sedde 'overcharge'; -kuula/-kudde 'pluck/pull/uproot'; lwala>lwadde 'be sick'; beera/-bedde '

2.5.2.3.Double consonants:

-ggula/-gudde 'open'; -ggala/-ggadde 'close'

Like in Kinyarwanda, Luganda k and g as well as the alveolar liquid after short vowels mutate to z in front of the perfective aspect:
l, d, g>z/-ye

-kola/-koze 'work'; -kala/-kaze 'get dry'; -wera/-weze 'reach a number'; -gula/-guze 'buy'; -sula/-suze 'spend the night'
-linda/-linze 'wait'; -yiga/-yize 'study'
-fuga/fuze 'rule/govern'; -genda/-genze 'go'; -sanga/-sanze 'find'; -laga/-laze 'show'

Like in Kinyarwanda, k and t become s. The k in Kinyarwanda has become an affricate (ts).

-yita/-yise 'call/pass';
-teeka/-teese 'put'; -sirika/-sirise 'be silent'; -golokoka/-golokose 'wake up'; -tuuka/-tuuse 'arrive'.

Palatal consonants, in this language, don't change either. Palatalized and causativized stems have a special mutation and bilabial consonants become palatalized.

p/b/m>py, by, my

-yamba/-yambye 'help'; -yimba/-yimbye 'sing'; -tuma/-tumye 'send'; -naaba/-naabye 'bathe'; -siiba/-siibye 'pass the day'; -laba/-labye 'look'/visit; -funa/-funye 'get'

-ggya/-ggye 'take away; -zannya/-zannye 'play'

2.5.3.Causativized and palatalized stems

The perfective aspect of verb stems with causative extensions or which end with palatal consonants is realized as -ezza/-izza. Whether the vowel will be i or e depends on vowel harmony rules.

-liisa/-liisizza 'feed'; -koza/-kozesezza ' cause to work'; -siriza/-sirisizza 'cause to be silent'; -yingiza/-yingizza 'cause to enter'; -tegeeza/-tegesezza 'cause to understand'; -buza/-buuzizza 'ask/greet'; -tuusa/-tusizza 'cause to arrive'; -somesa/-somesezza 'teach'; -nyumya/-nyumyizza 'talk'.

2.6. Zulu

The perfective aspect in Zulu is -ile. It doesn't produce consonant mutation like in other languages.

-dla/-dlile 'eat'; -fa/- file; -pha/-phile 'give'; -fika/-fikile 'arrive'; -thuma/-thumile 'send'; -thuka/-thukile 'insult'

If the verb is polysyllabic, however, and ends with a liquid or an alveolar nasal, the perfective morpheme is not added but vowel coalescence takes place in the penultimate syllable.

-fingqana/-fingqene 'draw one another in'; -linganisana/-linganisene 'be in a straight line with one another/be parallel'; -sangana/-sangene 'be confused'
-fohlokala/-fohlokele 'get smashed'; -fufumala/-fufumele 'act courageously';
-gagadula/-gagadule 'do ineffectually'; -limala/-limele 'get hurt'; -salela/-salele 'remain for'; -sangulula/-sangulule 'bring round'.

These liquids are not part of the verb stem but suffixes indicating stativity, applicative, reversive, etc.

III. Analysis

3. 1. The underlying form

There are three competing possible underlying representations for the perfective aspect morpheme, namely (a) a discontinous morpheme in which -ir- is detached from the final vowel -e, (b) this morpheme is -ir- and -e is a default final vowel and (c) this morpheme is the same as the applicative morpheme.

3.1.1. A discontinuous morpheme

Treating the perfective aspect morpheme as a diccontinuous suffix is motivated by the fact that consonant mutation and vowel coalescence, as many examples from different languages indicate, are far away from the final vowel. The morpheme seems to be split and segments which belong to the verb stem are inserted inside it.
Another evidence supporting the discontinuous morpheme analysis is the behavior of the passive morpheme as illustrated by examples from Kinyarwanda.
It has been noted that the perfective aspect is -ye and that it causes consonant mutation. The Kinyarwanda passive morpheme is -w-. This morpheme instead of being inserted between the verb stem and the aspect, it is inserted between the y and e of the perfective aspect -ye.

ba-sek-a 'they laugh'
ba-sek-w-a 'they are laughed at'
ba-sek-ye [basetse] 'they just laughed'
ba-sek-y-w-e [basetskwe] 'they are just laughed at'
ba-reg-a 'they accuse'
ba-reg-w-a 'they are accused'
ba-reg-ye [bareze]'they just accused'
ba-reg-y-w-e [barezgwe] 'they are just accused'

The aspect morpheme seems to be clearly split, otherwise consonant mutation would not be able to take place.
Examples from Herero also show that indeed the passive morpheme -u- occurs inside the perfective aspect marker :

tuma/tumine 'send' > tumua/tuminue 'be sent'
pita/pitire 'go out' > pitua/pitirue 'be sent out'
tona/tonene 'beat' > tonua/tonenue 'be beaten'
rota/rotere 'dream' > rotua/roterue 'be dreamed'

3. 1. 2. -e as default final vowel

This vowel may be a default one. It is assigned because Bantu languages have to have a vowel in the verb final position. The evidence for this analysis comes from the fact that in some languages such as Kikongo, Silozi or Yao, the last vowel is not always -e. It can also be -i. In Luganda, the last vowel of verb stems which end with alveolar fricatives (s, z) and the palatal nasal ny as was noted earlier is -a instead of -e. The verb -many- 'know' is realized as -manyi in the perfect aspect instead of -manye. Lingala also has two types of vowels in the final vowel position: a or i. The vowel -a corresponds to the imperfective and -i to the perfective:

koloba 'speak'; kolobaka (infinitive habitual); nakoloba (future); nakolobaka (pres.hab.); náloba (subjunctive); nálobaka (subj. hab.); nalobí (ind.pst); nalobáki (definite past).

kosala 'work'; kosalaka (habitualy work); nakosalaka (future); nasala (subjunctive);
nasalaka (subjunctive habitual); nasali (indefinite past); nasakaki (definite past).

3. 1. 3. Relationship with the applicative morpheme -ir-

If -e is a default vowel, the perfective morpheme not only does it look exactly like the applicative morpheme -ir- but do the two morphemes also bahave phonologically the same as far as liquid deletion, vowel coalescence and vowel harmony are concerned. It is thus probably that it is the applicative morpheme which developed this meaning as well. The applicative morpheme has many meanings and functions (Kimenyi 1980, 1995). Sometimes it also acts as a default morpheme without any grammatical or semantic function, thus for purely phonetic reasons (Kimenyi, 1995).This default applicative is always inserted in Kinyarwanda and other Bantu languages such as Luganda whenever the verb stem has a causative morpheme -iish- or -y-. The causative morpheme -y- is sometimes reduplicated.

-soma/somye 'read'>someesha/someesheje 'caused to read'
/som-iish-ye/[som-iish-ir-y-ye]
-aandika/aanditse 'write'>aandikiisha/aandikishije 'caused to write'
/aandik-iish-ye/[aandik-iish-ir-y-ye]
-rira/rize 'cry'>riza/rijije 'caused to cry'
/rir-y-ye/[rir-y-ir-y-ye]
-vuga/vuze 'talk'>vuza/vugije 'caused to talk'
/vug-y-ye/[vug-ir-y-ye]

In the examples above, the liquid and the palatal glide (causative) mutate into z and z and the next y mutate also producing the palatal voiced fricative j.

...r+y > z; z+y > j

What is interesting about the applicative default insertion and the perfective morpheme is that they are sensitive to vowel quantity. Verb stems whose last consonant is z and which end with a long vowel in the penultimate position have both the applicative default rule insertion and the non-insertion form.

-gaanza/gaanje/gaanjije 'defeat'
-geenza/geenje/geeenjeje 'do'
-húuza/húuje/húujije 'cause to meet'
-cúruuza/cúruuje/cúruujije 'trade'
-taangaaza/taangaaje/taangaajije 'announce/advertise'

Verb stems with a short vowel on the other hand occur in only one form. Only bisyllabic verb stems undergo applicative default insertion and polysyllabic don't: There is no other option for each verb type.

-búza/*búje/bújije 'prevent'
-hóza/hóje/hójeje 'console'
-sókoza/shókoje/*shokojeje 'comb'
-súbiza/shúbije/*shúbijije 'respond'

A full account of the phonetic realization of causative and the default applicative morpheme insertion is given in Kimenyi (1999). The reason why the applicative is a default suffix it is because it is the most unmarked morpheme in this position as indicated by its polyfunctionality and frequency. In the prefix position, the morpheme which functions as a default is the infinitive marker -ku- which also stands for other tenses such as habitual past, negative, etc.
In Kiswahili for instance, it functions as a default morpheme with monosyllabic verbs in some tenses:

ninasoma /ni-na-som-a/ 'I read'
ninakula /ni-na-ku-la/ 'I eat'

In Kinyarwanda, the prefix ku- is inserted between long vowel tense markers and verb stems which start with vowels:

twaasoma /tu-aa- som-a/ [twaasoma] 'we would read'
twaakwaandika /tu-aa-andik-a/ [twaakwaandika] 'we would write'

The problem with treating tbe applicative as a polysemous morpheme whose many functions include that of marking the perfective aspect is again a phonetic one.
The regular applicative morpheme doesn't cause consonant mutation. Unless, this polysemy resulted in doublet creation. Doublets or twin linguistic morphemes which are characterized by near phonetic and semantic similarity usually develop from the same historical lexical item.
There exist in Kinyarwanda certain palatalization rules which are historically or morphologically motivated. Some words which belong to class 5 whose prefix marker is -ri- have the initial consonant palatalized as the following examples show:

ishaká <>amasaká 'sorghum'
ishyaánga <>amahaánga 'foreign countries'
ishyano <>amahano 'calamity'
ishyarí <> amaharí 'jealousy'
ishyérezo <>amahérezo 'end/conclusions'
ijaambo <>amagaambo 'words'
ijabo<>umugabo (man) 'courage'
ijana <>amagana 'hundreds'
ijíisho <>amáaso 'eyes'
iryíinyo <>améenyo 'teeth'
icúmi <>makúmi 'ten'

Words from numbers two to six have their initial consonants palatalized also if the regular class marker zi- is substituted with the vowel e-

ebyiri <>zibiri 'two'
eshatu <> zitatu 'three'
enyé <>ziné 'four'
eshaanu <> zitaanu 'five'
esheeshátu <> zitaandátu 'six'

This palatalization has created doublets in the language as seen in the following examples:

kunúunuuza 'to suck with a straw' <> kunyúunyuuza 'to suck blood'> to exploit'
iháanga 'foreign country' <> ishyáanga 'exile/diaspora'
kugabura 'to feed' <> kujabura 'to feed uncooked food'

In any case, even if the perfective aspect and the applicative are historically the same, synchronically, they are obviously separate morphemes. There is thus no motivation to consider the applicative and the perfective as one single morpheme.
The aspect morpheme seems clearly to be -ile not only because it is found in many languages and its different phonetic realizations occur in complementary distributions but also because its cognates in other languages have phonetic and historical accounts. The realization of -ile as -ye in Kinyarwanda and Luganda can be explained as a deletion of a liquid between two vowels. This phenomenon of liquid deletion is common in Bantu languages. In many languages such as Swahili, the applicative morpheme -ir- is realized as only -i- or -e- depending on the height of the preceding vowel. In many languages, class 11 marker -ru- has lost the liquid also, thus becoming homophonous with class 14 morpheme -bu- which also has lost the consonant in many languages such as Swahili again. The liquid -r- is also deleted in Kinyarwanda, when it is part of a suffix or is preceded by a long vowel and followed by the prefective aspect marker. The deletion of -r- of the perfective morpheme in both languages yielding -ye is thus independently motivated. The deletion of the liquid is found in other languages outside the Bantu family. It is found in the French morphology, for instance, when it is followed by the plural marker morphme -s as in animal and animaux (animal+s) 'animals', for instance. The alveolar stop d found in Kikongo and Luganda alternate with the liquid in many Bantu languages. Thus the applicative which is realized in many Bantu languages as -ir- is seen as -id- in Kikongo and Chiluba. The class 5 morpheme which is -ri- in other languages is -di- in Kikongo and Chiluba as well.

The relation of the perfective aspect marker with the agentive and the causative morpheme.

In Kinyarwanda, the perfective aspect behaves like the causative morpheme -y- and the agentive morpheme -yi. In other Bantu languages, however, the agentive morpheme is -i and doesn't cause consonant mutation.

-rima/-rimye 'toil' >umurimyi 'toiler'
-cúra/-cúze 'forge'>umucúzi 'blacksmith'
-rera/-reze 'raise'>umurezi 'educator'
-reeba/-reebye 'watch'>umureebyi 'watcher'
-riinda/-riinze 'guard'>umuriinzi 'guard'

-vuura/-vuuza 'cure/cause to cure'
-seka/-setsa 'laugh/make laugh'
-rwáara/-rwáaza 'be sick/to nurse a sick person'
-vúga/-vúza 'talk/to cause to talk'

The only difference noted between the perfective and the agentive and causative is the behavior of the liquid. It does not delete but becomes z instead:

-vuura/vuuye 'cure'>vuuza/*vuuya 'cause to cure' > umuvuuzi/*umuvuuyi 'healer'
-kiinguura/kiinguuye 'open' >kiinguuza/*kiinguuya 'cause to open' >umukiinguuzi/*umukiinguuyi 'the one who opens'

The only exceptions are: umubyéeyi 'parent'<byáara/byáaye 'give birth' and umurwáayi 'sick person' <rwáara/rwáaye 'be sick'.


3.2.The domain of the perfective aspect

The domain of the perfective aspect morpheme seems to vary from language to language. In Kinyarwanda and Luganda, this domain starts from the last sound of the verb stem and includes the passive morpheme. In some other languages, this domain seems to be the last sound of the second syllable. Bantu languages have indeed specific domains. The Bantu verb architecture consists of the preprefix, the prefix, the stem, the suffix, the final vowel and the postsuffix. Some of these domains are sensitive to certain rules only. In Kinyarwanda, for instance, the prosodic domain is the suprastem . This metrical domain consists of segments contained from the first mora of the second syllable of the stem to the first syllable of the object pronoun in the prefix position. Everything else is extraprosodic. Some phonological rules affect certain sounds of the verb stem only , whereas others affect those of suffixes. The domain of reduplication is bisyllabic verb stems only.
Polysyllabic stems or verb stems with extensions are excluded from reduplication.The perfective aspect is added to the right part of the reduplicated stem as seen in the following examples:

-shaaka/shaatse 'want/search'> shaakashaaka/shaakashaatse 'search/do research'
-geenda/geenze 'walk/leave'> geendageenda/geendageenze 'take a walk'

The aspect marker is obviously assigned in the final vowel slot but its phonetic effects namely consonant mutation and vowel coalescence or gliding are seen in adjacent suffixes or some parts of the verb stem which create a "modified form". The main question now is how these rules responsible for the "modified form" operate. There are three alternative solutions:
(i) the aspect marker insertion analysis, (ii) the part of the stem insertion analysis and (iii) the assimilation analysis.


3.2.1.The aspect marker insertion analysis:

The aspect marker insertion analysis entails splitting up the verb stem and inserting the aspect marker. Using the Yao verb dodoma as an example, the morpheme -il- is inserted between the vowel of the second syllable and the next consonant. The phonetic output is obtained from the following derivation: (a) morpheme insertion, (b) vowel coalescence, (c) final vowel default assignment.

-dodoma/dodweme 'hesitate'

-dodom verb stem
-dodo-il-m aspect marker insertion
-dodwem gliding and liquid deketion
dodweme default final vowel assignment

The deletion of the liquid can be accounted for by the fact that Bantu languages don't allow consonant clusters which are not homorganic and also that this liquid is deleted in many Bantu languages in some defined positions.

3.2.2.The part of stem insertion analysis

This analysis implies that once the domain of application has been identified, that part of the verb stem is inserted inside the aspect marker. This approach is also motivated by the fact that indeed some morphemes seem to be inserted in this morpheme as was noted earlier about Kinyarwanda with the passive morpheme. This also goes well with the conception of the aspect marker as a discontinuous morpheme.

-dodom verb stem
-dodo-i-m-l consonant m of the stem insertion
-dodwem-l gliding
-dodwem liquid deletion
-dodweme default final vowel assignment

3.2.3.The assimilation analysis

The assimilation analysis implies that the perfective aspect is assigned in its natural position which is before the final vowel and that other stem modifications and consonant mutation are due to assimilation.

dodom-il underlying representation
dodo-i-m-l vowel copy or assimilation
dodwem-l gliding
dodwem- liquid deletion
dodweme default final vowel assignment

Although all these three analyses are possible, the last one is more natural. It is similar to the Germanic umlaut in which vowel alternation in the verb stem to create new words or change categories is seen as an insertion as in the examples foot/feet, sing/song/sung/, give/gave, man/men, etc.when indeed it is due to an old vowel harmony rule in which the vowel quality of the vowel suffix affected the vowel of the verb stem.
The assimilation analysis is indeed the correct analysis. First, it is more natural and more universal than the other two. Second it also explains why both vowel coalescence or gliding and consonant mutation take place at the same time. It is the same phonetic feature namely tongue raising which spreads in the whole domain affecting all segments which are targets of assimilation .

IV. Conclusion

The examination of the phonetic realization of the perfective aspect in different Bantu languages has shown that it is indeed the same underlying morpheme which is strangely similar to the applicative morpheme. Since the applicative morpheme is polyfunctional, can be reduplicated, tripicated or act as a default suffix, it is more than likely that the two morphemes are historically the same but later on became doublets.
In many languages, the phonetic realization of the perfective aspect is predictable and occurs in complementary distributions. There is therefore no reason why both the perfective aspect and the imperfective aspect of the verb should be listed in the dictionary as many lexicographers from different theoretical frameworks are doing today. Further studies are necessary, however, because not only do they revive the old debate between the lexicalist and the transformationalist hypothesis and current research in grammaticalization and its importance in comparative Bantu morphophonology but also because the nature, assignment and application rules of the perfective aspect marker are not settled yet .
Before concluding, it should be noted that the behavior of this morpheme has many similarities with the suffix -ion in English when it is added to latinate verbs. Not only does the last consonant undergo both fricativization and palatalization as in divide/division, admit/admission but default suffix insertion as well! This happens when the verb stem ends with a liquid and the default suffix is
-at- as in admire/admiration, inspire/inspiration.
Bantu languages might have had two perfective aspect markers, namely the perfect and the plupefect and that the latter disappeared. Kikuyu has two aspects -ile and -ete, referring to the perfect and the pluperfect respectively. Some irregular verbs in some Bantu languages take the -ite instead of -ile as a perfective aspect morpheme as the following Yao examples show:

linda/lindite 'wait/guard'; lela/lelite 'raise'; ti/tiite 'say'; lola/loleete 'look'; kola/kwete 'possess'; waala/weete 'wear'; pila/pite' pass'.

This indicates that probably all Bantu languages had both aspects and that the pluperfect was later on dropped.

References

Bastin, Y. 1983. La finale verbale -ide et l'imbrication en Bantu. Tervuren, Belgique: Musée Royal de l'Afrique Centrale.
Chesswas, J.D. 1967. The essentials of Luganda.Oxford: Oxford University Press.
Doke, C.M.; D.M.Malcolm; J.M. Sikakana; B.W. Vilakazi. 1990. Johannesburg: Witwartersrand Univerity Press.
Guthrie, Malcolm. 1966. Grammaire et dictionnaire de Lingala. London: Gregg Press Limited.
Kimenyi, Alexandre.1979. Studies in Kinyarwanda and Bantu Phonology. Alberta, Edmonton: Linguistic Research Inc.
Kimenyi, Alexandre. 1995. The applicatives revisited. Boston University: Parasession on the syntax and semantics of applicatives.
Kimenyi, Alexandre. 1999. "Kinyarwanda causatives revisited" in Intergitations, essays for Irmengard Raush. eds. Gerard F. Carr, Wayne Harbert, & Lihua Zhang.New York: Peter Lang.
Koski, Augustus. Luganda Basic Course. State Department. Washington, D.C.
Sanderson, Meredith. 1922. A Yao grammar. London: The Macmillan Company.