Kinyarwanda Applicatives Revisited*
by Alexandre Kimenyi
California State University at Sacramento

The present paper discusses both the morphophonological and the morphosyntactic aspects of the applicative suffix -ir- in Kinyarwanda. This morpheme has different semantic functions namely temporal, locative, event-localizer, accusative, dative, benefactive, possessive, cause and goal, subjective and narrative. In a sense this morpheme is to the verb what the associative morpheme -a, which is also found in the majority of Niger-Congo languages, is to the noun where the latter happens also to be polysemous standing for multiple functions.(1) The applicative morpheme is used for null arguments for purely phonetic reasons also. The discussion will be focused on this null argument, the event-localizer case and the multiple applicative suffixation because none of these has even been addressed in Bantu syntax. The applicative morpheme is added in three predictable environments: (i) between the intensifier morpheme -iir- and the perfective aspect -ye , (ii) on causativized verbs with an applicative suffix, and verbs with the causative suffix -y- which end with the perfective aspect marker, including those which are mistakenly believed to have this causative morpheme. The trigger in most of these cases is either the perfective aspect morpheme -ye or the causative suffix -y-. I will show that in a theory which is interested in the relationship between form and function, the phonetically motivated addition of the applicative suffix to the verb stem is expected. Polyfunctional and polysemous morphemes are less marked than others and have tendency to undergo desemanticization. The locative and the event-localizer have been confused in Bantu syntax. I show with many convincing syntactic examples that these two are indeed separate arguments. Since Kinyarwanda allows multiple objects and multiple affixes (verb extensions and incorporated-object pronouns) and since the applicative morpheme assigns multiple NP objects in Kinyarwanda, there is no reason why multiple applicative morphemes would not appear within the same verb if different arguments assigned by this morpheme are present. All Bantu syntacticians, unfortunately, have failed to examine this phenomenon and to discuss its theoretical consequences. Although double objects in different languages have been found in many languages including English and have been analysed in different frameworks of Generative Grammar, multiple objects have not, so far. How does this discovery of their existence affect the argument structure, the case theory, grammatical relations, the projection principle and other modules of syntax? Since they exist in other Bantu languages (Moshi, 1995) but are marked differently, how can they be accounted for in a unified way?
Certain constraints imposed on this language namely one object per verb stem except for for ditransitive verbs which allow two unmarked objects, the clitic only argument-absorption and the strict word order of object-incorporated pronoun which is limited to only six clitics put a limit on the kind of sentences that Kinyarwanda grammar can generate. This type of constraint is desirable in generative grammar as originally formulated by Chomsky because the grammar would be too strong to produce all types of sentences. The data also convincingly show that indeed the grammar can generate sentences which have never been produced before. Certain types of sentences, especially complex ones, may not be used, not because they are ungrammatical but because of non-linguistic factors such as efficiency, memory, time, etc. This is a very interesting area for cognitive linguistics also, since it involves both complex sentence generation and processing. It is a difficult task for the descriptive linguist, however, because since many native speakers have never used or heard these sentences, have problems processing them to determine their grammaticality and find their meanings. This is an area in which computational linguistics will be helpful because using all the combinations possible, it will generate all types of sentences that the applicative construction can create and determine which ones are acceptable and which ones are processable. It is very clear, therefore, that the "performance" approach of going through texts only for empirical purposes is clearly wrong. The best approach in the generation and processing of complex sentences is to rely on the native speakers' intuition, the "competence" approach, as previously done.
The other question that is discussed here is whether, complements of applicatives, are underlyingly oblique objects dominated by the verb phrase or belong to lower clauses outside the main verb domain as originally proposed in Relational Grammar or are they assigned by cases or are they arguments of underlying abstract verbs? In other words do they join the argument structure by promotion (advancement within the same clause) like dative-shift in English or by ascension (raising) like causativization in Romance languages where promotions and demotions occur as a result of clause-union? Many competing solutions are available to account for this phenomenon but I will argue that these applicatives and their complements are still better handled within a Relational Grammar framework. This theory has to be revised, however, by including case and the incorporation analysis of Baker (1988a, 1988b) and retaining only three grammatical relations namely the subject, the object and adjuncts. The argument module is shown to be unnecessary in the syntactic structure. The same analysis also solves the puzzle as to why possessives and goals behave differently from other complements of the extension.
This chapter is organized as follows. Section one presents all meanings and functions of the applicative suffix, section two discusses the event localizer argument. Section three deals with phonetically motivated applicative suffix addition. Section four looks at the multiple applicatives at both its generation and constraints. Section five reviews all possible representations of the underlying structures of applicatives. This is followed by the concluding section.

I. The polyfunctional role of the applicative morpheme

The applicative suffix -ir- which is also realized as -er- after midvowels , e and o becaue of vowel harmony or just as -i- or -e- because of the dropping of r in front of the perfective aspect morpheme -ye, has different functions such as possessive, locative, temporal, dative, benefactive, cause/goal and subjective as illustrated in the following examples:

1. a. Umugabo a-ra-som-a igitabo 'The man is reading a book.
man he-t-read-asp
b. Umugabo a-ra-som-er-a igitabo mu nzu. 'The man is reading a book in the
house' (event-localizer)
c. Siinzí igihe umugabo a-som-ér-a igitabo. 'I don't know when the man reads the book' (temporal)
d. Umugabo a-ra-som-er-a umugoré igitabo.
'The man is reading a book to the woman' (dative)
'The man is reading a book for the woman' (benefactive)
'The man is reading the woman's book' (possessive)
Underlyingly structure for possessive:
Umugabo a-ra-som-a igitabo cy'úmugoré
man he-pres-read-asp book of woman
e. Umugabo a-ra-som-er-a igitabo amatsiko.
'The man is reading the book because for curiosity' (cause/goal)(2)
Underlyingly structure
Umugabo a-ra-som-a igitabo ku matsiko
man he-pres-read book on curiosity
f. Umugabo a-r-íi-som-er-a igitabo.
'The book is reading a book' (subjectivity) (3)
g. Umugabo a-gu-som-er-a igitabo (narrative)
man he-you-read-app-asp book
'The man read the book'.

The subjective use of -ir- , (1f), called "the middle voice" in my previous studies, can be used with any kind verb, in any kind of tense and mood. It is not clear if it is an argument or a voice, because the verb with the subjective can appear in the passive form when one of its objects is passivized.
2a. Reka umwáana a-ryaam-é, a-ra-rwáa-ye. 'Let the child lie down, he is sick'.
let child he-lie down-subj he-pres-be sick-asp
b. Reka umwáana y-ii-ryáam-ir-e, a-r-íi-rwaar-i-ye.
let child he-refl-lie down-appl-subj he-pres-refl-be sick-appl-asp
This use of the applicative is common and very productive in other Bantu languages as well such as Kilega (Kinyalolo, 1995). He calls this use of the applicative " care free".
As shown in (3), the reflexive of the subjective doesn't prevent the occurence of another reflexive pronoun.
3. Umugoré a-r-íi-yi-som-er-er-a igitabo.
woman she-pres-refl-refl-read-appl-appl-asp book
'The woman is reading the book to/for herself'.

Used with transitive verbs, it produces ambiguities, however, because the reflexive can be thought as either being an argument to the verb (dative, benefactive or possessive). 4
The narrative use of the applicative suffix, (1g), is found in verbs in past tenses only. Like the reflexive morpheme found in subjective verbs, the second or first person singular incorporated pronoun used in narratives doesn't not have any argument status either.
The applicative suffix is sometimes used for "accusative" arguments with verbs which are usually used statively.

4a Umugabo a-ra-ken-nye 'The man is poor'
man he-t-be poor-asp
b.Umugabo a-ken-e-ye amafaaranga 'The man needs money'
5a. Umugabo a-ra-tsíind-a. 'The man is going to win'
man he-pres-win-asp
b. Umugabo a-ra-tsíind-ir-a amafaraanga 'The man is going to win money'
6a. Umugoré a-ra-túm-a abáana. 'The woman is sending the children'.
woman she-pres-send-asp children
b. Umugoré a-ra-túm-ir-a abáana. 'The woman is inviting the children'.
c. Umugoré a-ra-tum-ir-a ibiíndi bitabo. 'The woman is ordering other books'.
woman she-pres-send-appl-asp other books

Before concluding this section, it should also be pointed out that the applicative suffix is used for nominalization with any kind of verb to indicate 'manners of'. The nominalized noun takes the preprefix and the prefix marker of class 4 (i-mi-) and the subjunctive suffix -e:

7. gusoma 'to read' > imisomére 'the ways of reading'
gukóra 'to work' > imikorére 'the way of working'
kugeenda 'to go/to walk'> imigeéndere 'the way of leaving/walking'
kuvúga 'to talk' > imivugíre 'the way of talking'
kuryá 'to eat' > imiriíre 'the way of eating'

Like other applicative suffixes, this one also is doubled or causes the addition of another one if the verb stem has the same structural description as verbs which cause this morpheme addition as discussed in section three:

8. gukóreesha 'to cause to work' > imikoreéshereze 'the way of causing to work'
kuvúga 'to cause to talk' > imivugiíshirize 'the way of causing to talk'

Unlike other Bantu languages such as Chichewâ, Kichaga, Kilega, etc., the Kinyarwanda applicative doesn't have instruments as its complements. These are marked by the causative suffix -iish-.

II. Event-Localizer Argument

This argument marked by the suffix -ir- indicates location or the time of the event . It is thus different from the locative as examples convincingly show. It places in space or time the doer and the event whereas the locative focuses on the theme/action itself. In other words, the event-localizer is a ground. The only feature that both the locatives and the event localizer share is the use of the locative prepositions mu, ku, and i or their respective clitics mó, hó and yó when promotion takes place. Thus (9) and (10), have distinct underlying structures.

9.Umwáana y-a-juguny-i-ye igitabo mu máazi. (event localizer)
child s/he-pst-throw-appl-asp book in water
'The child threw the book (while) in the water'.
10. Umwáana y-a-jugun-ye igitabo mu máazi. (locative)
child s/he-pst-throw-asp book in water
'The child threw the book into the water'

In (10), mu máazi shows destination whereas in (9) it is ground.
Position verbs such as "to sit"; "to lean"; "to lie down", etc, don't allow the applicative suffix because they already have the event-localizer argument in them, they don't need a ground. The applicative suffix is allowed only if the locative preposition is not there.

11a. kwéegama ku gití 'to lean on the tree'
b. kwéegama hó igití
c. kwéegamira igití
d. *kwéegamira ku gití
12a. kuryáama ku mwáana 'to lie on the child'
b. kuryáama hó umwáana
c. kuryáamira umwáana
d. *kuryáamira ku mwáana
13a. kwiicara kuu ntébe 'to sit on the chair'
b. kwiicara hó intébe
c. kwiicarira intébe
d. *kwiicarira kuu ntébe

Semantically, the locative refers to source, destination, location and conduit. Some conduit meanings, however, eventhough they share the same preposition as other locatives, don't allow promotion:

14a. Umugoré y-a-gii-ye muu módoká. 'The woman went by car'
woman she-pst-go-asp in car
b. *Umugoré y-a-gii-ye mó imódoká.
woman she-pst-go-asp in car
15. Umugoré y-a-gii-ye n'íimódoká. 'The woman went by car'

Another evidence that indeed the event-localizer is not a locative is that another locative can appear with the event-localizer within the same verb:

16a. Umwáana y-a-juguny-i-ye mu máazi igitabo mu gihuru.
child he-pst-throw-appl-asp in water book in bush
'From the water, the child threw the book in the bush'.
b. Umwáana y-a-juguny-i-ye mó amáazi igitabo mu gihuru.
c. Umwáana y-a-juguny-iy-e amáazi mó igitabo mu gihuru.
d. Umwáana y-a-juguny-i-ye mó igihuru igitabo mu máazi.
*'The child threw the book in bush (while) in the water'.
'The child threw the book in the water while in the bush'.

It should be pointed out from the outset, however, that the locative cannot be "objectiivized" and acquire structural syntactic properties of other direct objects when the verb has an event localizer argument. Thus the preposition mu of mu máazi can not become a clitic to the verb and amáazi come closer to the verb. The noun amáazi of the locative phrase cannot be pronoun-incorporated, cannot relativize, passivize, cleft, etc.
As we have already noticed with the examples shown, the locative prepositions for both the locative and the event-localizer can become clitics to the verb. This clitic becomes an object pronoun in both cases when the head is deleted.

17a. Umwáana y-a-juguny-i-ye amáazi mó igitabo
child he-pst-throw-appl-asp water in book
'The child threw the book from the water'
b. Umwáana y-a-juguny-i-ye mó amáazi igitabo.
child he-pst-throw-appl-asp in water book
'The book threw the book from it/there'
ci. Umwáana y-a-juguny-i-ye mó igitabo
child he-pst-throw-appl in book
'The child threw the book from it'
ii. Umwáana y-a-(ya)-juguny-i-ye mó igitabo
child he-pst-throw-appl-asp in book
'The child threw the book from it'
iii. Umwáana y-a-(ha)-juguny-i-ye mó igitabo
child he-pst-it-throw-appl-asp in book
18a. Umwáana y-a-juguny -e igitabo mu máazi.
child s/he-pst-throw-asp book in water
'The child threw the book in the water'
bi. Umwáana y-a-juguny-e amáazi mó igitabo.
ii.Umwáana y-a-juguguny-e mó amáazi igitabo
ci. Umwáana y-a-juguny-e mó igitabo
child he-pst-throw-asp in book
'The child threw the book in(to) it'
ii. Umwáana y-a-(ya)-juguny-e mó igitabo.
child he-pst-it-throw-asp in book
iii.Umwáana y-a-ha-juguny-e mó igitabo.
child he-pst-it-throw-asp in book

Note that when the preposition becomes a clitic without a head, it absorbs the case. It acts as both preposition and pronoun. (in+Pro, on+Pro, at+Pro...)
The prepositional phrase NP cannot be assigned two arguement at the same time.
Thus whenever a preposition is promoted to a clitic, it acts as both a preposition and a clitic. This seems to be true for many languages including English.

19. Put the cover on the cooking put > Put the cover on.
She hid the book between the shelves > She hid the book between.
The dictionary was under the table > The dictionary was under.
20. Ils ont glissé la clé dessous la porte > Ils ont glissé la clé dessous.
'They slipped the key under the door' 'They slipped the key under'
Nous étions dedans la maison > Nous étions dedans.
'We were inside the house' 'We were inside'

Another phenomenon which seems to be interesting also, is that eventhough a temporal can act as a complement of the applicative morpheme, it doesn't licence the applicative suffix on the verb if the temporal occurs in the main clause.

21. Umugabo a-zaa-garuk-ir-a i Kigalí
man he-fut-return-appl-asp at Kigali
'The man return from Kigali'.
22a. *Umugabo a-zaa-garuk-ir-a ryáarí?
man he-fut-return-appl-asp when
'When is the man coming back'?
b. Ni ryáari umugabo a-záa-garuk-(ir)-a?
be when man he-fut-return-appl-asp
c. Sii-n-zí igihe umugabo a-záa-garuk-ir-a.
be not-I-know time man he-fut-return-appl-asp
'I don't know when the man will come back'.
23. Abáana ba-á-du-hamagar-i-ye i Kigalí.
children they-pst-us-call-appl-asp at Kigali
'The children called us from Kigali
24a. *Abáana ba-á-du-hamagar-i-ye ku cyuumwéeru.
children they-pst-us-call-appl-asp on Sunday
'The children called us on Sunday'.
b. Ni ku cyuumwéeru abáana ba-á-du-hamagar-(i)-ye.
be on Sunday children they-pst-us-call-appl-asp
'It is on Sunday that the children called us'.
c. Ku cyuumwéeru ni hó abáana ba-á-du-hamagar-i-ye.
on Sunday be when children they-pst-us-call-appl-asp
Sunday is when the children called us'.
d. *Ku cyuumwéeru ni bwó abáana ba-á-du-hamagar-i-ye
on Sunday be when children they-pst-us-call-appl-asp
e. Ku cyuumwéeru ni bwó ba-á-du-hamaga-ye.
'Sunday is when the children called us'.

Usually, the majority of grammatical processes in Kinyarwanda occur in situ. For instance, wh-qustioning is allowed only in situ. The applicative suffix marking on temporal event localizers only in derivational phrases is an exception.
Although hó and bwó are used interchangeably in pseudo-cleft sentences with temporal antecedents, it is again clear in 24d that bwó cannot licence the applicative because only locatives are inherently ground generators. Note that (25) is ambiguous.

25. Sii-n-zí ahó abáana ba-záa-du-hámagar-ir-a.
not be-I-know where/when children they-fut-us-call-appl-asp
'We don't know where the children will call us from'.
'We don't know when the chidren will call us'.

III. Null argument or Phonetically motivated applicative addition

Not all applicative morphemes are assigned complements. This section shows many instances in which an applicative suffix is added to the verb without adding any morphological, syntactic or functional information. It is actually wrong to call this absence of a complement as a null argument like the case of locative clitics found with locatives or event localizers as seen in the previous section. There, a complement exists in the underlying structure. In this section, there is no complement at all. The appearance of the applicative is only phonetically or historically motivated.
There are thus three important questions to be answered: First why is the applicative suffix added? Second, why in some cases, is the causative morpheme doubled when the applicative morpheme is added and third, why polysyllabic stems which end with z don't have the applicative suffix?
This addition of both the applicative morpheme and the causative morpheme also takes into consideration other phonological factors such as basic segments and derived consonants, short and long vowels, and the number of syllables.
This study is very important because first it shows that some rules have to remember the derivational history of their structural descriptions whereas others only apply to the output structures. It also shows that native speakers sometimes mistakenly confuse words which have the same phonetic form as having the same underlying structure thus making errors into rules. The polysyllabic stems which end with z don't allow -ir- addition because this z, whether mutated or basic, is thought of having been derived from the combination of r and the causative marker -y-.
This applicative morpheme addition although initially seems to apply in an arbitrary fashion, under careful analysis, however, supports the iconicity view of language that signs and structures are not arbitrary but motivated.

3.1. The effect of the applicative suffix on phonology:

Before instances of null argument applicatives are presented, we should first discuss the effect of this morpheme to the verbal phonology of Kinyarwanda.
The perfective aspect -ye when added to the last consonant of the stem causes what is known as consonant mutation resulting in the majority of cases in palatalized consonants. In other Bantu languages, the verb stem last consonant becomes totally unpredictable and this is what René Bastin calls imbrication and linguists writing Bantu dictionaries have to list both the verb stem and the verb stem combined with the perfective aspect. Below, are presented the Kinyarwanda mutated consonants.

26. b ----> by kureeba 'to look' areebye 'he just looked' (4)
m ----> my kurema 'to create' aremye 'he just created'
p -----> py gukopa 'to lend' akopye 'he just lended'
d -----> z kudóda 'to sew' adoze 'he just sawed'
t -----> s gukúbita 'to hit' akubise 'he just hit'
z -----> j kubaanza 'to start' abaanje 'he just started'
s -----> sh kurása 'to shoot' arashe 'he just shot'
r -----> z gukóra 'to touch' akoze 'he just touched'
VVr ---->Ø kuráara 'to spend the night' araaye 'he just spend the
night'
n -----> nny gukena 'to need' akennye 'he needs'
sh ----> sh guhísha 'to hide' ahishe 'he just hid'
g ---> z kuroga 'to poison' aroze 'he just poisoned'
k ---> ts kureka 'to abandon' aretse 'he just abandoned'
h ---> shy kuriha 'to pay' arishye 'he just paid'

z, s, and ts are the only mutated consonants which are not palatal. It is also worth noting that these are all alveolar.
Palatal consonants don't change, ofcourse, as the following examples show:

27.ny ----> ny guséenya'to destroy' araséenye/a-ra-séeny-ye/ 'he just destroyed'
sh ---->sh guhísha 'to hide'arahíshe/a-ra-hish-ye/ 'he just hid'
j---->j gukóonja'to get cold' arakóonje/a-ra-kóonj-ye/ 'he is cold'
shy--->shy kubéeshya 'to lie' arabéeshye /a-ra-béeshy-ye/ 'he just lied'

It is also important to note that some consonants have different "mutated" allophones depending on whether the preceding vowel is short or long, or belongs to the stem or is a suffix.
Thus r becomes z after a short vowel but zero after a long vowel and n is nny after a short vowel but ny after a long one or a suffix

28. r---> z akoze /a-kor-ye/ 'he just worked'
---> ø araaye /a-raar-ye/ 'he just spent the night'
n--->nny avunnye /a-vun-ye/ 'he just broke'
--->ny if -an- akoranye /a-kor-an-ye/ 'he worked with'

3.2. Incompatibility between -y- and -yi

Three morphemes which share the palatal glide, the perfective aspect marker -ye, the causative morpheme -y- and the nominalizer -yi all produce consonant mutation.
It is important to realize that there is neutralization when both -y- and -yi occur in the same stem:
A polysyllabic verb stem which ends with z doesn't mutate into j because this z is thought of having resulted from consonant mutation (r+y--> z)(5)

29. umucúruuzi /u-mu-cúuruz-yi/ 'seller' <--- gucúruuza 'to sell'
umubúguzi /u-mu-búguz-yi/ 'igisoro player' <--- kubúguza 'to play igisoro'
umusiinzirizi /u-mu-siinziir-ir-ir-y-yi/ 'sleeper'<-- gusiinziriza 'to close eyes'

The above examples must be reinterpreted as /umu-cúruur-yi/, /umu-búgur-yi/... in the underlying structure.
Note that bisyllabic verb stems mutate, however:

30. umubaaji /u-mu-baaz-yi/ 'sawyer' (6)
ingaanji /in-gaanz-yi/ 'winner'

3.3.. Markedness of palatalized consonants

The palatalized consonants are usually marked and so is the verb stem last syllable. These palatal consonants appear freely elsewhere but not in the verb stem final syllable. The following palatals are found in Kinyarwnda:

31. 1. palatalized bilabials: by, py and my
2. palatalized alveolars: dy, ty, sy, nny, and ry
3. alveo-palatals: sh, c, j
4. velo-palatals: cy, jy, shy and ny

Palatalized alveolars (ty, dy, sy, ry, nny) are never found in the stem final position except sy, ry and nny in three monosyllabic verbs but also at the phonetic level as a result of patalalization:

32. sy: gusya /ku-se-a/ 'to grind';
ry: kuryá /ku-rí-a/'to eat';
nny: kunnya/ku-ne-a/ 'to defecate'.

Not found either are palatalized velars cy and jy, except again in two verbs kujya /ku-gi-a/ 'to go' and gucyá /ku-ké-a/ 'to be day', in which again they have been obtained through palatalization. There is no single verb either which has the voiced alveopalatal fricative j and the voiceless alveo-palatal affricate c, in the stem final position but what is interesting, prenasalized j (nj) and prenasalized jy (njy) are found as in kuvúunja 'to exchange money', gukíinja 'to butcher', gushíinja 'to inculpate' and kubóonjya 'to taste an alcoholic beverage', respectively.

Verb stems with palatalized bilabials (py, by, my) in the final position although numerous don't seem to behave like basic palatals, either. They behave exactly like mutated consonants.
Thus ny, shy, and sh are the only true palatal consonants which can appear unmutated in the verb stem final position.

There are simple consonants which also seem to be marked. These are all alveolars namely, d, s and z. But z and s can also be either basic or "mutated" consonants, as seen earlier.
The voiced alveolar stop d is the most marked. Very few verbs are found which have it in the stem final position.

3. 4. The applicative morpheme -ir- with the perfective aspect

Several phonological rules apply when the applicative suffix and the perfective aspect are in contact.
The r of the applicative drops before -ye. This rule is formulated and illustrated in (30).

33. V-ir-ye --->V-i-ye

akin-ir-ye ---> akiniye 'he just played'
aror-ir-ye ---> aroreye 'he just looked at'

The r of all polysyllabic verb stems and bisyllabic stems with long vowels before the aspect -ye is also dropped:

34. a-kurur-a 'he pulls' ---a-kurur-ye [akuruye] 'he just pulled'
a-koror-a 'he coughs'---a-koror-ye [akoroye] 'he just coughed'
a-vuur-a 'he cures' --- a-vuur-ye [avuuye] 'he just cured'
a-tiir-a 'he borrows' --- a-tiir-ye [atiiye] 'he just borrowed'

There is some kind of irregularity, however, with bisyllabic stems which start with a vowel:

35. h-iir-a 'it gets dark' ---ha-iir-ye--> [hiije] 'it just got dark'
h-aar-a 'it gets big' --- ha-aar-ye--> [haaye] 'it just got big'
h-eer-a 'it grows' --- ha-éer-ye--> [heeze] 'it just grew'
a-ra-hágarar-ye --->arahágaraye or arahágaze 's/he is standing up'

3. 5. The perfective aspect with a stem with the causative morpheme -iish-

When the perfective aspect -ye is added to a verb stem with the causative morpheme -iish-, not only is the applicative suffix added but the other causative -y- is also found just before the perfective aspect marker, resulting in causative doubling. (36) also formulates and illustrates this phenomenon.

36 ...iish-ye ---> ...iish-ir-y-ye

a-kin-iish-ye --->akin-iish-ir-y-ye [akiniishije] 'he just caused to play'
a-kin-iish-a --->akiniisha 'he causes to play'
a-ror-iish-ye --->aror-iish-ir-y-ye [aroreesheje] 'he just caused to look'
a-ror-iish-a --->aroreesha 'he causes to look'

The r of the applicative morpheme and the causative morpheme mutate into z and in turn, the mutated consonant mutates again into j when followed by the aspect marker -ye.

37. (i)...r+y..-----> z; (ii) z+ye ------>j

The applicative suffix is added to the verb stem with the causative morpheme -iish- regardless of aspect if there is another applicative morpheme in the verb (applicative suffix doubling). This is again stated and exemplified in (37).

38. ...iish-ir...----> ...iish-ir-ir-y-{a}
{ye}

a-kin-iish-ir-a --->akiniishiriza /a-kin-iish-ir-ir-y-a /'he causes to play for'
a-kin-iish-ir-ye --->akiniishirije /a-kin-iish-ir-ir-y-ye 'he just caused
to play with'
a-ror-iish-ir-a --->aroreeshereza/a-ror-iish-ir-ir-y-a/
'he causes to look with'
a-ror-iish-ir-ye --->aroreeshereje /a-ror-iish-ir-ir-y-ye/
'he just caused to look with'

3. 6. The perfective aspect with the -y- suffix

Certain causativized verbs with the morpheme -y- have the applied suffix added in front of the perfective aspect marker -ye. The applicative addition seems to be obligatory with bisyllabic stems only.
There seems to exist a rule of the following type:

40. ...V+y+ye ---> ...V+y+ir+y+ye
avuura 'he cures' --- avuuza(a-vuur-y-a) 'he causes to cure'
avuuye 'he just cured'--- avuuje/avuujije(a-vuur-y-ir-y-ye) 'he just caused to cure'
araara 'he sleeps' --- araaza(a-raar-y-a) 'he sleeps with'
araaye 'he just slept' --- araaje/araajije(a-raar-y-ir-y-ye) 'he just slept with'

What is interesting about these bisyllabic verbs which allow both forms is that they all have a long vowel in the stem and all end with a fricative s, z, j or sh. Short vowel bisyllabic stems fall in either one of the two categories.

41a. arasa 's/he shoots' ---arashe
agesa 's/he harvests millet' ---ageshe
amesa 's/he washes' ---ameshe
zigisha 'they go into transmance' ---zigishe
asasa 's/he makes bed' --- ashashe
ashisha 's/he gets fat' --- ashishe
abisha 's/he becomes angry'---abishe
agásha 's/he harasses' ---agashe
agosha 's/he swalls' ---agoshe
aheenja 's/he walks awkwardly' --aheenje
agasa 's/he puts in suspens' ---agashe
b. abisa 's/leaves alone' --- abishije
abaza 's/he asks' --- abajije
akosa 's/he makes mistake' ---akosheje
ahusha 's/he misses target' ---ahushije

Note that all bisyllabic stems with a short vowel which end with z (either basic or mutated) always have the -ir- added.

42. abaza 'he asks' ----abajije/*abaje
agaza 'he sits in large space' ---agajije/*agaje
akoza 'he touches with' --- akojeje/*akoje
abuza 'he prevents' ---abujije/*abuje
akiza 'he cures' --- akijije/*akije

It is thus a mirror image of polysyllabic stems which end with z. These don't allow -ir- addition whereas the -ir- addition is obligatory with the former.
Other mutated consonants of polysyllabic verb stems, except -z-, always add the applied suffix.

43. araambuka 'he crosses' --- araambutsa /a-ra-ambuk-y-a/ 'he makes cross'
araambutse 'he just crossed' ---araambukije/-a-ra-ambuk-y-ye/ 'he just made cross'
atuuba 'he decreases' ----- atuubya /a-tuub-y-a/ 'he causes to crease
atuubye 'he just decreased' --- atuubije /a-tuub-y-ye/ ' he just caused to decrease'

3. 7. Causative doubling : ...y....y..

If the last consonant of the stem is alveolar (t, d, s, z, r, n) thus mutating into s, z, sh, j, z and ny, respectively, the causative y is doubled in front of the perfective suffix -ye and one is kept near the stem and ofcourse the applied morpheme is also added:

44 ...{t, d, s, z, r, n}+y+ye ---> ...{t, d, s, z, r, n}+y+ir+y+ye
[V] [V]
atigita --- atigisa /a-tigit-y-a/ ----atigishije /a-tigit-y-ir-y-ye/ 'he just shook'
atogota -- atogosa /a-togot-y-a/ ---atogosheje /a-togot-y-ir-y-ye/ 'he just boiled'
agaanda -- agaanza ---agaanjije /a-gaand-y-ir-y-ye/ 'he just won'

If the causative morpheme was not doubled, there would indeed be neutralization between the stem with the applicative suffix and the one without it in the perfective aspect:

45. / arir-y-ye/ --->arir-y-ir-y-ye [arijije] 'he just made cry'
/ arir-ir-y-ye/ --->aririz-ye --- [aririje] 'he just made cry for'

If the last consonant of the stem is the n belonging to a suffix, causative doubling is optional:

46. a-toong-an-a 'he argues' ---- a-toong-an-y-a 'he argues with'
a-toong-an-ye 'he just argued'---a-toong-an-y-ye--->a-toong-an-(y)-ir-y-ye ---
atoonganyije/atoonganije 'he just argued with'
a-rwaan-a 'he fights' --- a-rwaan-y-a 'he fights with'
a-rwaan-ye 'he just fought' --- a-rwaan-y-ye --->a-rwaan-(y)-ir-y-ye
arwaanyije/arwaanije 'he just fought with'

It seems like the causative doubling -y...y... with causativized verbs in the perfective aspect is the rule than the exception. This implies that an applied causativized verb (-iish-) has this y doubled also, one next to -iish- and the other one just before the perfective aspect.

47. akoreesheje 'he used' ---/a-kor-iish-y-ir-y-ye/

The one next to -iish- doesn't show up because of neutralization next to a palatal consonant.
The rule given earlier has thus to be revised as:

48. ...V+iish+ye ---->...V+iish+y+ir+y+ye

3. 8. The obligatory insertion of -ir- without the causative morpheme

Some verbs which clearly don't have the causatived suffix with them, have the applicative -ir- added also. What all these verbs have in common is the last consonant which is a coronal fricative or a palatalized consonant. They thus ressemble mutated consonants which also fall into these two categories: palatals or coronal fricatives.

49. abaza 'he asks' --- abajije 'he just asked' *abaje
afasha 'he helps' ---afashije 'he just helped' *afashe
arusha 'he surpasses' --- arushije *arushe
abisa 'he leaves alone' --- abishije *abishe
ahusha 'he misses' --- ahushije *ahushe
acugusa 'he shakes' --- acugushije *acugushe
akosa 'he makes mistakes' --- akosheje *akoshe
ahoosha 'he stops' ---ahoosheje/ *ahooshe

3. 9. -ir- doubling with the causative morpheme -y-

Whenever a verb stem with the causative morpheme -y- has an applicative suffix, the latter is always automatically doubled and the causative suffix comes at the end just preceding the aspect marker.

50. kwaambuka 'to cross' --- kwaambutsa /ku-ambuk-y-a/---kwaambukiriza /ku-ambuk-ir-ir-y-a/
gushyúuha 'to be warm' --- gushyúushya /ku-shyúuh-y-a/ ---gushyúuhiriza
/ku-shyúuh-ir-ir-y-a/
gutogota 'to boil' --- gutogosa /ku-togot-y-a/ --- gutogotereza
/ku-togot-ir-ir-y-a/

Note again, however, that this applicative doubling is not found with z mutated from r+y:

51. kwóoga 'to wash the body' ---kwóoza /ku-óg-y-a/ ---kwóogereza
/ku-og-ir-ir-y-a/
gukíra 'to recover/be rich' ---gukíza /ku-kír-y-a/ ---gukíriza
/ku-kír-ir-y-a/
but not *gukíririza /ku-kír-ir-ir-ya/
gutiira 'to borrow' --- gutiiza /ku-tiir-y-a/ --- gutiiriza /ku-tiir-ir-y-a/
'to lend'
but not *gutiiririza /ku-tiir-ir-ir-y-a/
guhóra 'to stop crying' --- guhóza /ku-hór-y-a/ ---guhóreza
/ku-hór-ir-y-a/
but not *guhorereza

gutiinda 'to be late' --- gutiinza /ku-tiind-y-a/ --- gutiindiriza
/ku-tiind-ir-ir-y-a/
kubúunga 'to wander' --- kubúunza /ku-búung-y-a/ ---kubúungiriza
/ku-búung-ir-ir-y-a/
kugeenda 'to go/walk'--- kugeenza /ku-geend-y-a/ ---kugeendereza
/ku-geend-ir-ir-y-a/

3. 10. Other cases of obligatory -ir- addition.

The -ir- is also added to stems which end with -iir- and end with the perfective suffix -ye.
As examples also show, -iir- can shorten before the applied suffix.

52. asiinziira 'he sleeps' ---asiinziiriye/asiinziriye 'he is sleeping'
avoomeera 'he waters' ---avoomeereye/avoomereye 'he just watered'
aheengeera 'he waits for' ---aheengeereye/aheengereye 'he just waited for'
araangiira 'he resonates' ---araangiiriye/araangiriye 'he just resonated'
arahíirwa 's/he is lucky'--- arahíiriwe 's/he is lucky'
arabéerwa 's/he looks nice' ---arabéerewe 's/he looks nice'

Monosyllabic verbs with an applicative suffix have it doubled also before the perfective aspect:

53. a-ca 'he cuts' a-c-iir-a 'he cuts for' a-c-iir-ye --->a-c-iir-i-ye
a-ba 'he is' a-b-eer-a 'he favors' a-b-eer-ye --->a-b-eer-e-ye
a-ta 'he throws' a-t-eer-a 'he throws for' a-t-eer-ye --->a-t-eer-e-ye
a-nywa 'he drinks' a-ny-eer-a 'he drinks for' a-nyw-eer-ye --->
a-nyw-eer-e-ye
a-pfa 'he dies' a-pf-iir-a 'he dies for' a-pf-iir-ye ---> a-pf-iir-i-ye

The nominalizer -yi also adds the applied suffix to verb stems which end with -iir-:

54. -V-iir-yi ---> -V-iir-ir-yi

kuzéreera 'to wander' --->inzérerezi /in-zéreerer-ir-yi/ 'wanderer'
guteembeera 'to take a walk'--->umuteemberezi /u-mu-teembeer-ir-yi/
'migrant worker'
guséembeera 'to look for lodging'--->umuseemberezi 'lodgee'
/u-mu-seembeer-ir-yi/
guséengeera 'buy alcoholic drinks'-->umuséengerezi 'beer customer'
/u-mu-séengeer-ir-yi/

3. 11. Verbs which allow both forms

Some verbs, as examples in (55) show, allow both forms: With or without the addition of the applied suffix before the perfective morpheme -ye. (55)a gives examples of verb stems with basic consonants and (55)b verb stems with mutated consonants:

55a. asiiza 'he levels' ---ashiije/ashiijije
abaanza 'he begins'---abaanje/abaanjije
atuuza 'he remains calm'---atuuje/atuujije
agaanza 'he wins' --- agaanje/agaanjije
amiisha 'he bleeds' --- amiishe/amiishije
avuunja 'he changes' --- avuunje/avuunjije
asooza 'he finishes' --- ashooje/ashoojeje
iruuza 'it chews' --- iruuje/iruujije
areesa 'he knocks down' --- areeshe/areesheje
yiitaaza 'he avoids' --- yiitaaje/yiitaajije
yaasa 'he axes' --- yaashe/yaashije
aceeza 'he danses' --- aceeje/aceejeje
abaasha 'he is able' --- abaashe/abaashije

b. araara/araaza 'he sleeps' ---araaje/araajije
atyaara/atyaaza 'he sharpens' ---atyaaje/atyaajije
avuura/avuuza 'he cures' --- avuuje/avuujije
ahuura/ahuuza 'he meets' --- ahuuje/ahuujije
yeera/yeeza 'he cleans' --- yeeje/yeejeje
ateera/ateeza 'he attacks' ---ateeje/ateejeje
atiira/atiiza 'he lends' ---atiije/atiijije
ageenda/ageenza 'he drives' ---ageenje/ageenjeje
ireenga/ireenza 'it goes beyond' ---ireenze/ireengeje

The reason why verbs in 55a allow both forms is because their infinitives also allow both forms.

56. kubaanza 'to start' kubaanzira/kubaanziriza
kugaanza 'to win' kugaanzira/kugaanziriza
kweesa 'to knock down' kweesera/kweesereza

These verbs have become doublets so to speak.
There is another category of uncausativized verbs which also undergoes applicative addition not because they have doublets but because they are thought of being causativized. This is revealed when the -ir- suffix is added to them. When this morpheme is added to them, it is doubled exactly like other verbs which are causativized.

57. kubáasha 'to be capable' --- kubáashiriza (*kubáashira)
kurusha 'to surpass' --- kurushiriza (*kurushira)
kubísa 'to give way' --- kubísiriza (*kubisira)
gucugusa 'to shake' --- gucugusiriza (*gucugusira)
gukosa 'to err' --- gukosereza (*gukósera)
gufásha 'to help' --- gufáshiriza (*gufáshira)
gushushaanya 'to draw' --- gushushaanyiriza (*gushushanyira)

It is verb stems which end with -z only which don't allow doubling of causative morphemes, because this z is always mistaken as having resulted from consonant mutation as we observed earlier:

58. kubáza 'to ask' --- kubáriza *kubázira
kubaaza 'to saw' -- kubaariza *kubaazira

There are two tests which clearly show that this is the case, namely the addition of the y suffix when the applied suffix is added, and the addition of the causative suffix to the reciprocal morpheme -an-:
When the recipral suffix -an- is added to a causativized verb stem, this nasal is also palatalized. It is also noted that these same uncausatived verbs which allow the applicative addition, have this reciprocal nasal palatalized also.

59a kubáasha{nya}'to share the same strength' <---> kumesa{na} 'to wash with'
kurusha{nya} 'to surpass each other' <--- >kurása{na} 'to shoot with;
kubísa{nya} 'to avoid each other' <--- >gusáaza{na} 'to get old with'
kubaaza{nya}'to ask each other' <---> gukíinja{na} 'to butch with'
gukosa{nya} 'to make mistakes with' <--- >guhísha{na} 'to hide with'
gufásha{nya}'to help each other' <--->gushísha(na) 'to frigthen with/to become fat with'
b.kumesa 'to wash' --- kumesera
kurása 'to shoot' --- kurásira
gusáaza 'to age' --- gusáazira
gukíinja 'to butcher' --- gukíinjira
guhísha 'to hide' --- guhíshira

This phenomenon thus seems to be less arbitrary when some verbs with palatal/alveolar fricatives don't allow the -ir- addition

60. ahisha 'he hides' --- ahishe / *ahishije
asaaza 'he gets old' --- ashaaje/ *ashaajije
asoonza 'he gets hungry' --- ashoonje/ *ashoonjeje
ashisha 'he gets fat' --- ashishe/ *ashishije
arasa 'he shoots' --- arashe/*arashije
agasa 'he refuses' --- agashe/*agashije
acweeza 'he keeps quiet --- acweeje/*acweejeje
agosha 'he has rushes' --- agoshe/*agosheje
atuuza 'he keeps quiet' --- atuuje/*atuujije
asasa 'he makes the bed' --- ashashe/*ashashije

Unlike those in 55a and 55b, these ones don't have doublets or mutated last consonants.

3. 13. Causativized verbs without -ir- addition

Causativized bisyllabic verb stems have always to have -ir- added before the -ye morpheme except ofcourse verbs discussed above which have doublet stems. Polysyllabic verb stems which end with z either basic or mutated (obtained with the causative morpheme -y- : r+y) don't have -ir- added, however.

61. ataakaza /a-taakar-y-a/ 'he loses' ---ataakaje /a-taakar-y-ye/ *ataakajije
ababaza /a-babar-y-a/ 'he makes suffer' ---ababaje /a-babar-y-ye/ *ababajije
akiinguuza/a-kiinguur-y-a/ 'he opens' ---akiinguuje /a-kiinguur-y-ye/ *akiinguujije
araangiza /a-raangir-y-a/ 'he finishes' --- araangije /a-raangir-y-ye/ *araangijije

62. abuguza 'he plays igisoro' ---abuguje *abugujije
acuruza 'he does business' ---acuruje *acurujije

The reason why -ir- addition is not found with poysyllabic stems which end with z, is because this z whether mutated or basic, is thought of as being resulted from the combination of r and y. This is revealed when the applied morpheme is added to the stem. The segment z always follows the applicative morpheme:

63. abuza /a-bur-y-a/ 'he prevents' ---aburiza /a-bur-ir-y-a/ 'he prevents for'/*abuzira
avuuza /a-vuur-y-a/ 'he cures' ---avuuriza /a-vuur-ir-y-a/ 'he cures for'/*avuuzira
ahamiriza /a-hamiriz-a/'he dances' ---ahamiririza /a-hamiriz-ir-a/
*ahamirizira 'he dances for'
acuruuza /a-curuuz-a/ 'he sells' --- acuruuriza /a-curuuz-ir-a/
*acuruuzira 'he sells for'

If the applied suffix were added to the polysyllabic stem with a z ending with a perfective aspect marker, there would be neutralization between the applied and non-applied form because both have to undergo the same rules namely the reanalysis (z: r+y ) and consonant mutation (r+y > z and z+y > j).

64. a-curuuz-ye --->a-curuur-ir-y-ye [acuruurije]
a-curuuz-ir-y-ye ---> a-curuur-ir-y-ye [acuruurije]

The frequentative -iriz- doesn't allow any other -ir- insertion.
If it did, there would be neutralization also because these verb forms would have a benefactive reading.

65. avugiriza 'he whistles' --- avugirije'he just wistled'/ *avugirijije 'he whistles for'
ategereza 'he waits' --- ategereje 'he just waited'/ *ategerejeje 'he waits for'
ahamiriza 'he dances' --- ahamirije 'he just danced'/ *ahamirijije
ateekereza 'he thinks' --- ateekereje 'he just thought'/ *ateekerejeje
asiinziriza 'he sleeps' --- 'he just slept'asiinzirije/ *asiinzirijije
asabiriza 'he begs' --- 'he just begged' asabirije/ *asabirijije

This morpheme consists also of the doubling of the applicative suffix and the causative morpheme:
-iriz- : ir-ir-y
Note that when the reciprocal morpheme -an- is added, there is also causative doubling (-y- ...-y-)
like other verb stems which have the causative morpheme:

66. asabirizanya 'he begs with' /a-sab-ir-ir-y-an-y-a/
ategerezanya 'he waits with' /a-teg-ir-ir-y-an-y-a/
ahamirizanya 'he dances with' /a-ham-ir-ir-y-an-y-a/

There are thus three factors which are responsible for the -ir- addition, first it is to avoid neutralization or ambiguity, second it is because some verbs have doublet stems ( with or without a causative suffix), and third it is because of historical reasons. Some verbs eventhough synchronically do not have a non-causativized corresponding verb stem are shown to have historically had one:

(67). gufásha (-fashije) 'to help'/ 'to hold for sbdy' <---- gufáta 'to catch'/ 'to hold'
kubáza (-bajije) 'to ask' <---- kubára 'to tell'
kurusha (-rushije) 'to surpass' <---- kuruta 'to be bigger in height/size/age/
character
gushush-aan-ya (-shushaanyije) 'to draw' <---- ishusho 'picture'
kuruund-aan-ya (-ruundaanyije) 'to pile' <---- kuruunda 'to pile'

Palatalized labial stops in the verb stem final position are also shown to be a new development or shown to have counterparts which are not palatalized, thus not being part of the stem at all:

(68). kurábya 'to blossom' ---- ururábo 'flower' (7)
kuramya 'to praise' ---- kurama 'to live long'
kugumya 'to keep doing'-- kuguma 'to stay'

This z in the verb stem final position has also resulted in some cases from linguistic contamination. The frequentative morpheme -agur- is sometimes realized as -aguz- or -aguzw- eventhoug there is no causative meaning involved:

70a.kuráaraguza 'to spend nigths here and there'
kubúundaguza 'to ask for lodging'
kuryáagaguza 'to keep eating'
kureebaguza 'to look here and there'
kuraandaguza/kuraandagura 'to uproot'
gutáagaguza 'to waste'
gushwáanyaguza/gura 'to tear up'
gucáagaguza/gura 'to tear into pieces'
gucúnaguza 'to reprimand'
gukébaguza 'to look around'

b.kuvúgaguzwa 'to talk nonsense'
gutéeshaguzwa 'to talk nonsense'
kuráangaguzwa 'to be distracted'
kwíineemfaguz(w)a 'to pretend not to like something'

This morphophonological study on the realization of the perfective aspect -ye supports the iconicity view in the genesis of linguistic signs and structures. The motivation of the addition and the selection of the applicative -ir- morpheme is clear. First, the language chooses among morphemes which already exist instead of creating new ones. It should also be pointed out that applicative suffix addition and doubling is not unique to Kinyarwanda. This phenomenon is common to many Bantu languages.
There are other cases of phonetically motivated segment and morpheme addition, in Kinyarwanda, namely, the addition of z with nouns of classes 9 and 10 which start with a vowel and the addition of -ku- in some tenses of verbs which also start with a vowel.
a. -z-:

71. urwáara /uru-ára/ 'nail' -- inzáara /inø-ára/ 'nails'
icyúuma /iki-úma/ 'knife' --inzúuma /inø-úma/ 'pins'

This -z- addition is not arbitrary. This consonant is historically motivated. It is a lost consonant which comes back. There are indeed words with y as the initial stem consonant but which changes into z when preceded by a nasal:

72. inzóka 'snake' --- akayóka 'small snake'
inzira 'road' --- akayira 'small road'
inzovu 'elephant' ---akayovu 'small elephant'

The addition of -ku- which in the conditional tenses of verb stems which start with a vowel is ofcourse phonetically motivated. The reason it is -ku- which is chosen it is again because first it is the infinitive marker and secondly because it is again unmarked semantically. It doen't affect meaning or has any new information to add.

73a. ku-bón-a 'to see'
ku-gir-a 'to have'

b. ba-aa-bón-a -->baabóna 'they would see'
ba-aa-gir-a --> baagira 'they would have'

c. ba-aa-bon-á --> baaboná 'if they could see'
ba-aa-kor-á --> baakorá 'if they could have'

If a verb stem starts with a vowel, however, the morpheme -ku- is inserted between the subject marker and the tense marker as seen below.

74a. ku-úumv-a 'to hear'
ku-áang-a 'to refuse'

b. ba-aa-úumv-a-->bakwúumva /ba-aa-[ku]-úumv-a/ 'they would hear' *búumva
ba-aa-áang-a -->bakwáanga /ba-aa-[ku]-áang-a/ 'they would refuse' *báanga

c. ba-aa-uumv-á -->bakwuumvá /ba-aa-[ku]-uumv-á/ 'if they could hear' *buumvá
ba-aa-aang-á -->bakwaangá /ba-aa-[ku]-aang-á/ 'if they could refuse' *baangá

If the morpheme -ku- were not added to these verb stems starting with a vowel, a neutralization would take place and these conditional tenses would be realized exactly as the participial tense and the present conditional, respectively.
The same thing happens in Swahili as well. Monosyllabic verbs, also get back the infinitive prefix in finite forms:

75.a. -la 'eat', -ja 'come', -nywa 'drink', -fa 'die', -cha 'dawn', -pa 'give', -wa 'be'
-enda 'go', -isha 'finish'
b. ninakula 'I am eating'
wamekufa 'they died'
atakwenda 's/he will go'
tumekwisha 'we are finished'

In Swahili the applicative suffix is marked by the vowel i or e depending on whether, the preceding vowel is mid or non-mid. If the verb stem ends with a vowel however, a liquid is added. Clearly, it is the same consonant of the applicative which comes back.

76a. soma 'read' >somea
andika 'write'>andikia
b. nunua 'buy' >nunulia
tia >tilia
pokea >pokelea
zaa >zalia

Second, it is the least marked one which is selected. Polysyllabic verb stems with a mutated final consonant always have the applied suffix added before the perfective aspect marker to avoid neutralization with simple verbs with a perfective aspect; but those which end with z, however, basic or mutated, don't change. This constraint which is imposed by the number of syllables and the length of the stem vowel is due to the fact that in some cases, the consonant is thought of having been resulted from the applicative suffix addition. The consonant mutation due to the perfective aspect and its effect to the applicative morpheme is found in many Bantu languages and has been discussed by other linguists such as Bazin (1982) in what he calls umbrication. The latest discussion of this phenomenon is by Chikane, K. and V, Manfredi and in Setswana

IV. Multiple applicative suffixation

In this section, we will limit ourselves to the combination of arguments marked by the applicative suffix -ir-. It is indeed possible to find these applicative marked arguments combined with other cases such as instrumentals, comitatives, and so on. These more complicated sentences are treated elsewhere.

Like English, Kinyarwanda also allows possessor ascension as seen below.

77a. The child is reading the book of the teacher. >
b. The child is reading the teacher's book.
78a. Umwáana a-ra-som-a igitabo cy'úumwáalimú. >
child s/he-pres-read-asp book of teacher
b. Umwáana a-ra-som-er-a umwáalimú igitabo.

English also allows multiple possession ascension as shown in (79).

79a. John saw Peter's neighor's friend's wife's children's teacher's car key.
b. John saw the key of the car of the teacher of the children of the wife of the friend of
the neighbor of Peter.
Does Kinyarwanda do the same as attempted in 80?

80a. *Yohaáni y-a-bon-er-er-er-er-er-er-e-ye Peetero umutuúranyi inshuti umugoré
abáana umwáarimú imódoká urufuunguuzo.
John he-pst-see-appl-appl-appl-appl-appl-appl-appl-asp Peter neighbor friend
woman children teacher car key
b. Yohaáni y-a-boon-ye urufuunguuzo rw'íimódoká y'úumwáarimú w'áabáana b'úmugoré w'ínshutí y'úmutúuranyi wa Peetero.
John he-pst-see-asp key of car of teacher of children of woman
of friend of neighbor of Peter

If not what are the reasons? This section will be concerned with this issue.
It is indeed possible in Kinyarwanda also to have multiple suffixation but there are certain constraints which limit the number of arguments the verb can have. The difference between English and Kinyarwanda as far as multiple possessor ascension is concerned is due also to the fact that cases are marked on the verb in the latter but on the noun in English.

We noted earlier that the subjective case is not an argument but rather a voice. For this matter it can be combined with any argument marked by the applicative morpheme therefore creating double suffixation.

81(i). Umugoré a-ra-som-a igitabo.
woman she-pres-read-l-asp book.
'The woman is a reading a book for the man'.
(ii).Umugóre a-r-íi-som-er-a igitabo. (Subjective)
woman she-rpes-read-appl-asp book
'The woman is a reading a book '.
(iii).Umugoré a-ra-som-er-a igitabo muu nzu. (event-localizer)
woman she-pres-read-appl-asp book in house
'The woman is reading the book in the house'.
(iv). Umugoré a-r-íi-som-er-er-a igitabo mu nzu . (subjective+event localizer)

81iv, subjective+event localizer can be combined with any other argument marked by the applicative suffix, namely dative, benefactive, possessive and goal.

82a. Umugoré a-ra-som-er-a abáana igitabo. (dative/benefactive)
woman she-pres-read-appl-asp children book
'The woman is reading the book to/for the children'
b. Umugoré a-r-íi-som-er-er-er-a abáana igitabo muu nzu.
woman she-pres-refl-read-appl-appl-appl-asp children book in house
'The woman is reading the book to/for the children in the house'. (subjective+event localizer+dative/benefactive)
83a. Umugoré a-ra-som-a igitabo cy'úmugabo. (possessive)
woman he-pres-read-asp book of man
'The woman is reading the book of the man'.
b. Umugoré a-ra-som-er-a umugabo igitabo. (possessive)
woman she-pres-read-appl-asp man book
'The woman is reading the man's book'.
c. Umugoré a-r-íi-som-er-er-er-a umugabo igitabo muu nzu.
woman she-pres-relf-read-appl-appl-appl-asp man book in house
'The woman is reading the man's book in the house'.
(subjective+event localizer+possessive)
84a. Umugoré a-ra-som-a igitabo ku mafaaranga. (goal)
woman she-pres-read-asp book on money
'The woman is reading the book for money'.
b. Umugoré a--ra-som-er-a igitabo amafaraanga. (goal)
woman she-pres-read-appl-asp book money
'The woman is reading the book for money'.
c. Umugoré a-r-íi-som-er-er-er-a igitabo amafaaranga muu nzu.
woman she-pres-refl-read-appl-appl-appl-asp book money in house
'The woman is reading the book in the house for money'.
(subjective+event localizer+goal)
85a. Umugoré a-ken-e-ye iki? (Accusative)
woman she-need-appl-asp what
'What does the woman need'?
b. Ibyo umugoré a-bi-ken--e-ye ku ki? (accusative+goal)
that woman she-it-need-appl-asp on what
'Why does the woman need that'?
c. Ibyo umugoré a-bi-ken-er-e-ye iki? (accusative+goal)
that woman she-it-need-appl-appl-asp what
'Why does the woman need that'?
d. Ibyo umugoré a-bi-ken-er-er-e-ye ikí muu nzu? (accusative+goal+event localizer)
that woman she-it-need-appl-appl-appl-asp in house
'Why does the woman need that in the house'?

It should be pointed out that since both time (temporal) and space (locative) can act as event localizers, if both of them appear in the same clause, only one applicative marker can be suffixed to the verb:
86a. Ni i Kigalí umugoré a-záa-du-hámagar -ir-a. 'It is in Kigali that the woman will call us'.
be in Kigali woman she-fut-us-call-appl-asp
b. Ni ku cyuumwéeru umugoré a-záa-du-hámagar-ir-a. 'It is on Sunday that the woman will call us'.
be on Sunday woman she-fut-return-appl-asp
c.*Ni ku cyuumwéeru umugoré a-záa-du-hámagar-ir-ir-a i Kigali.
be on Sunday woman she-fut-us-call-appl-appl-asp in Kigali
'It is on Sunday that the woman will call us from Kigali'.
d. Ni ku cyuumwéeru umugoré a-záa-du-hámagar-ir-a.
be on Sunday woman she-fut-us-call-appl-asp
'It is on Sunday that the woman will call us'.
e. Ni ku cyuumwéeru umugoré a-záa-du-hámagar-ir-a i Kigalí.
'It is on Sunday that the woman will call us from Kigali.

86e is ambiguous. It is not clear which is one is the ground: the locative or the temporal.
Before we examine more combinations which create more complex structures, it is important to look at the syntactic properties of the complements of the applicative extension that other objects to the verb undergo such as passivization, relativization, pronoun-incorporation, topicalization as done in Kimenyi (1980). We will limit ourselves to passivization, pronoun-incorporation and relativization. We will test 82b, 83c and 84c, subjectivized and event localized dative, benefactive, possessive and goal, respectively.

87i.. Umugoré a-r-íi-som-er-er-er-a abáana igitabo muu nzu. (dative/benefactive)
'The woman is reading the book to/for the chidren in the house
iia. Abáana ba-r-íi-som-er-er-er-w-a igitabo muu nzu n'úmugoré.
'The children are being read the book to/for in the house by the woman'
b. Igitabo ki-r-íi-som-er-er-er-w-a abáana muu nzu n'úmugoré.
'The book is being read to/for the children in the house by the woman'.
ci.Muu nzu ha-r-íi-som-er-er-er-w-a abáana igitabo n'úmugoré.
'The house is being read in the book to/for the children by the woman'.
ii.Inzu i-r-íi-som-er-er-er-w-a mó abáana igitabo n'úmugoré
iiia. Umugoré a-ra-b-íi-som-er-er-er-a igitabo muu nzu.
'The woman is reading to/for them the book in the house'
b. Umugoré a-ra-cy-íi-som-er-er-er-a abáana muu nzu.
'The woman is reading it to/for the children in the house'.
ci. Umugoré a-ra-h-íi-som-er-er-er-a abáana igitabo.
'The woman is reading there the book to/for the children'
ii. Umugoré a-r-íi-som-er-er-er-a mó abáana igitabo.
'The woman is reading in it the book to/for the children'.
da. Umugoré a-ra-há-ki-b-íi-sóm-er-er-er-a
woman it-pres-there-it-them-refl-read-appl-appl-appl-asp
'The woman is reading it to/for them there'.
b. Umugoré a-ra-yí-ki-b-íi-som-er-er-er-a mó.
'The woman is reading it to/for them in it'.
iva. Dore abáana umugoré y-ii-sóm-er-er-er-a igitabo muu nzu.
'These are the children to/for whom the woman is reading the book in the house'.
b. Dore igitabo umugoré y-ii-sóm-er-er-er-a abáana muu nzu.
'This is the book that the woman is reading to/for the children in the house'
ci. Dore muu nzu umugoré y-ii-sóm-er-er-er-a abáana igitabo muu nzu.
'This is the house in which the woman is reading the book to/for the children'.
ii.Dore inzu umugoré y-ii-sóm-er-er-er-a mó abáana igitabo.
'This is the house in which the woman is reading the book to the children'.
88ia. Umugoré a-r-íi-som-er-er-a igitabo cy'úmugabo muu nzu.
woman she-pres-refl-read-appl-appl-asp book of man in house
'The woman is reading the book of the man in the house'.
b. Umugoré a-r-íi-som-er-er-er-a umugabo igitabo muu nzu.
'The woman is reading the man's book in the house'.
iia. Umugabo a-r-íi-som-er-er-er-w-a igitabo muu nzu n'úmugoré.
man he-pres-refl-read-appl-appl-appl-pass-asp book in house by woman
?'The man is read his book in the house by the woman'
b.* Igitabo ki-r-íi-som-er-er-er-w-a umugabo muu nzu n'úmugoré.
'The book is read of the man in the house by the woman'
ci. Muu nzu ha-r-íi-som-er-er-er-w-a umugabo igitabo n'úmugoré.
'In the house is being read to/for the man the book by the woman'
*'In the house is being the man's book by the woman'
ii. Inzu i-r-íi-som-er-er-er-w-a mó umugabo igitabo n'úmugoré.
'The house is being read in the book to/for the man by the woman'
'*The house is being read in the man's book in the house'.
iiia. Umugoré a-ra-mw-íi-som-er-er-er-a igitabo muu nzu.
'The woman is reading his book in the house'.
b. Umugoré a-ra-cy-íi-som-er-er-er-a umugabo muu nzu.
'The woman is reading it to/for the man in the house'
*'The woman is reading the man's in the house'
ci. Umugoré a-ra-h-íi-som-er-er-er-a umugabo igitabo.
'The woman is reading the book to/for the man there'
*'The woman is reading the man's book there'.
ii. Umugoré a-ra-y-íi-som-er-er-er-a mó umugabo igitabo.
'The woman is reading in it the book to/for the man'.
*'The woman is reading in it the man's book'
di. *Umugoré a-ra-há-ki-mw-íi-som-er-er-er-a.
'The woman is reading that of him there'.
ii.*Umugoré a-ra-yí-ki-mw-íi-som-er-er-er-a mó
'The woman is reading that of his in it'

88di and dii would be correct ofcourse if the pronoun -mu- had a benefactive interpretation.

89i. Umugoré a-r-íi-som-er-er-er-a igitabo amafaraanga muu nzu.
'The woman is reading the book for money in the house'.
iia. Igitabo ki-r-íi-som-er-er-er-w-a amafaraanga muu nzu n'úmugoré.
'The book is being read for money in the house by the woman'.
b.? Amafaraanga a-r-íi-som-er-er-er-w-a igitabo muu nzu n'úmugoré.
'The money is being read for the book by the woman in the house'
ci. Muu nzu ha-r-íi-som-er-er-er-w-a igitabo amafaraanga n'úmugoré.
'In the house is being read the book for money by the woman'.
ii. Inzu i-r-íi-som-er-er-er-w-a mó igitabo amafaraanga n'úmugoré.
'The house is being read in the book for money by the woman'.
iiia. Umugoré a-ra-cy-íi-sóm-er-er-er-a amafaraanga muu nzu.
'The woman is reading it for money in the house'.
b.? Umugoré a-ra-y-íi-sóm-er-er-er-a igitabo muu nzu.
'The woman is reading the book for it in the house'.
ci. Umugoré a-ra-h-íi-sóm-er-er-er-a igitabo amafaraanga.
'The woman is reading there the book for money'.
ii.Umugoré a-ra-y-íi-sóm-er-er-er-a mó igitabo amafaraanga.
'The woman is reading in it the book for money.

As examples show, the possessive and the goal behave differently than other arguments which licence the applicative suffix. When the possessive is marked by the applicative suffix, other arguments lose their syntactic properties. In the case of the goal, it is the only one which loses these properties. Before an explanation is provided, let us look at more combinations.
It is possible to combine the regular cases such as dative+benefactive; dative+possessive; benefactive+goal, when these are combined with the subjective and the event localizer, more complex structures are created.

90a. Umugoré a-ra-som-er-a abáana igitabo. (dative)
woman she-pres-read-appl-asp children book
'The woman is reading the book to the children'.
b. Umugoré a-ra-som-er-a umugabo. (benefactive)
woman she-pres-read-appl-asp man
'The woman is reading for the man'.
c. Umugoré a-ra-som-er-er-a umugabo abáana igitabo. (dative+benefactive)
woman she-pres-read-appl-appl-asp man children book
'The man is reading the book to the children for the man'
d. Umugoré a-r-íi-som-er-er-er-er--a umugabo abáana igitabo muu nzu.
woman she-pres-read-appl-appl-appl-appl-appl-asp man children book in house
'The woman is reading the book to the children for the man in the house'.
(subjective+event localizer+benefactive+dative)
91a. Umugoré a-ra-som-er-a igitabo amafaraanga. (goal)
woman she-pres-read-appl-asp book money
'The woman is reading the book for money'
b. Umugoré a-ra-som-er-er-a abáana igitabo amafaraanga. (dative+goal)
woman she-pres-read-appl-appl-asp children book money
'The woman is reading the book to the children for money'.
c. Umugoré a-ra-som-er-er-er-a umugabo abáana igitabo amafaraanga. (8)
woman she-pres-read-appl-appl-appl-asp man children book money
'The woman is reading the book to the children for the man for money'.
d. Umugoré a-r-íi-som-er-er-er-er--a umugabo abáana igitabo muu nzu amafaraanga.
woman she-pres-refl-read-appl-appl-appl-appl-asp man children book in house on money
'The woman is reading the book to the children for the man in the house for money'.
(subjective+event localizer+benefactive+dative+goal)
92a. Umugoré a-ra-som-a igitabo cy'úumwáalimú.
woman she-pres-read-asp book of teacher
'The woman is reading the book of the teacher'.
b. Umugoré a-ra-som-er-a umwáalimú igitabo. (possessive)
woman she-pres-read-appl-asp teacher book
'The woman is reading the teacher's book'.
c. Umugoré a-ra-som-er-er-a umwáalimú igitabo abáana. (possessive+dative)
woman she-pres-read-appl-appl-asp teacher book children
'The woman is reading the teacher's book to the children'.
d.? Umugoré a-ra-som-er-er-er-a umugabo umwáalimú igitabo abáana.
woman she-pres-read-appl-appl-appl-asp man teacher book children
'The woman is reading the teacher's book to the children for the man'.
(benefactive+possessive+dative)
e.? Umugoré a-ra-som-er-er-er-er-a umugabo umwáalimú igitabo abáana amafaraanga.
woman she-pres-read-appl-appl-appl-appl-asp man teacher book children money
'The woman is reading the teacher's book to the children for the woman for money'.
(possessive+dative+benefactive+goal)
f. Umugoré a-ra-som-er-er-er-a umwáalimú igitabo abáana amafaraanga.
woman she-pres-read-appl-appl-appl-appl-asp teacher book children money
'The woman is reading the man's book to the children for money'.
(possessive+dative+goal)
g. Umugoré a-r-íi-som-er-er-er-er-er-a umwáalimú igitabo abáana amafaraanga muu nzu.
woman she-pres-refl-read-appl-appl-appl-appl-appl-asp teacher book children money in
house
'The woman is reading the teacher's book to the children in the house for money'.
(subjective+event localizer+possessive+dative+goal)
h.*Umugoré a-r-íi-som-er-er-er-er-er-er-a umwáalimú igitabo umugabo abáana amafaraanga mu nzu
woman she-pres-refl-read-appl-appl-appl-appl-appl-appl-asp teacher book man children money in house
'The woman is reading the teacher's book to the children for the man for money in the house'.
(subjective+event localizer+possessive+benefactive+dative+goal)

In the examples above, combinations become questionable only when possessives and benefactives are combined or goals with possessives or benefactives also. These findings are again consistent with our observations earlier about the behavior of possessives and goals. 92d and e would be okay without possessor promotion:

93(92d)a. Umugoré a-ra-som-er-er-er-a umugabo abáana amafaraanga igitabo cy'úumwáalimú.
woman she-pres-read-appl-appl-appl-asp man children money book of teacher
(92e)b. Umugoré a-ra-som-er-er-er-a umwáalimú igitabo abáana ku mafaraanga.
woman she-pres-read-appl-appl-appl-asp teacher book children on money

Note again that except possessives and goals, even complements of more complex "applicative architectures", acquire all the syntactic properties of NP objects as indicated in 94.

94ia. Umugoré a-r-íi-som-er-er-er-er-a umugabo abáana igitabo muu nzu.
'The woman is reading the book to the children for the man in the house'
b. Umugoré a-r-íi-som-er-er-er-er-a mó inzu umugoré abáana igitabo.
c. Umugoré a-r-íi-som-er-er-er-er-a inzu mó umugabo abáana igitabo.
'The woman is reading the book in the house to the children for the man'.
ii. Umugoré a-ra-yí-ki-bá-mw-íi-sóm-er-er-er-er-a mó.
'The woman is reading it to the them for him in it'.
iia. Umugabo a-r-íi-som-er-er-er-er-er-w-a abáana igitabo muu nzu n'umugoré
'The man is being read-for the book to the children in the house by the woman'.
b. Umugabo a-r-íi-som-er-er-er-er-w-a mó inzu abáana igitabo n'úmugoré.
'The man is being read for in the house the book to the children by the woman'.
c. Abáana ba-r-íi-som-er-er-er-er-w-a umugoré igitabo muu nzu n'úmugoré.
'The children are being read the book for the man in the house by the woman'.
d. Igitabo ki-r-íi-som-er-er-er-er-w-a umugabo abáana muu nzu n'úmugoré.
'The book is being read to the children for the man in the house by the woman'.
iva. Dore umugabo umugoré y-ii-sóm-er-er-er-er-a abáana igitabo muu nzu.
'Here the man for whom the woman is reading the book to the children in the house'.
b. Dore abáana umugoré y-ii-sóm-er-er-er-er-a umugabo igitabo muu nzu.
'Here are the children to whom the woman is reading the book for the man in the house'
c. Dore igitabo umugoré y-ii-sóm-er-er-er-er-a umugabo abáana muu nzu.
'Here is the book that the woman is reading to the children for the man in the house'
d. Dore inzu umugoré y-ii-sóm-er-er-er-er-a mó umugabo abáana igitabo.
'Here is the house in which the woman is reading the book to the children for the man'.

Why do possessives and goals behave differently? It is due to the constraint on the number of arguments that the verb can have. A verb with an extension can have a maximum of four objects as we will see later, but a simple verb always has one except ditransitive verbs which have two as seen below. What is interesting about these ditransitive verbs is the fact that the accusative can be dropped but the dative cannot.

95a. Umugoré a-ra-há abáana ibiryó. 'The woman is giving food to the children.
woman she-pres-give children food
b. Umugoré a-ra-há abáana. 'The woman is giving to the children'.
c. *Umugoré a-ra-há ibiryó. 'The woman is giving food'.
d.*Umugoré a-ra-há. 'The woman is giving'.
96a. Umugabo a-r-éerek-a abáana amashusho. 'The man is showing pictures to the children'.
man he-pres-show-asp children pictures
b. Umugabo a-r-éerek-a abáana. 'The man is showing to the children'.
c. *Umugabo a-r-éerek-a amashusho. 'The man is showing the pictures'.
d. *Umugabo a-r-éerek-a. 'The man is showing'.

It is interesting to note that the complement of the applicative suffix behaves exactly like the dative of ditransitive verbs. Other verb complements may be deleted but the complement of the extension has always to be expressed.

97a. Umugoré a-ra-som-er-a abáana igitabo.
woman she-pres-read-appl-appl-asp
'The woman is reading the book to/for children'.
b. Umugoré a-ra-som-er-a abáana.
'The woman is reading to/for the children'.
c.*Umugoré a-ra-som-er-a igitabo.
'The woman is reading the book for/to'.
d.*Umugoré a-ra-som-er-a.
'The woman is reading to/for'.

Kinyarwanda has only four structural case assigning suffixes namely, the comitative/reciprocal morpheme -an-, -the causative -iish-/-y- and the applicative -ir- and the locative.
-an and -iish-/-y- may be used without their arguments being overtly expressed.
None of these extensions, however, the applicative included, can be assigned two arguments at the same time.
As said earlier, a verb with an applicative extension can have a maximum of four complements only, namely locations, patients, recipients and benefactives but six if combined with other extensions(9). Because of this, a collapsing or neutralization of arguments takes place. The goal collapses with the theme and the possessive collapses with the benefactive. This is better illustrated by pronoun-incorporation which creates four pronouns only whereas overt NP objects may be more but without any grammatical relation to the verb any more.
Like French, Kinyarwanda has a word order constraint when there are multiple incorporated pronouns prefixed to the verb. This word order hierarchy is dictated either by case and by person(10). If all the incorporated pronouns belong to the third person, the worder order is as follows: patients, recipients and benefactives. Except the locative which has a special status. It can float occupying any prefix position.
This word order hierarchy is given in 98 and exemplified in 99.

98. accusative < dative < benefactive
Later on, the labels accusative and dative will be used not to refer to arguments but to inherent cases instead.

99a. Umugoré a-r-éerek-er-a abagabo abakoóbwa abáana muu nzu. (11)
woman she-pres-show-appl-asp men girls children in house
'The woman is showing the children to the girls for the men in the house'.
b. Umugoré a-ra-há-ba-bá-b-eerek-er-a.
woman she-pres-there-them-them-them-show-appl-asp
locative-patient-dative-benefactive
'The woman is showing them to them for them there'.
100a. Umugoré a-r-éerek-er-a umugabo umukoóbwa umwáana muu nzu .
woman she-pres-show-appl-asp man girl child in house
'The woman is showing the child to the girl for the man'.
b. Umugoré a-ra-há- mu-mú--mw-eerek-er-a.
woman she-pres-there-him-her-him-show-appl-asp
'The woman is showing him to her for him there '.

In the person hierarchy, it is the reflexive morpheme (-ii-) which comes next to verb stem, preceded by the first person singular or plural , the latter preceded by the second person (singular or plural) and then the third person further away from the verb stem:
The person hierarchy as examples show creates ambiguities.
101. 3rd person < 2nd person < 1st person < reflexive

101. Umugoré a-ra-mú-kuu-m-p-éer-a igitabo
woman she-pres-him-you-me-give-appl-asp book
'The woman is giving my book to him for you'
'The woman is giving your book to him for me'
'The woman is giving my book to him for you'
'The woman is giving my book to you for him'
'The woman is giving his book to you for me'
'The woman is giving his book to me for you'
102 Ba-ra-ku-m-p-a. 'They are giving me to you'
they-pres-you-me-give-asp 'They are giving you to me'

103. Ba-ra-ny-íi-h-á 'They are giving themselves to me'
they-pres-me-refl-give-asp 'They are giving me to themselves'
104. Ba-ra-mú-kw-iiy-erek-er-a
they-pres-him-you-refl-show-asp
'They are are showing themselves to him/her for you'
'They are showing him/her to themselves for you
'Theyare showing you to him/her for themselves'

Because of this strict word order of incorporated pronouns and the maximum number which is four, possessives and goals, take the space of benefactive and accusative, respectively. And indeed an ambiguity is also produced because of this neutralization as examples were shown earlier and which are repeated here again for illustration purposes indicate.

105a. Umugoré a-ra-som-er-a abáana igitabo cy'úmugabo.
woman she-pres-read-appl-asp children book of man
'The woman is reading to the children the book of the man'.
b. Umugoré a-ra-som-er-er-a umugabo igitabo abáana.
woman she-pres-read-appl-appl-asp man book children
'The woman is reading the man's book to the children'
'The woman is reading the book to the children for the man'.
c. Umugoré a-ra-mu-som-er-er-a igitabo abáana.
woman she-pres-him-read-appl-appl-asp book children
'The woman is reading the book of his to the children'.
'The woman is reading the book to the children of his'.
'The woman is reading the book to the children for him'

The word order of complement NPs is not as rigid as that of incorporated pronouns. Usually the animate object, dative or benefactive comes closer to the verb. This word order can be broken by pragmatics, however. The argument which carries emphasis or new information always comes last.
When the adjunct or oblique goal NP is promoted to the internal argument structure of a verb which already has a direct object, the goal becomes an accusative and the former accusative, a "dative". This explains probably why the goal eventhough promoted doesn't enjoy its new status since it is still the lowest argument in the hierarchy not only for the place it occupies but also for its failure to have the syntactic properties that other objects have.

106a. Umugoré a-ra-som-a igitabo ku mafaraanga.
woman she-pres-read-asp book on money
'The woman is reading the book for money'
b. Umugoré a-ra-som-er-a igitabo amafaraanga.
woman she-pres-read-appl-asp book money
'The woman is reading the book for money'
c. Umugoré a-ra-gi-som-er-a amafaraanga.
'The woman is reading it for money'
d. *Umugoré a-ra-ya-som-er-a igitabo.
'The woman is reading the book for it'.
e. ?Umugoré a-ra-ya-gi-som-er-a.
'The woman is reading it for it'.
f. ?Amafaraanga a-ra-som-er-w-a igitabo n'úmugoré.
money it-pres-read-appl-pass-asp book by woman
'The book is read for money by the woman'

When the verb doesn't have any other complement, however, the goal object can undergo all the syntactic transformations.

107a. Umugoré a-ra-som-er-a amafaraanga.
'The woman is reading for money'
b. Umugoré a-ra-ya-som-er-a.
'The woman is reading for it'
c. Amafaraanga a-ra-som-er-w-a n'úmugoré.
'The money is read for by the woman'.
d. Dore amafaraanga umugoré a-som-ér-a.
'Here is the money that the woman is reading for'.
e. Amafaraanga ní yó umugoré a-som-ér-a.
'It is the money that the woman reads for'.

All these examples support our analysis that indeed the reason why goals and possessives refuse or don't allow other complements to undergo the major syntactic transformations of regular objects is the collapsing of arguments and their restricted numbers. Possessives and goals are only derived "external structural" arguments as opposed to "internal" arguments and this fact sets them apart from the latter. The term argument is used here to refer only to NPs governed by the verb, thus being in the VP domain. Prepositional phrases are outside the argument structure eventhough they have semantic roles, the verb doesn't have any control on them. This terminology will be changed, however, because the argument structure doesn't seem to play any significant role in syntax. Only grammatical relations and cases do.
These arguments and their respective word order, as shown by the verb morphology , are repeated informally in (108):

108a. Argument Structure: Locative+Accusative+Dative+Benefactive
This argument structure will be later on shown to be the same as the case structure because an argument can stand for different semantic roles or more than one semantic role at the same time. Thus the accusative stands for Instrumental, Manner, Cause/Goal, Patient and Agent, the dative for Recipient, the benefactive for Benefeciary and Possessor, the locative for location and time.

The accusative which is lower on the hierarchy can have more than one semantic role expressed at the same time. Others allow only one case at a time. This hierarchy which is based on personal pronoun and case in Kinyarwanda might be influenced by animacy also as suggested in Benji Wald.
In some Bantu languages such as Swahili, for instance, the object whether it is an accusative, a dative or a benefactive, assigns agreement on the verb if it is animate as the following Swahili examples show.

109a. Juma a-li-wa-ona watoto.
Juma he-pst-them-see children
'Juma saw the children'.
b. Juma a-li-wa-som-e-a watoto vitabu.
Juma he-pst-them-read-appl-asp children books
'Juma read books to the children'.

The prefix -wa- is not an incorporated pronoun like in Kinyarwanda but only an agreement marker with an animate object which is required in Kiswahili. In Kinyarwanda object agreement doesn't exist.
The hierarchy shown in (108) argues also for the different levels of representation of the syntactic analysis. At least three of them namely (i)the semantic level which is responsible for assigning semantic roles or cases, (ii) the syntactic level which maps semantic roles with argument structure
(accusative, dative, benefactive) and (iii) the surface level in which other morphological elements might be added for purely phonetic reasons. There is an asymmetry, obviously, among these different levels. There are more semantic roles than arguments and cases.
Before we conclude, we should look at the combination of both the alienable and the inalienable possessives and see how they behave with respect to this applicative architecture. It is usuallly body parts or innate characteristics such 'somebody's name'; 'somebody's character', which are found with the inalienable possession. The ascension of inalienable possessions behaves exactly like that of English (door handle; car key; diamond ring,...) namely by lack of any marker or like ditransitive verbs in Kinyarwanda: two unmarked objects.

110a. gutéma ishámi ry'ígití. 'To cut the branch of a tree'
to-cut branch of tree
b. gutéma igití ishámi . 'To cut a tree branch'
to-cut tree branch
111a. kwiibagirwa izína ry'úmuuntu. 'To forget the name of somebody'.
to-forget name of person
b. kwiibagirwa umuuntu izína . 'To forget somebody's name'
to-forget person name
112 has two inalienable possessions whereas 113 has three of them.

113a. Umwáana a-men-nye idírishyá ry'úruugi rw'íimódoká.
child he-break-asp window of door of car
'The child has just broken the window of the door of the car'
b. Umwáana a-men-nye imódoká idírishyá ry'úruugi.
child he-break-asp car window of door
'The child has just broken the car door window'.
c. Umwáana a-men-nye imódoká uruugi idírishyá
d. ?Umwáana a-men-e-ye imódoká uruugi idírishyá
g. Umwáana a-men-nye idírishyá ry'úruugi rw'íimódoka y'úmugoré.
child he-break-asp window of door of car of woman
'The child has just broken the window of the door of the car of the woman'
h. Umwáana a-men-e-ye umugoré idírishyá ry'úruugi rw'íimódoká
child he-break-appl-asp woman window of door of car
'The child has just broken the woman's car door window'.
i. Umwáana a-men-e-ye umugoré imódoká uruugi idírishyá.
j. ?Umwáana a-men-er-e-ye umugoré imódoká uruugi idírishyá

114a. Umugoré a-ra-c-á inzáara z'ámano y'íkireenge cy'úumwáana w'úmugabo
woman she-pres-cut-asp nails of toes of foot of child of man
'The woman is cutting the nails of the toes of the foot of the child of the man
b. Umugoré a-ra-c-á umwáana w'úmugabo ikireenge.
woman she-pres-cut-asp child of man foot
'The man is cutting the foot of the child of the man'.
c. Umugoré a-ra-c-á umwáana w'úmugabo amano.
woman she-pres-cut-asp child of man toes
'The woman is cutting the toes of the child of the man'.
d. Umugoré a-ra-c-á umwáana w'úmugabo inzáara.
woman she-pres-clean-asp child of man nails
'The woman is cutting the nails of the the child of the man'.
e. Umugoré a-ra-c-á umwáana w'úmugabo inzáaza z'ámano y'íkireenge.
woman she-pres-cut-asp child of man nails of toes of foot
'The woman is cutting the nails of the toes of the foot of the child of the man'.
f.? Umugoré a-ra-c-á umwáana w'úmugabo ikireenge inzáaza z'ámano.
woman she-pres-clean-asp child of ma n foot mails of toes
g. Umugoré a-ra-c-íir-a umwáana w'úmugabo ikireenge inzáara z'ámano.
woman she-pres-cut-appl-asp child of man foot nails of toes
'The woman is cutting the nails of the toes of the foot of the child of the man'.
i. ?Umugoré a-ra-c-á umwáana w'úmugabo ikireenge amano inzáara.
woman she-pres-clean-asp child of man foot toes nails
'The woman is cutting the foot toes nails of the child of the man'.
j. Umugoré a-ra-c-íir-a umwáana w'úmugabo ikireenge amano inzáara.
woman she-pres-cut-appl-asp child of man foot toes nail
'The woman is cutting the foot toes nails of the child of the man'.
k. ?Umugoré a-ra-c-íir-a umugabo umwáana ikireenge amano inzáara.
woman she-pres-cut-appl-asp man child foot toes nails
'The woman is cutting the man's child's foot toes nails'.
l. Umugoré a-ra-c-íir-ir-a umugabo umwáana ikireenge amano inzáara.
woman she-pres-cut-appl-appl-asp man child foot toes nails
'The woman is cutting the man's child's foot toes nails'.

To respond to the question asked earlier as to whether, Kinyarwanda allows multiple possessive raising like in English, the answer is no. Only many alienable possessors can be raised. But there is a limit because they raise into the accusative slot which allows only three objects. Only one alienable can be raised because it collapses with the benefactive case, the highest argument on the hierarchy. This benefactive slot unlike the accusative allows only one argument.
As was mentioned earlier, although all these multiple applicative suffixes are grammatically correct and are accepted and understood by all Kinyarwanda speakers, they are rare. First because Rwanda was tradionally an oral society, simple sentences are more used. Second the complex structures are avoided because they render communication slow in both sentence generation and sentence processing. Third, grammatical mistakes might be committed because the speaker and the hearer have to remember how many object clitics or full complements the verb has to assign a corresponding number of suffixes. Fourth, eventhough multiple objects don't occur very often in the sentence structure, multiple incorporated pronouns do. For them to do so, it would necessitate intermediate structures in which full NPs occur and later on drop. This again supports generative grammar that grammars should not only generate sentences which occur on the surface. They also generate intermediate structures which exist to feed transformations which are responsible for the output structures.
In a sense then, sentence generation and processing don't differ in any way from the manipulation of other symbols. The human mind can compute only simple symbols.
Complex sentence production and processing is an area where indeed computational linguistics will be playing a very important role. It will be able to generate all possible types of multiple applicatives and their respective semantic interpretation.

The underlying structure of applicatives

This section is concerned with the underlying representations of complements of the applicative suffix and at which level of representation they are assigned.
So far we have discussed the descriptive adequacy of Generative Grammar, how it can accurately describe the language and create new structures, real and possible. We have so far used intermediate or visible structures but nothing has been said so far about the abstract strucures of applicative architectures, the nature of this applicative suffix and what kind of phrase structure trees they have. There are four competing approaches: a grammatical relation analysis, an incorporation analysis, a case analysis and an abstract verb analysis.
In the traditional relative grammar analysis, two analyses are possible: the presence of the applicative suffix may be seen as either a case of advancement within the same clause or a case of clause-union. In other words is this a case of dative-shift like in English or clause-union like the Romance causative construction? In my own earlier treatment of this phenomenon (Kimenyi, 1980), datives and benefactives were treated like dative shift in English and only causatives were treated as a case of clause-union. Multiple applicative advancements were not considered that time. Within Relational Grammar, the complements of the applicative suffix, is seen as a promotion of oblique objects to the direct object status or a case of objectivization. In the case of clause-union, a demotion process takes place, a subject becoming direct object, a direct object an indirect object , an oblique object or a chômeur as illustrated by the following causatived French sentences in which the former subject becomes a direct object (115), an indirect object (116) and an oblique object (117), as seen below.

115a. Les enfants lisent.
the children read
'The children are reading'.
b. Paul fait lire les enfants.
Paul makes read the children
'Paul is having the children read'.
116a. Les enfants lisent les livres.
the children read the books
'The chidreln are reading the books'.
b. Paul fait lire les livres aux enfants.
Paul makes read the books to children
'Paul is having the books read to the children by the parents'.
117a. Le garçon lit les livres aux enfants.
the boy reads the books to children
'The boy is reading the books to the children'.
b. Paul fait lire les livres aux enfants par le garçon.
Paul makes read the books to children by the boy
'Paul is making the boy read books to the children'.

In Relational Grammar's views, major transformations involved grammatical relations or affected them. Descriptively, Relational Grammar can still account and predict the sentence structure of applicatives. To get the desired phonetic structure, the preposition of the oblique object is deleted.
The object is advanced next to the verb and the applicative suffix is added to the verb.
There are serious problems with this approach, however. The first one : It is not clear at all that indirect objects exist in Kinyarwanda or for that matter in Bantu languages. How do they differ from other oblique NPs or even direct objects?
What is also the status of these multiple objects? RG doesn't tell how many a language can have. "Multiple direct objects " is already one of the major problems. If indeed, grammatical relations are primitives, then at least formal properties should be provided. It had already been pointed out earlier that even with the existence of two direct objects an asymmetry existed (Dwyer, 1977; Kimenyi, 1980; Bresnan and Moshi, 1984). They don't share the same structural properties namely, the place within the VP structure as either full NPs or clitics to the verb. The object of the extension comes closer to the verb. For instance, the initial direct object can be deleted but the dative or object of the extension cannot. These notions don't therefore seem to help at all. Even, in this essay, it was shown that benefactives are different from datives and datives from accusatives. Finally, Relational Grammar doesn't explain the status of the applicative suffix and where it comes from. Is the suffix a verb or a case marker? There is no answer to this question. I show later, however, that Relational Grammar is still better suited to deal with this phenomenon than any other existing theory if it is modified.
The other analysis is the incorporation analysis proposed by Baker (1988). In this analysis, the preposition of the oblique object is incorporated into the verb and the former oblique object becomes a direct object, an analysis in a sense similar to the relational grammar analysis. This is analysis is supported by the promotion of locatives and manners in Kinyarwanda and as well as the incorporation of the associative morpheme na 'also'. There is indeed some kind of cognate relationship between the former preposition and the new clitic to the verb.
It was shown earlier that the locative prepositions ku 'on', mu 'in' , and i 'at' become hó, mó, and yó respectively when the prepositional phrase is promoted to the direct object status and that the new clitic is either attached as a post suffix to the verb or to the head noun. A cognate relationship also exists between the underlying preposition and the clitic.
More examples are again provided for illustration purposes.
118a. Umugabo a-ra-shyír-a igitabo ku méezá.
man he-pres-put-asp book on table
'The man is putting the book on the table'.
b. Umugabo a-ra-shyír-a améezá hó igitabo.
man he-pres-put-asp table on book
'The man is putting the book on the table'.
c. Umugabo a-ra-shyíra hó améezá igitabo.
man he-pres-put-asp on table book
'The man is putting the book on the table'.

The preposition ku is incorporated into the verb and becomes the clitiic hó. The former locaticwe, an oblique object now becomes part of the argument structure of the verb.
Another evidence for incorporation comes from objectivized manner oblique objects.
Oblique manners are marked by the preposition na which is similar the coordination morpheme, the reciprocal and the comitative. When the manner is 'objectivized', the same morpheme is suffixed to the verb stem as seen in the following examples.

119a. Umwáana a-ra-vúg-a n'íshavu.
child s/he-talk-asp with sorrow
'The child is talking with sorrow'.
b. Umwáana a-ra-vug-an-a ishavú.
child s/he-talk-with-asp sorrow
'The child is talking with sorrow'.

It is again clear that the preposition and the suffix are related.
Finally, another clear case of incorporation is illustrated by the cliticization of the morphme ná. This morpheme is used in complementary distribution with kaándi. Both act as coordinators and also mean 'also'. The morpheme kaándi is used to coordinate verbs and adjectives whereas ná coordinates nouns and infinitives as examples in 120 and 121 show.

120a. Umwáana a-fit-e ikáramú n'úrupaparo.
child s/he-have-asp pen and paper
'The child has a pen and a paper'.
b. Umwáana a-zi gusoma nó kwaandika.
child s/he-know to-read and to-write
'The child knows how to read and to write'.
121a. Umwáana mugufí kaándi mutó.
child short and small
'A small young child'.
b. Umugoré y-ará-andits-e kaandí a-ra-hámagar-a.
woman she-pst-write-asp and she-t-call-asp
'The woman wrote and called'.

When there are two conjoined finite VPs like in 121b, the coordinator can be incorporated into the verb but in this case it is na which is chosen. It follows all the tense markers but precedes object-pronouns.

122. Umugoré y-ará-andits-e a-ra-ná-hamagar-a.
woman she-pst-write-asp she-t-also-call-asp
'The woman wrote and called'.

The problem with the incorporation analysis of applicatives is (i) lack of phonetic similarity between the oblique object and the verbal suffix and (ii) how multiple oblique objects can be incorporated. One of the requirement for incorporation is adjency. How would incorporation be possible if indeed there are other direct objects between the verb and the oblique object to be incorporated?

The third analysis which we will pursue here is the complex verb phrase analysis: This analysis entails the existence of abstract verbs in the verb phrase structure. This is the analysis which has implicitly been proposed by many linguists working with Bantu languages such as Givón, Wald who recognize that all these complements are arguments of the extension eventthough they don't offer a formal account of this analysis. The same type of analysis is being proposed for Universal Grammar in both PP and the Minimalist Program for languages which have double objects.
In these languages, then, the underlying structure would have as many verbs as there are complements of the applied suffix.
The complex verb analysis finds support in diachronic facts of Bantu languages. Verb extensions and the tense-aspect-modality system are indeed supposed to come from full verbs.
It is worth noting for instance that the applicative suffix seems to be a doublet of the perfective aspect morpheme -ye which is realized as -ile(-ir-e) in other Bantu languages. The loss of the liquid intervocalically especially in the penultimate or antepenultimate position is common in these languages as is also evidenced by Kinyarwanda data in which this r is deleted when it precedes this aspect morpheme. All these observations suggest that all different uses of the -ir- suffix may have entered the language at different stages of its history and also synchronically at different levels of the sentence derivation.
The locative prepositions (ku and mu ) of underlying locative and cause/goal cases look exactly like other class markers. It is also important to notice that when these prepositions appear before a demonstrative or pronoun, the syllable -rí which looks exactly the same as the copular the auxiliary -rí is added. The proper analysis of this phenomenon would be deletion rather than addition. That is -rí of prepositional locatives is deleted in front of full NPs and retained elsewhere. Otherwise it would be impossible to explain why -rí but not another syllable is added).

123. muu nzu 'in the house' but murí iyo nzu 'in that house'
ku méezá 'on the table' but kurí ayo méeza 'on that table'
ku báana 'on the children' but kurí bo 'on them'
mu bitabo 'in the books' but murí byó 'in/among them'

If indeed, this suffix -ri is the same as the copular -rí, it supports the analysis that event localizers and goals belong to separate clauses which when raising takes place, making them arguments of the higher verb, deletes the auxiliary verb and adds the applicative suffix.
This abstract applicative verb might be indeed the same as the copular -ri. Verb phrases with event localizers are exactly paraphrases with phrases with the copular -ri in the participial form as shown by 124.

124a. Umugoré a-ra-hámagar-ir-a abáana i Kigalí.
woman she-pres-call-appl-asp children at Kigali
'The woman is calling the children from Kigali'.
b. Umugoré a-ra-hámagar-a abáana a-rí i Kigalí.
woman she-pres-call-asp children she-be at Kigali
'The woman is calling the children being in Kigali'.

Although these abstract underlying verbs have support in historical linguistics and the theory of grammaticization, the same analysis cannot be proposed for the synchronic syntax of Bantu languages. Kinyarwanda has other extensions as pointed out earlier namely the reciprocal/comitative morpheme, the causative morpheme and the locative.
All of these as mentioned can appear within the same verb. Like the applicative suffix, many of them too can appear on one single verb. How is then the LF structure going to look like if they occur in the same verb? Kinyarwanda also allows multiple auxiliaries! These multiple auxiliaries, multiple applicatives, and other multiple extensions can all at once occur in the same VP. If the grammar that the linguist is trying to come up with has to be similar to the child's grammar, simple, and economical, clearly the phrase structure rules of Kinyarwanda are going to look complex.
Analysing applicatives as verbs would thus be treating them like auxiliaries which is intuitively wrong. Auxiliaries are indeed verbs in this complex verb structure (Kimenyi, 1986). As a matter of fact auxiliaries are indeed regular verbs because they too can be referential. What makes them auxiliaries is their position within the VP structure. They also all carry both the agreement features and tense.

Relational Grammar is the only generative theory which describes and explains better the complex applicative architecture. It is simpler, less abstract and easily testable. Its main claims is that only grammatical relations (subjects, direct objects and indirect objects) undergo major syntactic properties. Nouns in relation to the head verb of the verb phrase are classified as subjects, direct objects, indirect objects, oblique objects and chômeurs. The argument strucuture or theta-roles or thematic roles don't seem to play any major role in this theoretical framework.
It has already been pointed out that the same theta-role can appear in different grammatical relations.
A theme can appear as a subject or direct object; an instrument can appear as subject, direct object or an adjunct, etc.
A grammatical relation can stand for two theta-roles at the same time. This has been pointed out by Alsena (1992) and Hale and Keyser (1995). In the example (125), for example, the noun abanyurúru is both agent and theme and in (126) umugoré is both possessor and experiencer.
125. Abapóolisí ba-ra-sunik-iish-a abanyurúru imódoká.
policemen they-pres-push-caus-asp prisoners car
'The policemen are making the prisoners push the car'
'The policemen are pushing the prisoners with a car'
126. Umugabo a-ra-som-er-a umugoré igitabo.
'The man is reading the woman's book'
Thus abanyurúru is the causee but also the patient of the extension -iish-. For more arguments see (Alsena, 1992). Similarly, in (126) umugoré is possessor of the book but it is also an 'argument' to the verb. The action affects the possessor making her beneficiary or maleficiary if for instance, she resents the fact the man is reading her book.
These theta-roles or semantic roles as the applicative suffix has shown happen also to be marked by the same suffix or preposition. For instance goals, locatives and temporals share the same preposition ku. Manners, instrumentals, comitatives and agents also share the same preposition na. The polyfunctional or polysemantic nature of these morphemes is either due to polysemy or homonymy. Polysemy implies a tropical relationship among all the referents, that is association or similarity. When there is a similarity among the referents, this relationship is metaphoric but if there is an association then this relation is metonymic. The similarity may be accidental, however, due to either historical factors both internal and external or to the finite nature of linguistic signs and structures which causes them to refer to unrelated concepts. Thus extra-linguistic factors are necessary to get the correct semantic interpretation.
Kinyarwanda doesn't have intransitive verbs, either. Transitive verbs can be detransitivized, however. The following verbs, for instance, which in other languages are considered intransitives in Kinyarwanda they have direct objects and these direct objects manifest all the behavioral properties of other direct objects such as being able to passivize, to undergo pronoun-incorporation, relativization, extraposition, etc.

127a.Umwáana a-ra-rwáa-ye.
child s/he-pres-be sick-asp
'The child is sick'
b.Umwáana a-rwaa-ye ibicuráane
child s/he-be sick-asp cold
'The child is sick of cold'
c. Ibicuráane bi-rwaa-w-e n'úumwáana.
cold it-be sick-pass-asp by child
*'The cold is being sick by the child'.
d.Umwáana a-ra-bí-rwaa-ye.
child s/he-pres-it-be sick-asp
'The child is sick of it'.
128a.Uríiya mugabo y-a-báa-ye perezida.
that man he-pst-be president
'That man was the president'.
b. Perezida y-a--báa-w-e n'uúriiya mugabo.
president he-pst-be-pass-asp by that man
*'The president was been by that man'.
c. Uríiya mugabo y-ará-mu-báa-ye.
that man he-pst-him-be
'That man was him'
d. Perezida, uríiya mugabo y-ará-mu-báa-ye.
president that man he-pst-him-be-asp
'The president, that man was him'.

These grammatical relations and non-terms can change their grammatical status at different levels or strata during different cycles of the derivations which promote or demote NPs within the verb phrase structure. Relational Grammar obviously has to be slightly modified to account nicely for Kinyarwanda multiple objects. The grammatical relations needed for Kinyarwanda are only subjects, direct objects and adjuncts. Indirect object has to be removed because it doesn't seem to exist or to play any role in the grammar of Kinyarwanda and many other languages including English. The direct object is any complement controlled by the main verb. How many direct objects a language can have will depend on that specific language. Adjuncts are complements which are introduced by prepositions such as instrumentals, goals, locatives thus outside the verb domain. To be effective, however, Relational Grammar has to be coupled with a case theory. That is each grammatical relation has to be assigned case. The case theory is motivated by the universality of abstract case (Chomsky, 1995) and case filter (Vernaud, Chomsky, 1995). This case filter analysis has been introduced to account for the morphological marking on some complements. There are two types of case: inherent and structural. Inherent cases are usually not marked and structural cases are. Inherent cases in Kinyarwanda are datives and accusatives. They are shown strictly by the word order. Datives come close to the verb and accusatives follow. These two cases are found with ditransitive verbs. A full discussion on these inherent cases is presented in Kimenyi (in press, Ditransitive verbs). Structural cases on the other hand such as causatives, comitatives, reciprocals and locatives have morphological markers.
For instance if a case is assigned on the adjunct then that adjunct will appear with an appropriate preposition. If the same case is assigned on the direct object, however, the case marker will appear as a suffix on the verb. The only problem with this analysis is that the case marker is on the verb instead of being on the noun, the verb being the case-assigner and the noun the assignee. This problem is solved, however, if Baker's incorporation analysis is taken into consideration. That means that whenever a grammatical relation is assigned case the verb also raises to this complement to insure adjacency conditions. Once the verb, the noun and its case are adjoined, then incorporation takes place.
This analysis is again supported by locatives in Kinyarwanda. When they are promoted to direct object position, case is either marked on the direct object or on the verb. Example (118) presented earlier in support of Baker's incorporation analysis is again repeated for illustration purposes.

130a. Umugabo a-ra-shyír-a igitabo ku méezá.
man he-pres-put-asp book on table
'The man is putting the book on the table'.
b. Umugabo a-ra-shyír-a améezá hó igitabo.
man he-pres-put-asp table on book
'The man is putting the book on the table'.
c. Umugabo a-ra-shyíra hó améezá igitabo.
man he-pres-put-asp on table book
'The man is putting the book on the table'.

The underlying structure of 130a is the one we see in 131a and that of 130b and c is the one shown in 131b.

131a. (V, Su, DO, Adjunctloc)
b. (V, Su, DOacc, DOloc)

These examples clearly show that the case assigner is the verb and the assignee the noun but that the verb can also assign the case to itself through the process of case incorporation. Since these different cases are marked on the verb by the so-called verb extensions, I will use the standard Bantu terminology used to refer to these suffixes when refering to these cases. Kinyarwanda therefore has four structural cases namely the applicative case, marked by the suffix -ir-, the causative case shown by the morpheme -iish-, the comitative case marked by -an- and the locative case marked by hó, mó or yó.
The surface structure of the sentence namely whether complements will appear as direct objects or adjuncts depends solely on the information given in the underlying structure on case assignment. If these cases are assigned on the adjuncts then these complements will appear with appropriate prepositions but if they are assigned on direct objects then promotion will have to take place.
There are, however, structural cases which are internal. These cannot appear as adjuncts. They are in a sense similar to inherent cases such as accusatives and datives in Kinyarwanda. The only difference is that inherent cases are not marked morphologically but internal structural cases are. Thus what was traditionaly referred to as indirect objects or datives and benefactives of monotransitive verbs in Kinyarwanda are obligatorily marked for this applicative case as shown in the following examples as opposed to other complements such as goals or possessives which can have different surface structures depending on how cases are assigned. The example (132) below illustrates internal structural cases and (133) and (134) are examples of external structural cases.

132. Umugabo a-ra-som-er-er-a umugoré abáana igitabo.
man he-pres-read-appl-appl-asp woman chidren book
'The man is reading the book to the children for the man'.
133a. Umugabo a-ra-som-a igitabo cy'úmugoré.
man he-pres-read-asp book of woman
b. Umugabo a-ra-som-er-a umugoré igitabo.
man he-pres-read-appl-asp woman book
'The man is reading the woman's book'.
134a. Umugabo a-ra-híing-ir-a umugoré ku mafaraanga.
man he-pres-cultivate-appl-asp woman for money
b. Umugabo a-ra-híing-ir-ir-a umugoré amafaraanga.
man he-pres-cultivate-appl-appl-asp woman money
'The man is cultivating for the woman for money'.

(132) has only one surface structure, while (133) and (134) which have possessor and goal 'arguments', respectively have two surface structures, adjuncts and direct objects.
Thus (132) has only one underlying structure namely the one given in (135), whereas (133) and (134) have also two underlying structures respectively, shown in (136) and (137).

135. (V, Su, DO appl, DO appl, DOacc)
136a. (V, Su, DOacc, Adjunctposs)
b. (V, Su, DOacc, DOappl)
137a. (V, Su, DOappl, Adjunctgoal)
b. (V, Su, DOappl, DOappl)

138a. Umugabo a-rá-ang-a kó abáana ba-som-á ibitabo.
man he-pres-refuse-asp that children they-read-asp books
b. Umugabo a-rá-ang-ir-a abáana gu-som-a ibitabo.
man he-pres-refuse-appl-asp childen to-read-asp books
139a. Umugoré y-eeme-ye kó abakoóbwa ba-táah-a.
woman she-accept-asp that girls they-go home-asp
b. Umugoré y-eemer-e-ye abakoóbwa gu-taah-a.
woman she-accept-appl-asp girls to-go home-asp

This multiple case which is assigned in successive cycles is not done randomly. It respects the person or argument hierarchy shown earlier namely the reflexive being assigned first and the third person last in the case of person hierarchy and the benefactive first and the accusative last in the argument hierarchy.

The multiple case assignment in Kinyarwanda shows that this language, unlike others, even closer ones, geographically and genetically, is more iconic. There is an isomorphism or one-to-one relationship between the complement and the verb form. Each promoted grammatical relation is assigned a structural case to the verb matrix. This iconicity, unfortunately, creates a very complex structure and is therefore against the principles of the Minimalist Program and language acquisition namely, economy. It is the reason why it is different from Chaga a very close language which like Kinyarwanda also allows multiple complements as direct objects of the same verb but with only one marked applicative case. This difference is explained by the case assignment. Since only one applicative extension is allowed when the verb has multiple objects, it means that only one is assigned the applicative case and others are assigned the accusative case.

Since any verb can have a dative, a benefactive, an event localizer, a subjective and since these have to be marked by the applicative suffix on the verb, they are part of the extended projection principle of the main verb. This means that they are in this domain and therefore have all the object properties characteristic of inherent objects.
Now we also understand why, indeed goals and benefactives collapse with accusatives and benefactives, respectively. The latter are internal arguments or direct objects whereas the former become direct objects only structurally. This is was the intuition behind Relational Grammar. It is clear that what was meant is that if an element becomes part of the head domain then it acquires all the structural properties of other elements which occupy that domain.

Conclusion

This paper has discussed areas that have received very little attention or no attention at all in Bantu linguistics. The discovery of the multiple applicative suffixes was made possible by the metatheory and by comparative grammar. Without generative grammar these types sentences would indeed have gone unnoticed. Credit should be given to Relational Grammar since it was the one which introduced the notion of promotion to object status making it possible for adjuncts to acquire grammatical relations by moving into the head verb domain. It is very clear therefore that good theories don't only describe and explain but also affect grammar. As a result we can predict that Kinyarwanda linguistic performance will not remain the same. The analysis has raised more questions than it has answered. Its main contribution, however, was to confirm the original claims of generative grammar that a grammar should indeed be able to generate "novel" sentences that that have never been used or heard before. The multiple applicative constructions are still a fascinating area for further research not only for comparative Bantu but for cognitive linguistics because prescriptive grammar has not influenced these languages yet. The representation of syntactic rules still has a long way to go, however, before we are certain that they indeed correspond to their respective corresponding mental structures.

Notes:

*This paper was a keynote address at the 8th Niger-Congo Syntax-Semantics Workshop at Boston University in October 1995 organized by Victor Manfredi.

1. The associative construction is marked by the morpheme -a which is always deleted if the head noun starts with a vowel. It takes the class marker of the head noun.
It has different semantic functions such as identification, nationality, color, number, possession,
umugoré w'úmunyáameriká 'an American woman'
umugoré w'úmugabo (a woman of man) 'a courageous woman'
umugoré w'úmugabo 'a man's woman/wife'
umugoré waa njye (woman of me) 'my wife'
umugoré wa kábiri (woman of two) 'second wife'
umugoré w'íinzóbe (woman of swamp antelope) 'a light skin woman'

2. Cause and goal are not indeed the same case eventhough they are marked by the same morphemes the preposition ku and the applicative suffix -ir- when promoted. There is a metonymic relation however between the two, an association of cause and effect, trigger and target, source and destination.
gukóra ikiintu ku gahato, ku bushaáke, >gukórera ikiintu agahato, ubushaáke
'to do something because of being forced, on one's will'
gukóra ikiintu kubéera ubugomé, ishyarí >gukórera ikiintu ubugomé
'to do something because of meanness, jealousy'
Like in the case of locatives and temporals, an asymmetry exists between cause and goal also. The goal having less constraints than the cause in terms of licencing the applicative morpheme and acquiring the grammatical properties of other objects.
3. This is indeed similar to the Latin or Greek middle voice or the French use of reflexive pronouns
s'en aller 'to leave' , s'en foutre 'not to care' , s'en orgueillir 'to be naughty',
4. One verb has been lexicalized with the subjective morphology. This is kwíiyaangirá 'to refuse' from kwáanga. It is the only infinitive verb in the language which a high tone on last syllable. Others have it on the first mora of the verb stem. This lexicalized subjective verb can ofcourse appear in the subjective form: thus doubling both the reflexive morpheme and the applicative suffix:
kwíiyiyaangirira.
4. We have been assuming all along that -iz- which is added to these verb stems is the combination of -ir+y- , but the other alternative analysis that it might be just -iz- has been overlooked.
One of the reasons is that the morpheme -iz- by itself does not indeed exist. Secondly, it was shown that -ir- is added to verb stems which end with -iir- and monosyllabic verb stems with an applicative suffix before the perfective suffix -ye and (b) and before the nominalizer morpheme -yi after verb stems which end with -iir- as seen in . Third we saw that -ir- plus -y- always becomes -iz- by consonant mutation.
Another clear evidence that this -iz- is indeed -ir+y-, is the addition of the applicative morpheme to verb stems which already have the causative morpheme -y-. When -ir- is added to causativized verbs, the -y- follows the applicative suffix, unless the last consonant is alveolar in which case it is doubled, one next to the verb stem and the other one after the applicative suffix.

kuráara 'to spend the night' --- kuráaza /ku-ráar-y-a/ 'to spend the night with'
--- kuráaza+ir ---> kuráariza /ku-ráar-ir-y-a/ 'to spend the night with for'
gukáraba 'to wash hands' --- gukárabya /ku-kárab-y-a/ 'to help wash hands'
--- gukarabya+ir ---> gukárabiriza /ku-kárab-ir-ir-y-a/ 'to help wash hands for'
gutogota 'to boil' --- gutogosa /ku-togot-y-a/ 'to make boil'
--- gutogosa+ir ---> gutogosereza /ku-togot-y-ir-ir-y-a/ 'to make boil for'

All these examples clearly show that -iz- as morpheme does not exist but is the combination of the applicative suffix -ir- and the causative morpheme -y-.

5. There are many reasons as to why it has taken me a long time to discuss about these multiple applicatives: first it was because of other linguistic interests, second I became aware of this phenomenon late, third it was because of lack of informants. Since these types of sentences are so rare or may have never occurred at all, it is indeed very hard to be confident of the accuracy of these structures. Lastly, the theoretical bias played a role also. I thought in the beginning that linguistic theory should be concerned only with those sentences which are found, heard but not with the ones which are possiblle.
6. Although all our informants accept all these sentences, it is impossible to find in any text, written or oral such as folktales, historical narratives, poetry and the spoken language in which three or more applicative suffixes within the same verb appear.
7. In French the following incorporated pronouns:
me/nous 'me/us'
te/vous 'you sing./you pl.'
le/la/les 'it/him'/'it/her'/'them'
lui/leur 'to him/to her'/'to them'
en 'of'/y 'at'
If the incorporated pronoun are all third person, the accusative comes first and the dative comes second:
Je (le/la/les) (lui/leur) montre 'I show (him/her/them) to (him/her/them).
If it the third person and another person, the third person comes next to the verb.
Je te la/le/les montre. 'I show her/him/them to you'.
The only ones which never change their word order are en and y, y preceding en.
Y a-t-il des livres sur la table? Oui, il y en a.
'Are there books on the table. Yes, there are'.
8. exc. gufáta 'to hold' afashe 'he just held'
gutwíita 'to be pregnant' aratwíite 'she is pregnant'
afite 'he has' (but afise in Kirundi)
9.
When this suffix is added to -híig- 'hunt', no mutation takes place either: umuhiígi /u-mu-híig-yi/ 'hunter'
10.
There are some exceptions ofcourse as in the case of the loan-word gukosa (-kosheje) which has been borrowed from swahili and clearly doesn't have -kot- as a stem counterpart.
This supports the contamination thesis, however, since native speakers think that this s has resulted from consonant mutation.
11. Suffix reduplication is found with the frequentative morpheme -agur- also when this one is added to a monosyllabic verb:
gutá 'to lose' ---- gutáagagura 'to lose every where'
gucá 'to cut' --- gucáagagura 'to cut into pieces'
kuryá 'to eat' --- kuryáagagura 'to eat without interruption'
12. These historically motivated rules are found in English as well. For instance English has two nominalizers -er and -or. It seems that -er is added to Anglo-Saxon verb stems and -or to latinate
verb stems:
-er: smoker, drinker, driver, teacher, sleeper, swimmer, etc.
-or: actor, educator, professor, inspector, director, prosecutor, etc.
In English, also, we find a similar phenomenon of phonetically motivated morpheme insertion. For instance the morpheme -at- which is found in a lot of latinate verbs, is added to verbs stems which don't have it when the suffix -ion is added.

a. transport >*transportate > transportation
confront >*confrontate > confrontation
resign >*resignate > resignation
proclaim > *proclamate > proclamation
explain > *explanate > explanation
reveal > *revelate > revelation
provoke > provocate > provocation
publish > *publicate > publication
b. maximize > maximization
synchronize > synchronization
causativize > causativization
c. justify > justification
verify > verification
purify > purification

13. If the applicative is combined with the causative, the verb can have six arguments as shown below.
a.?Umugoré a-ra-som-eesh-eesh-er-er--ez-a umugabo umuhuúngu abáana igitabo
woman she-pres-read-caus-caus-appl-appl-asp man boy children book
imáshiíni muu nzu.
machine in house
'The woman is making the the boy read the book to the children for the boys in the
house'.
b. A-ra-ha-yi-ki-mu-ba-mu-som-eesh-eesh-er-ez-a.
she-pres-there-it-it-him-them-him-read-caus-caus-appl-appl-asp
loc inst pat ag dat ben
'She is making him read it with it to them for him in it'.
The accusative case slot can hold three incorporated pronouns at the same time: instrument/manner; patient and agent. Others namely dative and benefactive can hold only one case at a time.
Verbs with multiple object-pronouns are common but those with multiple full object NPs are rare. It is the reason why (a) is not good again because it is very difficult to know the role that each has in the sentence since there is no strict word order.
14. This is discussed at great length in Kimenyi (1990) "Genesis of auxiliaries".

References

Alsina, A. and S. Mchombo. 1990. "The Syntax of Applicatives in Chichewâ: Problems for a Theta Theoretic Asymmetry". Natural Language and Linguistic Theory 8: 493-506.
Baker, M. 1988a. Incorporation. Chicago: Chicago University Press.
Baker, M. 1988b. "Theta Theory and the Syntax of Applicatives in Chichewâ. Natural Language and Linguistic Theory 6, 353-389.
Bastin, Y. 1983. La finale verbale -ide et l'imbrication en en Bantu. Tervuren, Belgique:
Musée royal de l'Afrique Centrale.
Benji, W. (in press). "The animacy hierarchy in Bantu". Los Angeles.
Bresnan, J. and L. Moshi. 1990. "Object asymmetries in comparative Bantu syntax". Linguistic Inquiry 21: 147-83
Chicane, K. and V. Manfredi. "Applicatives in Setswana Fonosyntax". Paper presented at the 8th Workshop on Niger-Congo Syntax and Semantics. Boston University.
Chomsky, N. 1995. The Minimalist Program. Cambridge: MIT Press.
Haiman, J. 1988. Iconicity in Language. Cambridge University Press.
Hale, K.and S.J.Keyser. 1993. "Argument structure". A View from Building 20. eds.
Hale, K. and S.J. Keyser. Cambridge: MIT Press.
Halle, M. and A. Marantz. 1995. "Distributed morphology". A view from Building 20. eds.
Hale, K. and S. Keyser. Cambridge: MIT Press.
Kimenyi, A. 1979. Studies in Kinyarwanda and Bantu Phonology. Edmonton: Alberta. Linguistic Research, Inc.
Kimenyi, A. 1980. A Relational Grammar of Kinyarwanda. Berkeley&Los Angeles: University of California Press.
Kimenyi, A. 1990. "The genesis of Auxiliaries". Paper read at the 21 Annual Conference on African Linguistics. Athens: Georgia
Kimenyi, A. (in prep.). Sight, Sound, Sense, Style and Structure in Kinyarwanda: Evidence of the iconicity of language.
Moshi, L. 1995. "Multiple objects in Chaga".
Peirce, C. 1931-1959. Collected Writings. Cambridge: Harvard University Press.
Shapiro, M. 1980. Language as semeiotic. Bloomington: Indiana University Press.