Kinande Anaphora Sketch

Version 1.3

Ngessimo Mathe Mutaka, University of Yaounde 1 and Ken Safir, Rutgers University

There are some patterns of particular theoretical interest that distinguish the Kinande anaphora system from patterns found in other Bantu languages, but these only emerge in careful study of the details of particular morphemes and the constructions they enter into, since Kinande shares many features typical of the Bantu pattern.

First, some more typical Bantu properties shared by Kinande: The reciprocal marker (RCM) is expressed amongst the suffixal verb extensions and a reflexive marker (RFM) is expressed in prefixal position adjacent to the verb stem. The RFM is in complementary distribution with the object marker (OM) that is in turn in complementary distribution with a full nominal direct object. Unlike the OMs that they are in complementary distribution with, neither the RFM nor the RCM are sensitive to noun class distinctions, though reciprocal readings, of course, require semantically plural antecedents. Where RFM and RCM cannot represent a given argument position, the most typical outcome is the presence of a pronoun in the argumental position.[Note X1] Two emphatic reflexives, one a reduplicated version of the other, can be used in connection with a pronoun in any position, but it more typically modifies the subject adverbially to emphasize an action performed either alone or in person. The emphatic reflexives cannot render a predicate reflexive without the presence of the RFM, unless they are attached to pronouns in positions that the RFM cannot correspond to.

One of the patterns found in Kinande that is of particular interest involves the relation between intransitive reflexive verbs and reflexivized versions of transitive verbs. There are two verb roots in Kinande, -nab- “wash” and -hakab- “smear”, which have intransitive reflexive interpretations as many grooming or body preparation verbs do in other languages. It is not uncommon for such verbs to have transitive forms that can also be reflexive by the usual productive strategy for reflexivization, as in English, where John washed and John washed himself are both possible. In Kinande, however, it is not possible to use the productive form of reflexivization, the addition of the RFM, to form a transitive reflexive reading (similar phenomena are to be found in the AQs for Cinsenga section 2.1.3, (A2) and Ikalanga section 4 (X2a-c), for example). Rather the intransitive stem has to be made transitive, and this is achieved for these two verbs by the addition of what appears to be the causative verb extension -i-. In other words, intransitive -nab- and -hakab- must first be transitivized, permitting a non-reflexive object, as in (A13c”), before they can appear with either the RFM or the RCM.

A13c) tu-kándi-nab-an-i-a

we-TM-wash-RCM-CAUS-fv

“We will wash each other.”

c’) tu-kándi-yi-nab-i-a

we-TM-RFM-wash-CAUS-fv

“We will wash ourselves.”

c”) tu-kándi-nab-i-a Kambale

we-TM-wash-CAUS-fv Kambale

“We will wash Kambale.”

There are at least two larger points of particular interest here.

One issue of interest here is the nature of the transitivizing verb extension -i-. This form appears in the morphological shape and position of the second causative verb extension in Kinande, yet the reading that results from this suffixation for -nab- and -hakab- is not causative, although the suffixes in question more typically induce causative interpretations, or particular forms of causative interpretations. After all, the acts of washing and smearing are volitional and hence agentive even for the intransitive reflexive versions of these predicates. The extended verb does not mean that “X caused Y to be washed” (as in the causative versions of inchoatives in English for verbs such as melt or burn) or that X caused Y to wash (on direct and indirect causation see below). Thus the transitivizing role of the verb extension -i- is not causative in this environment in any conventional sense.

It is useful here to enlarge on how causative readings are usually constructed in Kinande. It has been observed in Bantu and specifically for Kinande (Mutaka & Kavutirwaki 2006) that causativization can come about by two co-occurring verb extensions, -is- and -i-. The first cannot result in a causative reading without the second, but the second can achieve causative interpretation in the absence of the first. A fully compositional analysis of causative seems required when the full two-part affix is present, -is-...-i-, though the role of -i- on its own seems much less consistently causative, at least in the sense of semantically causative, as the role of -i- as a transitivizer illustrates. There are, however, some other distinctions that arise between causative interpretations achieved by the two part extension -is...i- vs. the single extension -i-. In contexts where they are semantically causative, the two-part causative and the single causative contrast with respect to whether they induce a direct causative reading or an indirect causative reading, a possibility first noted by Bastin (1986) for a variety of Bantu languages and recently discussed by Good (2005)[Note X2] as illustrated by the way these different causativizations apply to the intransitive root -titir-, meaning ‘dance frenetically’ (a form of trembling dance, better rendered by the French verb se trémousser)(see AQ 2.3.4).

AS1a) ngátitiraya omwána

n-ka-titir-a-i-a omu-ana

SM-TM-dance-a-CAUS-fv CL1-child

“I make a child dance frenetically.”

b) ngátitirisaya omwána

n-ka-titir-is-a-i-a omu-ana

SM-TM-dance-CAUS-a-CAUS-fv CL1-child

“I have a child to dance frenetically.”

For (AS1a), the first person is holding the child and making him dance by direct physical intervention, but in the second case, causativization can be indirect, insofar as the first person can give the child a gift that will cause the child to dance frenetically.

Returning now to readings more akin to reflexive ones, there are other verbs that are transitivized by the presence of -i-. Consider the following contrast:

A15c.i. Kambale mó-a-ká-hutal-a okó byála

Kambale TM-SM1-TM-hurt-fv on hand

“Kambale got hurt on the hand.”

c.ii. Kambale mó-a-ká-yí-hutal-a-i-a okó byála

Kambale TM-SM1-TM-yi-hurt-(a)-CAUS-fv on hand

“Kambale hurt himself on the hand.”

c.iii. Kambale mó-a-ká-hutal-a-i-a ó-mu-aná y’ okó byála

Kambale TM-SM1-TM-hurt-(a)-CAUS-fv iv-C1-child Lk on hand

“Kambale hurt the child on the hand.”

The use of the causative and the reflexive in (A15cii) adds volition to Kambale’s injury, such that he intentionally hurts himself, whereas (A15ci) only reports an injury (and if -yi- were replaced by an OM, then the reading would be one where Kambale hurt someone else on that person’s hand, as shown in (A15ciii)). In this case, one could argue that the -i- affix is contributing a causative semantics, although the case is somewhat intermediate between transitivization and full causality. In both versions of the verb, the hand in question is Kambale’s, but if (A15cii) is transitive, then we must assume that (A15ci) is intransitive, and the question arises as to whether -hutal- is like -nab- insofar as it is a reflexive intransitive, except for the fact that -hutal- has an additional oblique argument (the body part). If this reasoning is on the right track, then it may be pertinent to reconsider English pairs like John hurt his hand and John hurt himself on the hand, where the introduction of the preposition in English appears to indicate that English intransitive reflexive hurt has been transitivized in much the same way as in Kinande (A15cii).

The issues concerning the causative are complicated, but there are perhaps a few interesting generalizations that can be made about the addition of only -i-. It would appear that when it is present on its own, it makes one or both of two changes to the intransitive verb it attaches to. It insures that there are two distinct role players and/or that the first causes a change in the second by direct action. Where both of these adjustments to the meaning of an intransitive verb are unnecessary, then extension with only -i- is anomalous. This appears consistent with a sample of such relations we supply here.

Some paradigms with causative extension

Eri-hék-a To carry	Eri-hek-i-a omundú y’okó mú-twê person LK on C3-head (in the sense of to help him carry a burden on his head; e.g. when I put the burden on his head	Eri-hek-es-i-a omundú y’ebi-rî person LK C8-potato To cause him to carry potatoes
Erí-húm-a To move in the sense of leaving a place in order to settle somewhere else.	Erí-hum-i-a ebíndu C8-thing To move stuff, to put things in a different place. Here, you are the one doing the moving of the things	Erí-hum-is-i-a ebíndu Here, you do not have to be the person doing the displacement. You cause s.o. to move the things
Eri-seny-a To collect (firewood) e.g. n-gandi-senya munabwire I FUT collect firewood today I will collect firewood today	*eri-seny-i-a	Erí-seny-es-i-a esyóngwi firewood To make s.o. or people to collect firewood
Eri-sóm-a To read	*eri-som-i-a omwána	Eri-som-es-i-a omwána child To cause a child to read
Eri-sóm-a To drink	eri-som-i-a omwána y’óbwabu to make a child drink (in the sense that you take the cup and you pour beer in his mouth	Eri-som-es-i-a omwána y’ óbu-abu Child LK C14-beer To cause a child to drink beer
Eri-hér-a To get lost	Eri-her-i-a To lose something or not to win	Eri-her-es-i-a ekingyó ky’ age Mortar LK me To cause my mortar to get lost
Erí-kohól-a To cough	? ?erí-kohol-i-a	Erí-yi-kohol-es-i-a To cause oneself to cough
? ?Erí-lánga	Eri-lang-i-a to get erected e.g. a-abiri-lang-i-a SM-TM-erect-CAUS-FV He is already excited	Eri-lang-is-i-a To cause one to get erected
Erí-sénda To flow	Eri-send-i-a ekirí ky’ omo magétse Potato LK in water To make the potato flow in the water (you are the one putting the potato in the water to make it flow)	erí-send-es-i-a ekiri to cause the potato to flow (for example when you give the order that s.o. should go and make the potato flow)

The failure of -i- with verbs meaning ‘read’ and ‘carry’ would be a result of the fact that these intransitives involve direct causation by an agent acting on a semantic role distinct from that of the subject, hence -i- does not add anything to their meaning. This could be extended to ‘cough’, where the physical relations are hard to imagine under direct causation, but if one were partially strangling an individual to make him cough, then erí-kohol-i-a omúndu becomes more natural. This hypothesis requires ‘move’ and ‘lose’ in Kinande to be analyzed as inherently reflexive intransitives or stative intransitives (hence -i- contributes to a change in meaning), the former being a plausible analysis for erí-hum-a ‘to move’, especially from the English perspective with respect to move. Verbs like eri-hér-a pattern with those like eri-tamír-a ‘to get drunk, to be tipsy’; eri-tamir-i-a to get tipsy as a result of something that one took, as in óbwabu bu-ámá-nyi-tamir-i-a ‘the drink makes me tipsy’, or omwátsi a-ámá-mu-tamir-i-a the news has rendered him tipsy, that is, it has rendered him like someone who is drunk. However, even this account does not quite explain all the cases, especially ‘drink’, which would not appear to be interpreted as an intransitive reflexive, still involves semantically distinct roles, yet still forms a causative with -i-. This approach to -i- is somewhere between those that treat its Bantu cognates simply as causative or direct causative (as in Bastin, 1986) and those that treat it as strictly a transitivizer (a possibility suggested by Good, 2005).[Note X3]

Another issue of theoretical interest arises because the evidence is clear that transitive ‘wash’ and ‘smear’ are formed in Kinande from intransitive forms, not the other way around. This would appear to contradict the assumption that unproductive, intransitive reflexive wash is generally derived from a transitive reflexive form, as is generally assumed by Reinhart and Siloni (2005). R&S assume English intransitive reflexive wash is derived from transitive reflexive wash by a process of thematic merger, whereby two thematic arguments are assigned to the same position. Their proposal abandons a general restriction on thematic assignments enshrined in the Theta Criterion (Chomsky, 1981) which requires that every thematic argument be assigned to a distinct argument, and thus represents a significant departure from earlier assumptions. However, if intransitive reflexive wash is the form on which the transitive is based, then thematic merger is ill-conceived as a way to relate the intransitive and transitive forms of wash. This does not mean that R&S are necessarily wrong about English, but it opens the possibility that the analysis they propose could be backwards, and a careful investigation of a transitivization analysis for English is warranted, perhaps now extending not only to verbs like move but also to verbs like hurt or cut analyzed as intransitives with oblique body-part complements (‘cut’ does not work like ‘hurt’ in Kinande, but cut does work like hurt in English).[Note X4]

Kinande is also a language in which perception verbs behave like ‘exceptional casemarking (ECM)’ or ‘raising to object’ verbs, which is to say that they are verbs which select for some sort of infinitive complement that has an overt subject and no complementizer. The overt nominals following such perception verbs are thematic subjects entirely selected by the infinitives that follow them. ECM verbs in Kinande essentially follow this pattern, insofar as the object-related markers on the perception verb (OM, RFM) correspond to the thematic subject of their ka-V complements, in which case the ka-V complement lacks a complementizer. In the absence of an object-related marker on the perception verb, full nominals following the perception verb correspond to thematic subjects of the subordinate verb. [Note X5]

AS2a)*erí-ba-ow-a uti ba-genda

INF-them-hear-fv that they-leave

“to hear them leaving”

b) erí-ba-ow-a bá-ka-génd-a

INF-them-hear-fv they-TM-go-FV

“to hear them leaving”

c) eri-yi-owa ú-ka-génda

INF-RFM-hear you-TM-go

“to hear oneself leaving” (the default translation - can also be “to hear yourself leaving”)”

A9g) Kambale mó-a-owíre Alícé á-ka-yi-píp-a [môwíre]

Kambale TM-SM1-heard Alice she-TM-RFM-praise-fv

“Kambale heard Alice praising herself.”

gi) Marya mó-á-yí-hulikir-íré á-ká-humúla

Mary TM-SM1-RFM-hear-TM SM1-TM-breathe

“Mary heard herself breathing.”

gii) Marya mo-a-yi-tungerer-ire á-ka-lu-á mo musási [mwáyítúngerere]

Mary TM-SM1-RFM-look-TM SM1-TM-leave-fv in blood

“Mary saw herself bleeding.”

giia) Marya mo-a-tu-tungerer-ire tu-ka-lu-á mo musási

Mary TM-SM1-OM2pl-look-TM SM1-TM-leave-fv in blood

“Mary saw us bleeding.”

X12c) mó-n-á-lángir-ire o-mu-ndú á-ka-génd-a

TM-I-TM-see-TM iv-C1-person he-TM-leave-fv

“I saw a person leaving.”

A14g) Kambale mó-a-a-owíre aba-síká bá-ka-pip-án-a [môwíre]

Kambale TM-SM-TM-hear C2-girl SM2-TM-praise-RCM-fv

Kambale heard the girls praising each other

The translation is rendered with the English gerundive, sometimes called the ACC-ING construction, as in I heard him leaving, which seems also to correlate with adverbial usages in both English and Kinande, such as those in (AS3), where it would appear that the ka-V clause is not a complement, in which case these may be adverbial control structures, perhaps similar to those in English, e.g., to talk to someone while leaving (see also AQ (A9giii)).

AS3a) eri-yi-bwira á-ka-génda

to-RFM-tell he-TM-go

“to tell to oneself while he is leaving”

b) erí-bwir-an-á bá-ka-génd-a

to-tell-RCM-fv they-SM-TM-go-fv

“to tell each other while they are leaving”

However, in cases like (A9giia), the most natural reading is one where Mary saw us not just while we were bleeding, which is the adjunct reading, in which case we might be bleeding and she might not know it when she saw us (also possible), but rather the favored reading is that she saw us bleeding because she saw the blood flowing, and this latter reading is the complement clause reading with the ‘direct perception’ effect discussed in the literature on perception verbs (see, for example, Safir, 1993).

For those less familiar with the considerations surrounding ECM as an analysis of infinitival complementation, a brief review: In English, I told John to leave and I heard John leave are usually given very different structures, e.g., I told John [PRO to leave] vs. I heard [John leave] (the absence of English infinitival to is not crucial for what I am pointing out here, and there are contrasts that work just as well that have it, e.g. I expect the rock to fall, but this phenomenon appears to be limited to perception verbs in Kinande.[Note X6] The perception verbs are argued not to be control-of-PRO structures, while verbs like tell are. The difference is detectable for examples like I told the rock to fall vs. I heard the rock fall. Since telling a rock anything requires a fairy tale animation of the rock, we assume that tell selects for the kind of direct object it can have (i.e., an animate that can understand a message). However, hear does not seem to select in any way the kind of NP that comes after it in these structures, so there is no animacy restriction or indeed any restriction (e.g., I heard all hell break loose, where all hell in English is only an NP for the predicate break loose, showing that hear is not selecting this NP). Because object control verbs like tell involve two sets of selectional restrictions, one imposed by tell and the other imposed by the infinitival verb (e.g., John told the men to scatter, involves animacy required by tell and plurality required by scatter), the proposal that underlies the PRO analysis is that the subject argument is represented by PRO and the object argument that controls it is represented by the overt nominal following tell. Since PRO has to get its identity from the object of tell, the object of tell must also indirectly meet the requirements that the lower verb selects for its subject (in this case, plurality). Similarly, for subject control verbs, like English want in I want to sleep, the subject of the infinitive is assumed to be controlled PRO, and there are two sets of selectional restrictions as well, since #The rock wants to crumble would require the rock to have volition as imposed by want. Complements to control verbs in Kinande take eri-V complements (true infinitives), as in AQ-(A9e).

A9e) Kambale á-sondir-e eri-yi-píp-a

Kambale he-want-fv INF-RFM-praise-fv

“Kambale wants to praise himself.”

The presence of the reflexive on the subordinate verb shows that the infinitive must have a structural subject, one that corresponds to Kimbale, or at least it must in any theory that treats the control complement as a proposition (most theories), and not as a property. For the perception verb, there is only one selecting predicate, namely the subordinate (bare) infinitive which selects its subject (e.g., Mary dropped the bag and I heard the marbles scatter). That is why there is no PRO posited for these perception verbs, since only one argument is required to absorb the single set of selectional restrictions.

Returning now to the examples in (AS2) and (A9), etc., one remaining issue is whether the OM/RCM/RFM on the matrix verb which is thematically selected by the perception verb complement, is associated with some dummy object position that in turn antecedes the missing subject of the infinitive, or if the argument to which these verbal markers correspond is directly the subject position of the lower clause. There are interesting syntactic issues to resolve in order to explain how Kambale can reside outside the apparently tensed full CP that follows it, and yet still be the thematic subject of that clause. We will set this very interesting issue aside (see Note 5) and focus just on the cases involving matrix OM/RCM/RFM.

With these distinctions in mind, we can test to see if the RCM, like the RFM, replaces the -ka- complement subject in the same way in perception verb complements, and the result, in (AS4), shows that when the perception verb has a reciprocal extension (the RCM), the interpretation of the -ka- clause tends to be treated as an adjunct in (AS4a-c) (the TM in the lower verb of (AS4b) is ka...a, which is why there are two final vowels).

AS4a) ba-a-hulikirir-án-á bá-ká-humúl-a

they-TM-hear-RCM-fv they-TM-breathe-fv

"They listened to each other while they were breathing."

b) ba-a-hulikirir-án-á kó bá-ká-humul-á-a

they-TM-hear-RCM-fv how they-TM-breathe-fv-fv

"They listened to each other breathing."

c) bá-a-lángir-an-á bá-ka-náb-a

they-TM-see-RCM-fv they-TM-wash-fv

"They saw each other while washing."

d) mó-ba-á-túngerer-an-ire bá-ka-génd-a

TM-SM-TM-see-RCM-TM SM-TM-go-fv

“They saw each other (while) leaving.”

From the syntactic point of view, the similarity of (AS4a,c) to the ECM construction is only apparent, since the -ka- clause patterns with the non-ECM structure in (AS3b). The syntactic adjunct status of the -ka- clause is even more perspicuous in (AS4b) where a complementizer appears to the right of the matrix verb, creating an opaque domain for anaphora (one way or another, in a variety of theories). However, (AS4d) clearly permits both an adjunct and complement reading. If the latter result is reliable (and with careful controls, it may even be more general), then the RCM is not necessarily detransitivizing, at least in this case, or else -tungerer- would not be able to have a clausal complement.

Summarizing the facts, the perception verb ‘hear’ does not have any complementizer when it occurs with a complement of the form SM-ka-V, and in these cases the matrix verb can occur with the RFM or an OM. When there is an overt NP, like Alice, then Alice must be the thematic subject of SM-ka-V and the lower SM agrees with Alice. If instead of an overt NP, the perception verb has an OM, then the SM of -ka-V agrees with the OM of the matrix verb or, when the matrix verb has an RFM, with the matrix SM (presumably agreement is with the RFM, but this cannot be seen overtly, since the RFM is morphologically invariant). When an RCM extends the verb, then a complement interpretation of the -ka- complement is impossible.

It is not yet clear whether or not these facts fully support to the view that reciprocal extension and the RFM act distinctly on the argument structure of the verbs to which they attach in Kinande. It is clear that the OM and the RFM do not change the adicity of the verb, but it is now not certain that the RCM always does - a surprising result (compare, e.g., Mchombo, 2004). If the RCM is a detransitivizer, and if the clausal -ka- complement of the perception verb is its object, then attaching the reciprocal would make the -ka- clause an impossible complement - no thematic role could be assigned to the overt complement of a detransitivized verb. This would explain why there is a tendency to interpret the ka-clause as an adjunct in (AS4), but it does not account for (AS4d).

Apart from (AS4d), this conclusion echoes others reached by others who distinguish the sorts of processes that RFMs and RCMs participate in other Bantu languages. It has been argued, (e.g., Mchombo, 2004: 83-85, 102-110) that various generalizations concerning tone and ellipsis indicate that the RCM renders the verb stem intransitive, but that verbs with RFMs remain transitive verbs. On this account then, RFM is not a detransitivizer, which suggests in some theories that RCM should attach to verbs in the lexicon (since argument reduction is generally rejected as a syntactic process, e.g., Chomsky, 2001, Reinhart and Siloni, 2005: 403) and RFM does so in syntax.

The issues that arise for theories of anaphora are those of locality and predicate formation. First of all, it is generally taken to be the hallmark of a syntactic relation between coconstrued items, as opposed to a lexical one, when the subject of a subordinate clause falls under the locality domain that normally applies to clausemate relations (e.g., Chomsky, 1981, Reinhart and Reuland, 1993, Reinhart and Siloni, 2005). On this rendering, reflexivization by an RFM takes place in syntax and reciprocalization by an RCM takes place in the lexicon (e.g., Mchombo, 2004) argues that various generalizations concerning tone and ellipsis indicate that the RCM renders the verb stem intransitive in Chichewa. This has consequences for the Reinhart and Siloni theory in that their Lex-Syn parameter predicts that if there is more than one affixal process deriving a coconstrued reading in the grammar, then they are either both in syntax or both lexical operations. They also predict that if the such affixal processes are in the syntax, on their theory, then both RCM and RFM should form reflexive nominals, but only RFMs do (see AQ (A16)).[Note X7] [Note X8] It would appear that the Reinhart and Siloni generalization is disconfirmed for Bantu, if reflexive readings are created by RFMs in syntax and reciprocal readings are created by RCMs in the lexicon.

It is a matter of some interest on independent grounds that the -ka- seems quite similar in semantic force to English progressive -ing, insofar as it also appears in typical progressive statements, as illustrated by AQ (B3b).

B3b) Kámbale álwé á-ka-nába

Kambale leaves SM-TM-wash

"Kambale was washing himself."

The similarity to English could be accidental, but somehow it does not seem so. One can only wonder how it might be predicted, if indeed it should be, that a language with a main clause progressive participle and an adverbial progressive participle might be expected to have perception verbs that take the same sort of progressive complement without an auxiliary.

One area where the Kinande facts seem somewhat clearer than in other Bantu languages, though not fundamentally different, concerns the distribution of proxy readings. Recall that proxy readings for reflexives and pronouns are instances where the pronoun or reflexive depends on an animate antecedent, even though the dependent form does not correspond to an animate object, but some sort of representation or representative token of the animate antecedent. The typical cases raised in English, originally by Jackendoff (1992), concern contexts where an individual sees his effigy at a wax museum and it can be remarked he saw himself at the wax museum, meaning ‘he saw his statue at the wax museum’. These readings are notable since they are not coreferent, but rather instances where there is a dependency of identity on the antecedent, even where there is no coextensionality.

It might be expected that in Bantu languages where pronouns are widely distinguished by noun class, the dependent pronoun would not agree in noun class with its antecedent, blocking the relevant reading in examples like (AS5), and providing a more literal interpretation, where the subordinate SM is in the inanimate diminutive noun class, but there are pronominal cases that appear to have proxy readings, such as the SM1 (B10a) (the vowel harmony sensitive extension -ek- here is sometimes translated as a middle and sometimes as the equivalent of the English affix -able).

AS5) ómu-ami a-ká-lengekanaya a-ti ka-gámb-íre

C1-chief SM1-TM-think Agr1-that SM12-be handsome-TM

(á-ká-hula aka-sanámu kíwe)

(SM1-calls C12-picture of his)

“The chief thinks it is beautiful (when talking about his picture)”

B10a) Tatsopa a-ti a-ká-som-ek-á ndeke omo Kiswahíli

Tatsopa SM1-say SM1-TM-read-able-fv well in Swahili

“Tatsopa says that his writings read well in Swahili.”

(note: the extension -ek- here renders the idea of X-able where X is the verb)

However, neither the RFM nor the RCM are marked for noun class, so the reading will not be blocked on account of noun class distinctions. In these cases, it is possible, with a bit of imagination, to illicit proxy readings. For (B9a), imagine that the wax museum is having a special event, which the wax statues of each celebrity will be washed and dressed by the celebrity they represent, and for (B11a), consider that both Tatsopa and Kavutirwaki are known to have written Kinande dictionaries.

B9a) ómw-ami a-byá á-ká-yi-ery-á na e-ánge, á-kasyá-tsand-i-a erí-bumbá éry’ okó-ri-íso[oko líso]

C1-chief he-be SM1-TM-yi-wash-Fv with c9-attention, he-FUT-spoil-CAUS-Fv C5- mould ASS5 on eye

“The chief washed himself carefully so as not to damage the mould on the eye.”

B11a) Tatsopa ná Kavutírwáki sí-ba-lí-bá-ta-som-an-a omo Kiswahíli

Tatsopa and Kavutirwaki not-SM2-be-SM2-neg-read-RCM-fv in Swahili

“Tatsopa and Kavutirwaki have never read each other in Swahili.

Notice also that there is a distinction between transitive and intransitive ‘wash’ in that transitive intransitive eri-nab-a cannot support the proxy reading, so a different verb ‘wash’ erí-eri-a is used here with the RFM, just the same sort of distinction as we see in English.

The significance of these distinctions concerns not only what we take to be the underlying conditions required to support proxy readings, which in Safir (2004) was argued to be the absence of a requirement for indistinctness (eri-nab-a requires indistinctness, but eri-nab-i-a does not), but also involves questions of what it means for a predicate to engage two distinct arguments. By this criterion, it would appear that reciprocals are not necessarily intransitive in Kinande, a possibility hinted at before. Moreover, if intransitive reflexives are derived in English (as in the theory of R&S) but underlying in Kinande, it must turn out for any such theory that the lexicon will treat intransitive reflexives the same whatever their provenance. Alternatively, if intransitive reflexives are not derived, but a possibility available in any language for a small class of verbs, then the fact that they all act alike does not require any such calibration.

There are a few other areas of interest that have caught our attention or that the passerby should take note of. One such phenomenon, discussed in the AQR in sections 3.1-4, concerns the AGR-ene strategy, which is a form of emphatic or adverbial reflexive in Kinande that typically appears in postverbal position.

T3) Kámbale á-li-asa iyówene

Kambale SM-TM-come alone

“Kambale came alone.”

Most typically, Agr-ene must refer to the subject argument of the clause it occurs in and its shape changes according to the class of this subject, as illustrated in AQ 3.1.1 (T4a-c). There are occasional cases where Agr-ene can correspond to an object if the object is an OM, as illustrated in AQ 3.6.3 (T20). As described in AQR 3.4.3, the -ene portion and the w/b alternation is a noun class agreement phenomenon, C1 vs. C2. The class markers that agree are consistent with the -ene- portion meaning ‘alone,’ but only if -ene- is an agreeing adjective or adverb, since this is not a canonical noun meaning. As is typically the case in the Bantu languages that have them, these emphatic forms do not normally form reflexive predicate interpretations in the absence of the RFM, but rather permit readings attributing to the subject that he or she acted in person (direct proximate involvement in the event), without assistance (alone), or else the subject did something that someone other than the subject might have been expected to do, i.e., a contrastive reading.

There is also a reduplicated form Agr-ene-ene which appears to have a somewhat more emphatic interpretation. In addition to the postverbal position, the reduplicated forms are also acceptable in initial position as in the following examples from AQ 2.2.2.

AS6a) ingyówenewene ngándigendáyô

"Myself, I will go there."

b) itwíbenebene tukándigendáyô

"Ourselves, we will go there."

In such cases, the reduplicated forms can be reinforced by the strong pronouns as in (AS7).

AS7a) ingyówenewene ingyé nyinamundigendáyô

myself me I-will- go-there

"I am the one to go there."

b) itwíbenebene itwé tunámundigendáyô

ourselves us we-will-go-there

"We are the ones to go there."

The reduplicated form can also appear directly following full nominals it modifies (as in (T19d) and (T22a,b)) Agr-ene-ene can also be used in focus position as in (T14). However, an account of the differences in distribution between the single and reduplicated forms will require more research.

There are a number of other interesting phenomena that appear to involve forms of anaphora but that we will not pursue here, such as the Kinande complementizer system, including factors determining the choice of complementizer type and the nature of complementizer agreement. For example, there is a complementizer related to the verb meaning ‘say’ that bears the agreement matching the matrix subject (Agr-ti, as in (AS5), for example). The ‘say’ complementizer is typically selected as the head of CP complements by propositional attitude verbs and it agrees with the subject of the propositional attitude verb. There are open questions as to whether this putative complementizer should be treated as a synchronic verb, but if so, it lacks both a tense marker and an infinitive marker, which would make it unlike other verbs in Kinande. Aspects of complementizer agreement in Kinande have been examined in work by Baker (forthcoming, chapter 4), and will not be explored here. We have also elected not to discuss the reciprocal-comitative construction, but see AQ 4.1.4.2.

NOTES

Note X1: Kinande also has a compositional reciprocal that occurs in environments where the RCM cannot be associated with the target nominal, but the usage is regarded by our consultant (and others consulted) as understandable but very artificial.

A12c) abá-kara mó-ba-ow-íre emy-átsi é-yi-bá-lóleré kó ówundi okó wundi

C2-priest TM-SM2-hear-TM C4-story which-them-concern on other on other

"The priests heard stories concerning each other."

See the discussion of these cases in AQ 2.3.3.

Note X2: That the difference between the single and two part causative might be one between direct and indirect causation, respectively, was brought to our attention by Larry Hyman, personal communication, who led us to Bastin’s (1986) discussion of the matter. Good (2005) suggests that Bantu languages that make distinctions between direct and indirect causation by means of these suffixes are less common. Bastin (1986:116) gives a paradigm with two different forms of the causative for the stem meaning ‘bathe’ in Nyoro that is almost exactly like the one detailed here for -nab-. It is notable that the Nyoro root (-og-) appears to be morphologically unrelated to the Kinande root, another indication that transitivization of this sort, however, inconsistent in some of the synchronic descriptions, has had a productive history.

Note X3: Good (2005) groups the first causative, applicative and reciprocal verb extensions in a separate class of extensions from passive and the second causative/transitivizer, such that the first class of extensions is always inside the second class. He argues against Baker (1988) and following Hyman (2003) that a templatic approach to verb extensions is a better approach to Bantu than Baker’s productively generated structures respecting the Mirror Principle. Good points out that Kinande is consistent with this view in that it has both first and second causatives, and the RCM occurs before -i- in the cases detailed above. However, the transitivizer clearly is associated with a particular causative meaning in its most productive use, which appears inconsistent with the view that it has a meaning less essential (‘relevant’ in Good’s terms) to the verb root than the applicative extension, which it follows. Notice also that it is possible in certain instances to have the reciprocal both before and after the first causative

C25f) abákali ba-kándisyá-kwam-an-is-an-í-a

women they-will- follow-RCM-CAUS-RCM-CAUS-a

"The women will accompany one another."

This would not be an obvious expectation for a fully templatic approach.

Note X4: It cannot be said that inherently semantically reflexive verbs in Kinande are never syntactically reflexive. Just as English allows for verbs that are inherently reflexive but syntactically transitive in (e.g., to behave oneself, to perjure oneself), Kinande also has some lexical reflexives formed with -yi-.

T21) Eri-yi-butik-a 'to sit down' (-butik- is not a root with any identifiable meaning)

Eri-yi-fun-a ‘to boast’ (-fun- does not have an identifiable meaning per se)

Eri-yi-handik-a ‘to resist’ (-handik- means to write and has nothing to do with ‘resist’)

In addition, there are idioms related to body parts with null possessors and the RFM, such as AQ (A4c) (see comment there) and some additional examples below.

A4c) Kámbale a-ká-yi-bulá m’ omútíma

Kambale SM1-TM-RFM-miss- in heart

"Kambale misses the heart in himself, i.e. K. is worried."

i. Kámbale a-kábí-yi-luma okó lúlími

Kambale SM1-TM-RFM-bite on c11-tongue

"Kambale bit himself on the tongue."

ii. Kámbale á-amá-yi-tonika oko ri-inô

Kambale SM1-TM-RFM-stumble on C5-toe

"Kambale stumbles (himself) on the toe."

iii. Kámbale a-ká-yi-buhirirá kô

Kambale SM1-TM-RFM-blow on

"Kambale is well dressed." (lit. K. blows on himself)

Note X5: The complementizer that is used in those cases where an overt nominal precedes the complementizer ko and where the subordinate verb takes a different tense (it is no longer -ka-). Ko is not the complementizer that normally introduces tensed complements (i.e., the uti/ati complementizer).

i. erí-ow-a Kámbalé ko á-amâ-gend-a

INF-hear-fv Kambale that SM-TM-go-FV

“to hear Kambale leaving”

This structure raises interesting challenges for syntacticians because Kambale occurs in an unexpected position, insofar as it appears to be the unselected object of erí-ow-a and the understood subject of the following clause. Other perception verbs such as eri-hulikirir-a (to listen), eri-langir-a (to see) can be substituted to the eri-ow-a slot (cf. eri-hulikirira Kambale ko amagenda, eri-langira Kambale ko amagenda) See Schneider-Zioga (2002) for a proposal that in Kinande the nominal understood as the subject of the subordinate tensed clause is structurally in the subordinate clause, but in a specifier higher than Spec CP. The issues here are interesting, but I shall set them aside.

Note X6: The verb translated as ‘consider’ in AQ (D8) is -langir-, which is a perception verb with a SM-ka-V complement. When translated as ‘consider’ it has a reading and structure reminiscent of a small clause. However, the unacceptability of (A9f”) would appear to indicate that epistemics do not permit ECM the way perception verbs do, though examples in AQ 4.2.1.2 raise the question again. The construction could use more scrutiny.

Note X7: There is apparent backwards anaphora that is possible in reflexive nominals, insofar as the genitive argument is postnominal whether the nominal is infinitival AQ (A17) or not AQ (A16). It is an open question whether or not the postnominal argument is really the antecedent here, or if it is some sort of adjunct and the real ‘subject’ of the nominalization, the antecedent of the RFM, is some abstract structural antecedent (e.g., like pro or PRO in principles and parameters theories).

Note X8: The case of intransitive reflexives is hard to evaluate with respect to the Reinhart and Siloni theory, since they do not consider the possibility that such forms could fail to be derived. They treat arguments that are added to the matrix of a verb as lexically derived (see their Lexicon-Interface Guideline, p.403).

References

Baker, Mark. 1988. Incorporation: A theory of grammatical function changing. Chicago: University of Chicago Press.

Baker, Mark. (forthcoming)

Bastin, Yvonne. 1986. Les suffixes causatifs dans les langues bantoues. Africana Linguistica: 10: 55-145.

Chomsky, Noam. 1981. Lectures on government and binding. Dordrecht: Foris Publications.

Good, Jeff. To appear. Reconstructing morpheme order in Bantu: The case of causativization and applicativization. Diachronica.

Hyman, Larry. 2003. Suffix ordering in Bantu: A morphocentric approach. Yearbook of Morphology, eds., Geert Booij and Jaap van Marle, 245-281. Dordrecht: Kluwer.

Mchombo, Sam. 2004. The syntax of Chichewa. Cambridge: Cambridge University Press.

Mutaka, Ngessimo M. & Kambale Kavutirwaki. 2006. Kinande-English Dictionary with an English-Kinande Index (to appear as Tervuren publication). Published on the website of African Anaphora Project, Rutgers University, <http://www.africananaphora.rutgers.edu/>.

Reinhart, Tanya and Eric Reuland. 1993. Reflexivity. Linguistic Inquiry 24: 657-720.

Reinhart, Tanya and Tal Siloni. 2005. The lexicon-syntax parameter: Reflexivization and other arity operations. Linguistic Inquiry 36: 489-436.

Safir, Ken. 1993. Perception, Selection and Structural Economy,” Natural Language Semantics 2:47-70.

Safir, Ken. 2004. The syntax of anaphora. Cambridge: Oxford University Press.

Schneider-Zioga (2002) Schneider-Zioga, P. "Anti-Agreement and the Fine Structure of the Left Periphery." University of California Irvine Working Papers in Linguistics, Vol. 6