T HE KIRUND I V E R B

John GoLnsMm( and Firmard Sha(M4,NA

l. INTRODUCTION
In this paper, we will present an analysis of the tonal rules and some of the morphology involved in the Kirundi verb. (fn. I) Kjrundi is a Rantu language, classified by Guthrie as being in the D group, and reclassified as J by the Tervuren Dantuists. a group of languages in lhe region generally south of Lake Victoria. It is the major language of I3urundi, and is closely related to Kinyarwanda. There are between four and five million speakers of Ki rundi.

I. Our primary source of data is our (Sabimana)own judgments. We are gratcful

to 3ae Ohk Cho and Wayne Smith for their contribution to this psper in its early stages, snd (o G,N, Clements and Larry ht. Hyman for csreful criticisms ofan earlier vcrsion of this paper. htecusxen'x anslysis ((959) hss heen of usc to us, ax we notc below, Nnne(hetess, this work ol' Meeussen's, unlike much ol' his Ister work, appears to make an efrort in many plsces to eschcw generslizationx. It is a descriptivc tour
de force, snd yet fails to producc general principles. Coupez (l980), on the closcly relstcd Kinyarwsnda, follows in the t radition of M e eussen in scversl respects,

including the crucial charsctcrization of tonal melodies of the various tenses in terms of independent "tiroirs", rsthcr thsn gcnersl rules where possible. Kimenyi (l979), also dcscribing Kinyarwsnda, uses a generative framework, but with somcwhat different simx thsn ours, and our rcsults are in many respects diA(cult to compsre. Sibomsna ((974), also dealing with Kinyarwsnds, presents sn ex(reinely orgsnired description ol' the vsrious vcrbsl tenses, but he neither presents the tonal pa(tcrns

found when object msrkcrs appear, crucial dats in our opinion, nor attempts sny
genersl formulstion of rulcs We uxe the following abbreviations SM Subject marker TM Tense marker FOC Focus markcr OM Object markt:r FV Finsl vowel LOC OM I ~ at i ve object marker Inf Infinitive marker This material is based upon work supported by the hla(ionsl Sciences Founds(ion under Grant No. BNS.8421245,

20

MOI)0LES EN TONOLOGIE

Tl(E KiitIJNDI VEitls

2I

'llie tonal system of the Kirundi verb dilTcrs considerably lrom the tonal systcms of related languages within the Cr'roup 3 area, Thus, in addition to the inherent interest of seeing the internal workings of a complex tonal system, the study of the Kirundi systems is especially interesting in that its tonal rulcs havc a number of propcrties that distinguish thcm from thc types of rules found in such languages as Luganda, Shi, Hunde, Haya, or the more distant Kikuyu, Tonga, Shona or Zulu. Of special interest arc thc scveral tonc rulcs that make reference to syllable or segmental structure, as well as the tone rules that arc scnsitive to highly grammatical, rathcr than purcly phonological and tonological conditions. In more familiar tonal systems, tone rulcs typically modify tonal segments on thc basis ol' the tones in the immediate environment, or add o r d elete association lines in the immcdiaie cnvironmcnt of a tone. Kirundi has reanalyzed its Bantu tonal inheritance in ways that are in certain respects quite radical, and has arrived at an elegant set of tonal rules that are quite different from its neighbors'. For examplc, whcrc ncighboring languages, and no doubt Kirundi's ancestor, possessed object markers which were in most cases inherently high-toned but which were subject to various lowering rules, Kirundi has opted for a system in which the object markers are underlyingly luw bu t subjcct to synchronically odd lowering rules, More significantly, there is considerable evidence that thc rcstructuring of thc tonal systcm in Kirundi is highly inllucnced by a mctrical, or rhythmic, structure that is imposed on the word. Certain of the tone shil't rules appear to be scnsitive to whether the number of mores preceding the site of the shift is odd 'or even. l his kind of global quantity sensitivity is highly unusual for a tone system, and has been found to date only in metrical stress systems (Mc Carthy 1979, Hayes I981). We ar c c o nvinced that only a s y nthesis of metrical and autosegmental analyses will allow us to arrive at a satisfactory account of the Kirundi system. If an account along thesc, lines provcs to be correct, it wi ll s uggest that mctrical and autosegmental modes of analysis are not two specialized formalizations of the same thing, but are rather independeni mechanisms that can be invoked in languagcs, and can be found togcther in certain languages, with the two systcms acting autonomously but affecting each othcr. Also of somc thcoretical interest is the layered morphology that is dcvelopcd in the Kirundi verb. In scction 6.7, we discuss this model cxplicity, and suggest a case where despite the appeal of a simple layered thcory of morphology, it is necessary to have "loops" in the ordered rules ol' morphology, rerniniscent of the Loop in Mo h anan (1981); we investigate the way in which this should be incorporated in tlie formal model, and discuss the non-cyclicity of Dahl's Law.

II. THE TENSES

T he linear structure ol the Kirundi verb is given in l. We w i l l rel'er to the radical with all the following extensions and the Final Vowel as stem, following traditional Bantu terminology.
I. The Kirundi verb. I. The Kirundi verb. subject tense focus object o b j ecl marker marker m iirker m a r kcr ba k ra ya mu t hey I'a r Ihern him pasi c l,6 cl. I "they cultivated them for

radical e xtension fmal vowel rim ir a cultivate for

him/her"

Fach of thesc positions deservcs soiuc discussion. The subject marker is of the lorm V or CV, or in the case of first person singular, a consonant, in thc last casc, the riasal n-. All these morphemes are endowcd with tone at the underlying level, though we will suggest that word-initial moras are incapable of bearing tones themselves ("extratonal", in the sense proposed by pulleyblank 1983), There is one subject markcr for each nnun class, in the traditional Bantu sense, This subject marker generally agrees with the subject NP, overt or implicite: this is by no means the whole story on "subjcct agreement", but will suffice for prcsent purposes. We shall concentrate on fivc mnrphcmes thai appear in the tense marker position in the first part of t his paper; ku, the infinitive marker, which itself cannot be preccded by a subject marker: zoo, thc uture tcnse marker, which is syntactically quite special, as we shall see below in section 7.1; the present te iise marker, phonologically null: a, the recent past marker, on a Low tone; and 6, the far past marker, on a H igh tone. We shall discuss several others below in section 6. The locus marker is somewhat unusual. While it is not uncomm mmon or I n u anguages to distinguish in an overt morphological way whether the inlormation focus, corresponding roughly to the intonation peak in an Fnglish sentence, is on the vcrb or the object, KiRundi has per aps the rnost thorough-going system for such marking. Of the four tenses mentioned above, the three non-future tenses show an overt distinction of this sort. When the focus is on an object, direct or indircct, following the verb (and this is the norrnal case when an object does follow the verb), the focus marker is absent lrom the verb and

22

MODL'LLS EN O'ONOLOG)E

THE KIRUNDI VERB

23

in most cascs High tones arc lost from the vcrb due to a process that we will discuss in detail below. In other cases, the appropriate focus marker for the tense must be used in thc verb. It follows as a special case that when the verb is linal in its clause, it must bear the focus marker. The focus markers are given in 2 and examples are provided
Ill 3 .

3.

Te n s e Fuiure Present Recent

Past

Tense Marker z6o

a

Focus Markcr
fa

a
ra

Far I ast 4. a. u

SM

TM

ra FOC
andik write

andik write
a FV

a FV
ibitabo books

ibitabo

books

" you write books." o b. w SM T ht "you wriic books,"

Object markers, spccified as the subject markers are, for noun class, exprcss the noun class of'an argument of the verb which is either u nspeciflcd - - i n w h i c h c ase the o bject m arker i s p r o nominal i n function, as in Sa or which precedes thc vcrb, as in Sd. 5. a. u ra mu hoo n yc SM F OC OM see FV "you sec him/her" b. u ra bbon ye umuniu FV person Shl FOC sce
"you see ihc person" ra mu bon ye C. u unlccs there is a pause before the object. h, umuniu u ra mu bon person Sht FOC OM scc

appear adjacent to t h e v erb r a dical. Al l o t her o bject markers a r e undcrlyingly Low toncd. T he radical is t y pically of t h e f or m C o VCo o r C o V V Co, a n d extensions are of th e f orm V Co. A s i n o t her Bantu languages, one f requently f i nd s c o m binations o f w h a t a p p car t o b e r a d i cal a n d extcnsion which are not synchronically decomposable, either because thcre is no indcpendcnt radical of the appropriate form, or because the meaning of the combination is not that which would be derived productively from the component radical and extcnsion. T hcre may be any number of cxtensions, from zero on up, Th e extensions contribute to the meaning of the verbal stem in generally p redictable ways, bu t a r e t h e mselves tonally i n e rt, a s t h e y a r e throughout the Bantu languages. In this paper we shall make relerence in several examples to the "applicative" extension, which increases the valence of the verh stcm by one, permitting an otherwise transitivc verb to take two objects, especially important for our purposes when we wish to investigatc the tonal properties of verbs with two object markers, The so-called "Final Vowel" is oddly misnamed, we adrnit, but is in keeping with Bantu terminology. Therc arc thrce morphcmcs that can appcar in this position: thc Final Vowel -c, used in the subjunctivc and in t h e i m p erative wit h a n o b j ect o t her t han th e f i rst p erson singular; the perfective -ye; and thc neutral -a. In most tenses, either - ye or -a ca n h e u sed, with a c l ear d i f ference in aspect or r e a l i s marking. Thus 6a, with -ye, indicatcs that the action was completed, while 6b, with -a, indicates an action not necessarily complcted.
6.a, ha a rim SM TM cul tiv a t e "ihcy werc cultivating-

a
IV

umuniu

a FV

b. ha Sh4

a rim TM cul tiv a t e "they cult iviited"

yc FV

The properties described above are indcpendcnt ol' deliniteness and animacy, unlike thc situation lound in many other Bantu languages, except insofar as topicalization (that is, preverbal placement of a noun phrase) presupposes that the NP is known and definite. This restriciion holds cqually for subject Nl's as for non-suhject NPs, and should thus be viewed as a restriction on surface word order, There is no l i mi t i n p r i n ciple to the number of surl'ace object markers in a single verb, though in practice the limit appears to be three, with the third a reflexive. The reflexive object marker is the only one that is not marked for noun class, and is consistently of the form -ii-, with a High tonc on the second vowel underlyingly; it must always

The meaning of the ye/a contrast in the present tense is worthy of note, ln a wide rangc of'verbs, the -yc lorm is used in the present tcnsc to indicate a single action, and the -a I'orm to i n d icate a habitual action. Tiiis is rather the revcrse of the parallel situation in Fnglish, where the aspcctual marker used to indicatc completion is also uscd to indicate habitual or " structural" properties, in the terminology of Goldsmith and Wolsetschlaeger (l982). See 7. 7. a. ba ra tbek a SM FOC cook I' V "thcy cook (e.g., I'or a living)" b. ha ra tt ets c ( fr o m / ye/ underlyingly) SM FOC coak FV

1'I I E K I RU N O I VERB

25

htODf.LFS FN I O NOLOGIF.

Certain tcnse forms, most notably the futurc, cannot appear with the perfective suffix -yc, though it would seem reasonable from a semantic point of view that such forms should exist, We shall consider this point in greater detail below in section 6.7.

c ontaining a single mora, and we will reformulate the Pullback ru l e within a rhythmic framework in the next section,
IO. Pullback rule

ku

OM

Cy

111. TONE
3.) Introduction inftnitives without object markers divide simply into two tonal classes, those composed with Low stems, as in 8a, with Low tone throughout, and thosc composcd from High stems, as in Sb, with a single High tone on the first mora of the verb stem.
g. a. Low stem infinitivcs ku rim a to cultivatc to raise (childrcn) ku rcr a ku rog a to poison ku rut a to surpass
ku raah a b. High stcm i n l i i iitivcs ku b6n a ku bur a ku bik a to look at

H erc, as elsewhere, we us c t h e f a m i liar c o nvcntion t ha t a n a ssociation line with an " x " t h r o ugh it is deleted by the rule, whil e a dotted association linc is added by the rule. A ll the object markers except the reflexive -ii- behave like th e object rnarker in 9 above. -ii-, howevcr, has an underlying High tone, a s we see in I li . " '
II, a H i g h tone stcm; ku ii bon a "to scc oneself' b. Lo w tonc stem: ku ii raab a -to iook at oncsclf"

ku baag a ku bbcr a

to to lo Io to

scc lack crow hutchcr suit

Recognizing -ii-'s underlying High tone, we must posit a rulc to account I'or the lack of High tone on the stem in l la, a rule given in l2, lowering a High tone immediately following a High tone. This rule is important in the tonology of Bantu languages, and has been called "Mceussen's R u le " i n th e l it e r ature ( see e specially G o l d smith l982,1983,1984). The tone of the reflexive marker is considerably more complicated than thesc initial observations suggest, however. We will discuss this matter in detail in section 4.2 below.
I2. Mccuscn's Rule

The basic tone of thc verb radical is thus realized on the flrst mora, the usual situation found in l i a n tu languages. This basic tone is realized on the preceding object marker, if therc is onc, in the infinitive, as illustrated in the cxarnple in 9a, lormed from a High stem. An otherwise parallel examplc is given in 9b, formed with a Low stem, where no High tone appears; we conclude that the High tone in 9a is truly the rcalization ol the stem High tone, displaced leftward,
9. a. ku gi tcm inl'. O h1 cut "to cut it (cl. 7)" b. ku ki rim inf, OM culti va t c : "to cultivatc it (cl. 7)
a

Co

H ~

FV

FV

2. We will oot discuss in detail the intercxting and importsnt question of the trcstcmcnt of vowct scqucnccs. We msy note (hst as with all Fsstcro Bsotu lsngusgcs, within the word scqucoccs ot two or more vowels reduce to two more scqucnccx, with the Issl vowcl dctcrraining the vowct quality of the I oi t snt long vowel. Both Ctcmcnts I984 l)979l siid Wall l984 cootaio analyses of this phcoomcooit thst wc find

We shall provisionally account for this " p u l lback" of H i g h t one by means ol' a rule whose formulation wc give in l0; we shall call this rule the "Pullback Rule" We will return briefly to the naturc of the rnorphemes other than -ku- that trigger the Puiihack rule, atl ol'them

useful. As hss hccn tiotcd ciscwhcrc (for cxsmplc, in the discussion of Mskos in Kcnstowicz and Kisscbcrth I 979), when scvcral tones are associatcd with a n underlying scqacncc of more thstt two moras, destined to become s two.mors syllahlc, the principle thst ststcs thc surfscc rcslizstion of tone is this: i l' thcrc are any Iligh toncs on the third or Istcr mors, these are rcslixcd on the second mora of the rcsuhing tong vowel.

26

MODBLES EN TONOLOCitE

THE KIRUNDI V F Rtt

The Pullback rule formulated in 10 applies in the Present Tense a s well, but no t i n t h c R c cent o r F a r Past tenses, In 13, w c g i v e repre presentative examples of verbs in the present te n s e with no and with s ,t o ne object marker. In al l cases, the focus marker must be pre,en since, as we shall sce below, verbs out of focus display greatly reduced t onal patterns, We reformulate 10, the Pullback rule, to apply in t e present tense as well. The -ra- focus marker which is said to trigger t he Pullback from stem to object marker in 10' must be said to b e distinct from the -ra- focus marker of the Far Past, where no pullbac occurs (this is illustrated in 15 below).
10' Pullback rule

OM
present

CV

%ith one ob)ect marker n da gi tcck a tu rs gi teek s u rs gi leek s mu ra gi teek a s rs gi teek s bs ra gi teck s d. Low tone verb -raab- "look at" Without object marker n ds raab a tu ra rssb a u ra rssb a mu ra rasb a s ra raah s ha rs raab s With one objcct markcr n da mu rssb a tu ra mu rasb a u ra mu raab a mu ra rnu rsah s a rs mu rsab s ba ra mu raab s

Consider the following data, given for 1st, 2nd, and 3rd pcrson (singular and plural) subjects, hcre as throughout this paper.
13. Present tense (focus form)
s. High tone verb -b6n- "scc" Without object marker n da b*n s tu ra b6n a urs hon s mu ra h6n a a ra b6n s ba ra b6n a With one object msrkcr (-mu- -him/her") n ds mu hon s tu ra mu bon a u rs mu bon s mu rs mu bon s a ra mu bon s bs rs mu bon a

The Pullback Rule also fails to apply in the Future, Recent Past, and Far Past tenses, but the tone of the affirmative main clause Future verb is complicated hy two tense-specific rules which we will discuss below, For present purposes, we may summarizc these lacts hy saying that the Pullback Rule applies when thc Tensc plus Focus Markcr is of thc form CV, but not in other cases, such as when the Tense plus Focus are ol the form CVV (futurc), VV (reccnt past), or VCV (far past). This is partially illustrated in 14, with the recent past forms; the Far Past is illustrated in 15 below. For the reader's convenience, in all but the first paradigm we give near-surface forms; in all cascs, [uj before a vowel becomes a glide, and sequcnces of three vowels simplify to two (as in ba a a rim a becoming baarima),
14. RecentPast (f ocus forms )
s. High tone verb -bon- "see": short vowcl after morphological rules surface forms naab6ns t ua s b 6 n s nasb6ns twasb6ns uaab6ns m us s b 6 n a waah6ns mwaah6na y s a b6n s bs s a b6n s y aab6na basb6 n a with one object marker nss mu bona tu a a mu b6n a u s a mu b6n s m us s m u h o n a y s s mu h6n s bs s a mu b6n a b. Low tone verb rim cultivate short vowcl nasrima tu s a ri m n u s a rim s mu a s rim s y a a rim a hs s a rim a With one object marker n a a ki rim s tu s a ki rim s u a a k i r i ms mu s a ki rim s y a s ki r i m a b aa a k i r i m a

b. Low tone vcrb -rim- "cultivstc" Without object marker

n da r i m a urs rim s s rs r im s tu rs rim a mu rs rim s ba ra rim a

With one object marker (-ki- "it", class 7) n da ki riin a tu ra ki rim s u rs ki rim a mu ra ki rim a a ra ki riru s bs ra ki rim s
c. lligh tone verb -t6ck- "cook" Without object marker n da thek a tu ra thek a u ra t6ek s mu ra t8ck a s rs t6ck a bs ra ttek a

28

MOI)PLLS EN YOY OLOOIF

'I HE K I R UNOI VP R II

29
twshrski rims mwahraki rima bahraki rim s

c. High tone verh -teek- "cook"; long vowel without object msrkcr n aa t hek a tu a s thek s u s s leek s m ua s t h ek s b as s t h c k s y s s teek s with one object marker tu s s gi thek a n a a gi thek a u a a g i t hek s mu aa g i t h eka

With one object marker tu a ra ki rim s nsrhki rima u a ra ki rim a mu h ra ki rim a w arh k i rims y A ra ki rim s ba h ra ki rim a yarh k i rims c. High tone verb -thek- "cook": long vowel Without objcct msrkcr n h ra thek a tu h rs thek s nark teeks uh r a l bek a mu A ra thek a warh teeks
nh r s k > r t m a

twshra theks mwshra theka

y a a gi thek a

ba aagi thek a

y h r a t h ek a

d, Low tone verb .rash- "look st": io ng vowel without object marker ns a rasb a tu s a rssb a m ua a r a s b s us s rssb s ba a a raab a y a a raab a with onc object marker ns a m u r ssbs t ua a m u r a a b a

u aa m u r a sb a
y s s mu rssh s

mu a a mu raab a
h ss s m u r s a b a

I II. I T O N E I N F A R P A S T The tonal pattern found in the Far Past diffcrsin certain respects from what wc would expect on the basis of thc forms discussed up to this point, As thc data in 15 below illustrates, there is no contrast bctween inherently High and Low toricd stems; both are assigned a High tone in th is t e nse, although this High tonc may he lo st b y Mccusscn's Rule.Furthcrmorc, thc tense markcr -A- has a High tone in this tense, though it may be displaced if it falls on the first syllahlc o f the w ord. Spacing in th c s urface fortns given is p urcly l o r orthographic convenience,
15. Fsr Past (focus forms) a. High tone verb -bon- "sec": short vowcl after mo rphologics I ruies surface form n A ra bbn s tu A rs b6n s narh hona u A ra bbn a mu h rs bbn s wsrh bona y h rs b6n s b a h ra b 6n a yarh bona With one objcCI msrkcr n a ra mu b6n s tu A ra mu bbn a nsrhmu bbns u A ra mu b6n a mu h ra mu bbn a warhmu b6na y A ra rnu b6n a b a h ra mu b6n a y ar h mu bona b. Low tonc verb -rim- "cultivate': short vowel n h ra rim s tu h rs rim a narh rima uhrarima m uh r a r i r n s wsrh rim a y A ra rirn a ba A rs r i r n a ysrh rim a

bs h ra thek s ysrh t e eks bahra theka With one object marker n h rs gi thck a tu A rs gi thek a naragi thek a twaAragi thcka u h r a g i t h ek s mu h r gi thek a war hgi thek s mwahrsgi theka y A rs gi thek s bs h ra gi thek a y arhg i t hek a bsaragi theka d. Low tone verb .raab-: tong vowei Without object marker nh rs rhaba tu A ra rhab a narh rasbs twahra rhabs uh rs rhsba mu h ra rhab a wsrh rasbs mwahra rhaba bs h ra rhab a yarh r a a ba bahra rhaba yh r a rhsba Wth one object marker n h r s mu rhsb s tu h ra mu rhsb a n arh m u rhaba twaAramu rhabs u hrs mu r hsb s mu A r mu rhsb w a r hmu rhsha mwshrsmu rhsbs y A rsmur hsbs bs h rs mu rhab a y a r hmu rhaba bshramu rhabs

In all cases in 15, the stem carries a High tonc, except when a High tone immediately precedes on the focus marker -ra-, which is the case when there is no object marker and the subject marker is singular (although the corrcct gcncralization depends ultimately on thc phonological shapc, not the grammatical number, of thc first morpheme). The loss of the stem tligh tonc in these cases is clearly the result ot Mecussen's Rule 12, lowering a Iligh after a High. The High tone on the stem is placed by a rule that we shall formulate as in 16,
16. Var Past Stem-tone Placement
stem

I
twshrs bbns rnwssrs bbns
I

bahra b6na

FAR PAST

IwaArsmu b6na mwshrsmu bbns bahramu b6na

twahra rima mwshra rims bshra rlma

We must look a bit further to sec what mechanism is responsible for shifting the High tone frorn the Tcnse Marker to the Fncus Markcr in forms like narhtema. Mccussen (p.26) notes that "un morphoton6mc haut appartcnant h unc voyelle initiale, ou mime A une nasale initiale, est reprhsentbc sur la voyelle suivante." Th is general principle is illustrated by the shil't of tone, for example, in thc demonstrative serics (Mcussen's series lll) in 17.

TIIL kI R U N n t V F R I I MO D & LES LN TONO I nGIE

3I

17. hloun class

3 4 5 6 7 8

Form u + n6 hk + no u + n6 i + n6 rl + no a + n6 ki + no bi + no, e t c .

(u) n6

versus hk

H
u ara tem a

Intermediate structure (after texical creation of the stem)

L H

tt

We suggest that all tones both Il igh and Low are in clTcct shifted off the fi rst vowcl of the word i n lCirundi, and this not due to an explicit rule to that effect, but rather to the specification that a word-initial tone-bearing unit is "extratonal", in the sense proposed by Pulleyblank (1983). A unit which would potentially be tone-bearing in an autosegmental I'ramework may be marked as extratonal and is thus disregarded by t onc assignmcnt rules of the language. This is e flected by r u l e l g . N o t e t h a t i t a p p l ies t o strictly w o r d-initial v owcl-positions, not m c rely t o t h e f i r s t v o wel-position ol' a w o r d (lootnotc 3). We will return to thc question of extratonality in more dctail in scction 6.5, where wc contrast extratonality at the phrase-level with extratonality, as here, at thc word-level.
lg, Initial Fxtratonality

Qu ara tem a

Extra t o nal marking

l
L II II
A ssoc i a t ion, l4teeussen's Rulc, Cilide formation

tem a ara '

' '

L H

(+ extratonal]

In the derivation in 19b, we anticipate our discussion helow of' the cyclic formation ol' the verb from the verbal stem -tcma. We assume that tone has already been assigned to the stem at the time when tone is assigned to the Subject Marker and Tense Marker, I'or reasons that wc will considcr in morc detail bclow. We assume, with most recent work in the autosegmental I'ramework, that Iloating tones automatically associatc with the lefmost accessible unassociatcd vowel, if, in particular, they become free during the derivation. 3.3. Verbs With Two Objcct Markers. When we turn to verbs with two object markers, we might expect to find no H igh tones present when the stem is underlyingly Low in tone. This expectation is not met, and throughout the tense system, in fact, the undcrlying tonal contrast on the verb stem is lost when the verb stem is preceded by two object markers. II' we look first at a tense such as the Recent Past, a tense in which the Pullback Rule l0 does not apply, we find tonal patterns as in 20. The Far Past is illustrated in 2I. F or ease of' cxposition we have placed a High tonc in each case on the left hand column on the tense-markcr -6-, cvcn though its actual first association is with the following vowel, as we have just suggested. We present the examples broken down morphologically with the tone assignments, according to our analysis, found after thc morphologically-scnsitive rulcs have applied; thcse are followed by (hc true surface forms.

Wc shall reprcsent extratonality henceforth by a circle drawn round the extratonal segment (recalling a floating tone, which is an "extratonal" tone, in the sense that it exempted frorn association principles). T he association of the Class I d emonstrativc takes placc as in l 9 a ; the shift of the High tone in a form such as w-ark-tem-a is illustrated in 19b. When the supcrficially parallel b6-tem-a occurs with a H i gh tone on the first vowel (in the participial form, discusscd bclow), no shift of the High tone occurs, bccause thcrc the crucial assignment of extratonality does not takc place.

3. We are grateful to Larry Hyman for suggestions leading to an improvemeni in the formulation of this rule,

MO[>f:LFS LN TONOLOCitE

THF. KIRUNDI V F R B

33

20. Recent Past: Two object markers (focus on verb) a. High tone stem -tem- "cut" short vowel "I cu! it for him " ln a a k i m u t ern er a] n a a ki mu thm er a "they cut it for him" [b a a k i mu tem er a] ba a a ki mu thm er a b. Low tone stem -rim- "cultivate" : short vowel [n a a ki mu rim ir a] n a a ki mu rim ir a ]ba a ki mu rim ir a]
c, High tone stem -thek- "cook" : l ong vowel d, Low tnne stem -raab- "look at" :

ll' we look now at infinitives and Present tense verbs, I'orms in which the Pullback rule ]0' applies, we find a curious result: the l l igh tone on the second object marker that has been placed by the Two Object Markcr rule 22 is indeed pulled back to the first object marker, as we would expect; but furthermorc, a High tone appears on thc verb stern w hether its basic tone is High or Lo w
23. Infinitivcs with a , High tone ; ku hg 1n( LOC -O h l "to work there for b . Low tone : ku hh I nf there "to cultivate there two object markers
nlu OM him"

y aa k i m u r a a b i r a ba a ki mu raab ir a

ly a a ki mu teek er a] lba a ki mu teek cr a] l ong vowel ly a a ki mu raab ir a] ]ba a ki mu raab ir a]

khr work

eI'

appl

a FV

2I. Far Past: Two object markers (focus on verb) surface After rnorphological rules a. High tone stem; short vowel y a rk ki mu tem era y k ra ki mu t0rn er a ba k ra ki mu tem er a ba a ra ki rnu t0m er a b. Low tone stcm: short vowel y a rk ki mu rim ir a y 1 ra ki mu rim er a ba 5 ra ki mu rirn ir a b a S ra ki mu ri m i r a c High tone stcm: long vowel y a rk ki mu teek er a y 5 ra k i m u t b ek er a ba 6 ra ki rnu teek er a ba R ra ki mu tbek er a d. Low tone stern; long vowe1 y a rk ki rnu raab ir a y ll ra ki mu rhab ir a ba 6 ra ki mu raab ir a ba tbl ra ki mu rkab ir a

mu rim OM cultivat e for him/her"

a a pple F V
lf

24. Present tense wlth tv o obyect markers a. High tone : a ra hk mu kor SM FO C LO C . OM OM work "He works there for him"
b. Low tone '. a ra hh mu SM FOC LO C OM OM "He cultivates there for him"

er appl

a FV

rim cultivate

ir appl

a FV

We scc that ineach case in 20, a High tone appears on the second object marker, and we can infer that the rule that accomplishes this applies before Mceussen's Rule, changing the underlying High tone on the stem to Low. This rule is provisionally given in 22. Much the same pattern is found on verbs with two objects markers in the Far Past, with a l l i gh tone on the sccond object marker, and a Low tone on the verb itself, illustrated in 21.
22. Two Object Marker Rulc
Obj Marker

We propose that there is a rule, which we shall call the Rhythm Rule, whose function is to place a second High tone two syllables to the right of the ll igh tone that is created by the Two Object Marker rule when that position is on the first stem syllabe (or on an object marker). This rule is formulated in 25.
25. Rhythm Rule

QM

OM

Syl l
I
I

Obj Marker

Stem

(thc dottcd line on the second H indicates that the structural change of the rute inserts the H), 3.4. Non-l'ocus Tone. Wc will provisionally formulate one more tonal rule in this section. Wc noted above that in n o n-focus main clause verb, where the principal focus of the sentence is on a phrase following

MODCLES EN TONOLOGIE

THI: KIRU NDI V E R B

35

the verb, High tones which would otherwise be expected on the verb are lost. It' we indicate non-I'ocus status by -F, we may formulate this generalization as in 26, which takes account of the fact that only tones to the Right of the tense marker are lost in the -F contexts. The High tone of the tense marker in the Far Past is not lost, in particular, in the non-focus forms', however, tones on the verb stem, the reflexive object msrker, snd the tones created by the Two Object Marker rule are all lost in ihe non-I'ocus context
26. Non-focus Rules Main ctsus

IV. A R H Y T H M I C

A PPROACH

4.1 INTRAD UCTIO N

H ~ L / T ense X--y F verb main clause

We give representative examplcs in 27 ( t hird p erson, singular and plural) of non-focus main clausc verb forms. Forms with long vowels arc cnlircly parallel.
27. Non-I'ocus main clsuscvcrbs ' . Pr esent Tense
s. High tone vcrh, shnrt vowel

Subj No O M O ne O' I sg. a b o n s a mu bon s pl. bs bon a bs mu bon s b. Low tone verb, short vowel sg s gur s s mu gur i r a bsgur s b s mu gur i r s pl

Reftex OM yii bon s b i i bon s

Two OM s ki mu bon er s bs ki mu bon r s

y i i gur i r s b ii gur ir s

s ki mu ir s
ha ki mu gur ir s

28. Non-focus main clause verbs: Recent Past s, High tonc verb, short vowel sg. y s bo n s y s mu bo n s y i i bon s y s ki mu bon er s ba a ki mu bon cr a p l. basb on s bs a m u b o na h i ibo n a b. Low lone verb, short vowel

W e have observed so fa r t ha t t h ere are tw o t e nses in w h ich t h e P ullback R ul e a p p l ies, th e I n f i n i tivc ( a n o m i nalization) an d t h e Present tense, formed with ku- and -ra-, respectively. In addition, there is another tensc, whosc tense marker is also of the form CV, in which thc Pullback rule applies, the consecutive -ka- tense. As we noted above, neither the Recent Past -aa- nor th e Fa r Past -ara- trigger Pullback, nor does the I'uturc marker. W c propnsc that the Pullback phcnomcnon skctchcd in IO' i s a ctually motivated b y a r h y t h mical, o r r n etrical, structure that i s constructcd on th e w or d b eginnning with th e l i rst O bjcci M a rker. Whether the Object Markcrs plus the following stem forms a constit uent is ao important and still outstanding problem in t h s t udy o f Bantu morphology, but in the Kirundi vcrb therc is evidence that this unit, which we shall call the suprastem, for want of a b e tter term, serves as the b asis fo r t h e i m p osition o f r n c trical structurc f rorn lcft-to-right. We shall see below (section 5.2) that (his metrical structure is motivated in the treatment ol subordinate, subjunctive, and negative vcrb forms, and that the necessary metrical structurc is again that erected on the suprastem. We propose, therefore, that binary-branching trochaic feet (Strong-Weak) be erected on the suprastem (thc s tem plus any p receding Objcct M a rkcrs). As w e shall show itt a rnoment, this metrical structure serves as the basis for thc Pullback rule in main clauses. We must thercforc address the question of how to dcal with the contrast hetween those tcnses which do trigger pullback ancI those which do not. lt appears that in the synchronic grammar ol' Kirundi, it is not possible to prcdict phonologically which tcnses trigger the Pullback rule. The h i storical o r igins o f t h i s d e velopment are d i scussed in Goldsmith (1984). Consider some examples from the Present Tense. With no object marker, as in 3la, the lexical tone of the verb stem falls on a Strong syllabc. In 3lh, with a single objcct marker in a Strong position, the

sg. y s gu r s y a mu g u r i r s p l. b s s gu r s bs s mu gu r i r s

y i i gur i r s b ii gur ir s

y s ki mu gur ir a ba a ki mu gur ir a

29. Non-focus main clauseverbs. Far Psst s, High tuoe vcrb: short vowel sg. y s h6 n a y a mu bo n s y ii bon s b ii bon s pl, bsh h u n s h s 5 mu bo n s b. Low tnoc vcrb : short vowcl sg. y a g u rs y a mu g u r i r s y i i gur i r s pl. b s A g u r s b s 0 m u g u r i r s b ii gur ir s

y s ki mu bon cr s bs h ki mu bon cr s

y s kl mu gur ir s bs A ki mu gur ir a

MOnLLFS LN rONOLOGi e

TIIE KIRLINI>I VI:RB

37

tonc is shifted onto the Strong position, by thc Pullback rule, now r eformulated in 32." '
3I. Prescnt Tense

32. Strong Pullback Rule

CV
a.a r a h o n a b. a r a m u bon a

l[a ra bon a[

a ra mu

bo n a

[a ra mti bon al

4, It is appealing to predia on a metrical basis whlch tenscs will trigger the Puilhack rule. One a(tractivc hypothesis that we have explorcd, and which appeared in an earlier draft of t h c p rcscnt paper, involved constructing mctrical structure starting from the subject markcr, and dropping the "CV" condition in the Pullback rule. Thus when thc Tense + Focus marker contained an cven number of moras, as with thc Near and Far Past and the Futurc, we would find that the ftrst object marker would always hc in a metrically Weak posi'tion. It would thus never attract the tone through the Pullback rule. When the Tense + F ocus containcd and odd numbcr of rnoras - likc I. as in the present -ra- -- then the Iirst object marker would fall in a metrically strong position, and auract the next High tone via thc Pullback rulc. This position is so aesthetically appealing that it is hard to rcsist, but as a synchronic analysis it has several problems. The first is that the -racha- tense has an odd number of moras in its Tensc-Focus marker lthree, to b prccise), but it does not trigger Pullback. Second, the infinitive has only one mora, and functions in (he same way as the other odd-moraed markers (like .ra-, etc.), but metrical structure cannot be consistently erccted to put ku in a metrically weak position, like the other markers ol the I'orm CV; this problem arises simply because the other CV markcrs appcar in finite clauses and are thus preceded by the monomoraic subject marker, Thirdly, the pattern of metrical structure beginning with the subject markcr is difricult to reconcilc with the mctrical structure that is erected on the hasc, which we will suggest below is necessary for the treatement of tone in subordinate. subjuntsive, and negative verb forms. Fourthly, and nnally, the alternative mentioned in this footnotc requires a more specilic formulation of thc Strong Putlback rule 32, in that it must be stipulatcd that the High must be pulled back onto an Object Markcr, and ncver onto a Focus Marker; if this stipulation is not addcd, and mctrical structure is crected starting with the

T he rule f o r m ulated i n 3 2 w i l l r e p l ace th e P u l lback r ul e i n I O ' formulated initially. As o r iginally I'ormulated, it ap plied equally t o verh forms with two object markers, as in 23 and 24. It should be clear that the reformulation in 32 applies equally correctly to these forms with the Two Object Marker rule formulated above, assigning a Iliglt tone to the second object marker when two object markers are present. This is illustrated schematically in 33. Thc Rhythm Rulc, in addition, always adds a Iligh tone to a metrically Strong position, an aspect not explicitly shown in the formulation presented. 33.

SM

TM

OM

OM

Stem

suhjcct marker, then the stem High tonc will be wrongly pulled back in a Rcccnt Past form such as y a a bbn a, producing 'tu k rg bon a, which would then lower by Meeussen's Rule to 'tu k ra bon a. This matter is discussed at lenght I'rom a historical point of view in Goldsmith (in prcss c).

4.2 Rellexive -ii- We commented earlier on the comple,x behavior of the tonc of the rellcxive markcr -ii-, and we have held the tonal patterns of thc reflexive forms aside until now, when we can make reference to metrical structure, In 34 we present thc relevant forms from the Infinitive, the present tensc, the recenl past, the far past, and the -ka- consecutive tcnsc. In each casc, we present forms with just the reflexive marker, as well as for tns with one non-reflexive object marker followed by thc refiexive. We give only cases with lhc Class I subject marker, except in the Far Past, where thc choice of subject marker can directly affcct where the High tone of the tense marker associatcs. In the infinitival form with just the reflexive marker, there

38

MOOBLES EN TONOLOGIE

1 HE RtttUNDI VEttft

are two fortns in frec variation. When only a High tone verb form is given, there is no contrast on the surlace between High and Low tone verbs. [Fn. 5]
34. Tone with Rcflcxivc Marker Rcficxivc Markcr OM + R eflexive a, Infinitive kwiitcmcra or ku mwli tkmcra kwiitcmcra
h. Prcscnt a r ii tcmcra c. Recent Pas t

35. Present Tense

a, Incorrcct

S W5 Q $

i tcmcra

a ra inwii t6mcra

'R, 'R"
h. Corrcct
a ra mu i t ct n c ra

y ii tcmcra d.Far Past sg. y ar ii tcmcra pl. ha ftr ii tcmcra

y aa mwii tcmcra

y a rh mwii tcmera ba 6 ra mwii temera We suggest, thereforc, that there is an e a rly ru le in e iTect, of allomorphy that applies bef'ore mctrical structure is assigned that shortens the reflexive marker when it fol)ows an object marker (in main clause forms, the only ones wc have considered so far). This rule is given in 36.
36. Rcflcxivc shortening (focus forms)

We suggest the following analysis of these fortns. The underlying Icngth of the rcllcxivc marker is difficult to determine directly from the surface lacts, and it sharcs this characteristic more generally with the vowel-initial radicals, where there is no contrast betwccn short ayd long vowels. We wi l l s u g gcst below, in o ur a c count of t o ne t n suhordinate f'orms, that there is strong reason to takc thc reflexive to be basically long (as it is in other Lacustrinc languages). The behavior of the High tone on the reflcxive in the forms in the left-hand column of 34 suggests the same result, The High tone is pullcd back from the second mora to thc lirst mora in exactly thc tenses where the Strong pullback rule applics (Present, Consecutive, and here optionally, thc Infinitivc). Since we know that the Strong Pullback rule does not pull back a High tone onto the -ra- Focus Marker, the existencc of a form such as Present a r li teincra points in the direction we are suggesting. However, the forms in the second column also show thc cffects of Strong Pullback in the relevant tcnses, and if the Reflcxive marker werc treatcd as bimoraic (thus having on a Wcak and a Strong mora in this case), Strong Pullback would not apply, as we illustrate in 35,

Summarizing, then, we have the following rules applying in the main clause forms considercd so far: 37. Summary (Main clause forms) 1. Assign High tone to radical in Far Vast 2. Initial extratona1ity 18 3. Rcflcxivc shortcning 36
4. Assign trochaic inctricai structure ori

5. Two OM rule 22 6. Mccusscn's rulc 12 7. Strong Puiihackrule 32 8. Rhythm rule 25 9. Non-1'ocus rulc
5. The 1'ar psst plura1 also has a possible pronunciation [ba a rii tctncra),

40

MODL'LLS FN TONOLOGIE

THF. KIRUNDI VERB

4I

V. SUBORDINATE CLAUSES.
5.1 I NITIAL GENERALIZATION S: E X T E NSION TON E
We will turn our attention now to the morphological and tonal behavior of verbs outside of the main clause indicative affIrmative. 1 he vast majority o f t h c r e m a ining f o rms share a g r eat d eal i n common; in f a ct, the general picture that one gets of th e K i r u ndi verhal system is that the main clause affirmative is, in a sense, the odd man out, in two primary ways: f i rst, the focus/non-focus distinction is lost almost c vcrywhcre outside of t h e m a i n c l a u se affirmative indicative. It is, in fact, this characteristic that distinguishes them from the other forms; thc other vcrb forms (all negatives, embcdded or not, subordinate clause verbs, Imperatives), with one very marginal exccption, nowhcre express this contrast. Second, and consequently, we f ind another generalization emerging in al l n o n -focus verb forms : there can bc only onc High tone per phonological word in non-I'ocus verbs, In this second respect, the non-focus forms look more like an accentual language than a tone language, perhaps, and the incursiolt of rhythmic metrical structure that we suggest sccms directly relat4cd to this observation. The developmcnt of rhythmic structure means thai, unlike the situation i n a t r u e t o n e l a nguage, not al l s y l lables are endowed equally with the potential to be tone-bearing; weakly-placed syllables will tend to lose their toncs. This appears to lead, in turn, to the one-ione-pcr-word surfacc restriction, about which there will be considerably more (o say bclow. We will begin by presenting representative data on the subordinate clause forms. Thcsc verb forms are used in all subordinate clauses t hat arc complements to v c rbs, complementizers, or ar e f o und i n r elative clauses; Meeussen misleadingly calls this pattcrn " r elatif ' , overcmphasizing onc of its uses. All tonal contrast is lost between verb stems which are lexically High and those which are lexically Low, so we present only one High toned stcm (-bon-); Low tone stems show the same pattern. Stems with long vowels behave somewhat diffently, for reasons that cmcrgc from a mctrical analysis; we will reiurn to this in Scction 5.2,
38. Subordinate: I'rescnt Tense Suhjcct N o O hl One OM Singular a bon 6r a a mu bon 6r a Plural ba bon 6r a ba mu b o n 6 r a reflexive OM y ii bon cr a b ii b6n er a

39. Subordinate: Recent Past Singular y a bon 6r a y a mu bon g r a Plural ba a bon 6r a ba a m u h o n 6 r a rcflcxivc OM y ii b6n cr a b ii b6n er a
4th Subordinate: Far Past S ingular y a h6n c r a y a mu bo n e r a Plural b a h bon er a b ag m u b o n c r a reflexive OM y ii bon c r a b iibon er a

y a ki mu hon cr a b aa ki m u b o n e r a

y a kl m u b o n c r a ba 6 ki mu bon er a

Two striking generalizations emerge from this mass of data. The first is that in many forms an unexpected High tone appears on the second mora ol the verb stem. The second, noted just above, is that there is, and can be, no more than one High tone per word. In the Far Past, this is the High tone on the 'Iense Marker, which may be in cf feot p ushed to t h e s e cond s y l labe b y r u l c s which w e h a v e a l ready cstablished'. in ihe others, it is a High tone on the second mora of the verb stem. The importance of hoth of these generalizations will cmerge as we proceed. [Fn. 6j, Let us begin hy I'ormalizing these two observations, There is first a High tone sul'fix that is added to the tone of the verb of all verbs in subordinate verbs (and, as we havc noted, this will he extcnded to n egatives, subjunctives, and imperatives). This High tonc w il l au to m atically associate with thc f i rst availablc mora, which w il l b e t h e sccond mora, since the Cirst is associaicd with th c stem tonc. Th e morphology thus has a suffsx -H which is suffixed in all subordinate forms (as wclt as negativcs and subjunctivcs). There is, in addition, a rulc 41 which lowers an immediately preceding tnne beforc a High in the stem. This r ul e d oes no t n eed t o m e n tion "Stem" e x plicitly, because it is a Level One rulc, in the sense described below (Section 6.7); the Stem comprises a layer in the sense of lexical phonology, as we shall see below,

Two OMs a ki m u b o n r a ba k> mu bon er a

6. In vinually ali of thc closly rclatcd group J Ianguagcs of thc Lakc Victoria region. and more gcncraily in nnnhcastcrn Bantu, tense and other grammatical inl'ormadon is convcycd by various paucrns of High tones distributcd ovcr the verbal cxtcnsion and the Final Vosvcl, Kirundi has lost virtually all of this tonalacdvity. For

other systcms. scc Iiyman and Byarushcngo I9g4 on Kaya, Stcvick I969 on Luganda, Massammba l984 on CIRuri, Goldsmith 19$4 on Tonga, GoIdsmhh (in prcss b) nn Kiliundc, Clcmcnrs I984 on KIKuyu, Polack-Bynon l975 on Shi, and Stappcrs I973 on Mhuku.

IviODf'LES L'N I ONOLOGIE

'I'IIF. KIRUND I V L'RII

43

4L T~

1. ( j M
s'tc Jn

Next we need a rule of tone lowcring in the subordinate forms, and more generally for the forms that cannot contain a f ocus-marker (subordinate, ncgativc, subjunctive). As we bricfly noted above, in subordinate clauses the principle can bc maintained that the first High tone "wins"; that is, although at a deeper level High tones may bc distributed by pr i n ciples whose independent functioning we can isolate, on the surface only the Ieft-most of these High tones surfaces, and all others are lowered to Low. This is expressed in 42. 42. Non-focus forms: First Migh Wins H ~ L/ I M x suhordinaie

the first syllable when there is one (Low toned) object tnarkcr present. When there are two object markers prescnt, of course, the Two Object Marker rule assigns a Eligh tone to the second, and this is the High tone which prevails on the surface,
43. Suhordinate: Present tense: Iong vowel No OM One OM Two OM a teek 6r a a mu tehk er a a ki mu teck er a ba teek hr a ba mu te6k er a ba ki mu teek er a Reflcxive Non.reflex + reflex y ii tkek er a a c ii teek er a h ii t6ek cr a h ac n teek er a

'Ihese two principles together account I'or almost all of the data in 38-40. In 40, the Far Past, the Tense Marker is underlyingly High, and bcing the leftmost High, always wins out, In 38 and 39, when there are two object markcrs prcscnt, thc Two Object Marker rule puts a Iligh lone an thc sccond ol thcm, and this High tone is the leftmost and wins out. TIic Pullback rulc ncvcr applies in thc suhordinate forms; we return to the metrical structure of thesc forms in thc ncxt section. In any cvent, we must stipulate that the Pullback rule is limited to the focus constructions, The reflexive is inhercntly High toned and, except iti the Far Past, wins out, although for a reason that we will turn to directly in section 5.2 where we discuss long vowels, it is frequently realized on the lollowing syllable, the verb radical. Only if nonc of these other conditions is met docs thc High tone assigned by 4l actually surface. S.2 Moras and syllables, and more rhythmic structure. We have just noted that the reflexive object marker presents unusual tonal properties, in that its High tone appears on the following (stem) syllabe in the cxamples in 38-39. This added complexity derives from the fact that the reflexive marker is a long vowel. This can be determincd by comparing the rclativcly simple subordinate pattern found in shortvowel radicals with the pattern found in long-vowel radicals. In the short-vowel radicals, thc High tone characteristic of the subordinate (assigned by the suffix in 4l) consistently falls on the second syllabe of the stem, which is the same as the second mora. When the verb radical contains a long vowel, however, we find that this High tone lalls sometimes on the second syllahe (i.e,, the third mora of the stem), and somctimes on the second mora of the lirst syllabe. The conditioning is straightl'orward: the High tone falls on the second syllahle when there are no object markers prcscnt and on the second rnora of

44. Subordinate: Recent Past: long vowel No OM One OM Two OM y a Ieek hr a y a mu teek er a y a ki mu teek er a ha a teek 0r a ba s mu Ie6k er a bx a ki mu teek er a Rcflexivc Non-reflex + reflex y ii t6ek er a y a c il teek er a b ii t6ek er a ba a c i i teek er a
45. Subordiuutc: Far I'ast: long vowel No OM One OM Two OM y a tkek cr a y mu teeker a y a kl mu tcck er a ba k teek er a . h a II rnu teek er a ba 0 ki mu leek er a Reflexive Nou-rcflcx + reflex y il teek er a ya c u t eeker a b il teek er a ba II c li teek er a

Since ICirundi appears everywhere else to be essentially a mora-counting language in its Ionc assignments, wc will take the pattern in a mu iehk er a to be closer to the underlying form. That is, we assurae that the High tone suffix to attach the second mora. ln general, and whcn there are no object markers present, this High tone will he shifted frorn the sccond half of a longvowel to the follov ing syllabIe. We submit that this shift is parallel in certain respects to the Strong Pullback rule, shifting a High tonc I'rorn a Weak position to a Strong, as is illustrated in 46.
46.

a, No object marker

Onc object marker:

t ee k

s w h

s~w
er a
b. mu te ek

er a

LH

LH

MODf!LFS FN TONOLOGII:

THE KIRUNI>I VI:RB

Thc second mora High of the subordinatc shifts in case 46a, but not in 46b; we express this as in 47, with the Long Vowel Shift rulc.
47. Long Vowel Shift ([-focusable tenseI)

V I, A D D I T I O N A L

FORMS

The othc:r casc wherc a long vowel occurs, as we have already noted, is with the reflexivc object markcr -ii-, which also has a High tone on its second mora. As we have observed in 43 and 44, the rcflexive tnarker's lligh tonc appears on the following syllablc (the first mora 'I'his is of thc radical) whcn it is the only object marker in the vcrb. illustrated, with the associated rhythmic structure, in 4 8a; w hen anothcr marker precedes ii as in 48b, the High tone does not shtft. These properties follow directly from the gencral rule formulation given in 47, (The Reflcx i s hortening rule 37 does not apply in subordinate forms.)
48. s w s w w
ir s s mu i i rim ir a

sJ I

y ii rim

lt is perhaps worth emphasizing that the mctrical structure that we have postulated for use with the Pullback rule in focuscd forms and for usc in subordinale clauses is the same it is erected on the supraslem ol' the verb, the sequencc ol morphemcs beginning with object markers and containing thc stem.

6.l M ai n clause ncgativc forms (indicative). Most featurcs of the negativ'e main clause verb are easily analyzed along the lines of the ncgativc forms appearing with the focus markcr. "' Superficially, howcvcr, thc ncgative forms are somewhat puzzling. AII main clausc negativc verbs begin with the negative marker nti-, a morpheme which has two special characteristics that must he noted from the outsct. Thc tonc ol'the negative marker nti- is what Meeussen aptly calla an "unstable lligh", Like the copula ni, the tnne ol' thc nti- is High if it is preceded by another word in thc phrase; othcrwise, it is Low in tone. The evidence suggests that both ni- and nti- have a High tone undcrlyingly as well as at the level of the word phonology, Neither morphctne can ever, of course, he phrasc-ltnal; whcn phrase-initial, they arc Low in tone, Applying the notion of "extrapositionality" (of which extratonality is a special case) oncc again, we may wish to explorc the possibility that thesc "unstable High" morphemes are extratonal at the phrase level (recall Ihat the earlier use of extratonality involved an extratonal vowel at the word level ( cf. rule 18 above). When the lexical High tone ol' this morphemc docs not associate with this morpheme by virtue of this extrapositionality, a low toned vowcl is the rcsult. Thcrc is, to be sure, a striking contrast here witt> what occurs within the word-level phonology when Ihc word-initial moras a re rcndcrcd extratonal. As w e n oted abave in th e case of t h e dcmonstratives (17 above; see also 6.6 below for parallel cases with High toncd subject markers in the participial form), ihe tonc of the vowel that has been rendered extratonal is attached hy the automatic conventions to the following vowel. Nothing of thc sort occurs with the High tone of the "unstable Highs". We will suggest why this difference should be found in a moment, Investigations into cxtratonality (primarily in Pulleyblank 1983) and extrametricality (Hayes 1980, H arris 1983 Archangeli 1983) continue to provide support for the proposal that extrapositionality is withdrawn from an clement oncc it is no longer peripheral, i.e. when it is not at the extrcmc Icft or rigltt of the domain in question. However, the syrnmctry Ihat we must assume exists betwcen tona1 and
7. Tltis contrasts strikingly with Ihc situstion in Hsya, s closcly telated languages described by Hyman and Hyarushengo l98, where negative forms are inherently + F, in a related sense; Hyman (personal communicstion) suggests Ihat the similarity is somewhat deceptive, and thst Ihc two types of focus marking found in the two languages reprcsent typologically distinct categories of focus-marking.

46

hIOI>ALF.S FN TONOLOGIE

THL' KIRII NB I V F R II

47

non-tonal segments leaves open the possibility that in thc case of thc "unstable Iligh toned elements", it is not the vowel (or, correspondingly, the skeletal position) vhich is extrapositional but rather the High tonc. We will see strong evidence below (section 6,5) that this latter is indeed correct; for now, we will use this result, and analyze the unstable High as a spccially marked extrapositional High tone, a marking which renders it unassociable if it is peripherical at the phrase level. At the phrase lcvcl, v herc both nti- and its unstable High tonc may be preceded by material, the High is no longer extrapositional, and it is free to (and must) associate. Thus we find pairs as in 49.
49. a. Yohakni ntkhonfr. "John docs nol see" b. ntrrhnnfr. "Ile does not see."

Consider the relevant data in 50-52. We give lorms for Iligh tone verbs only (both short and long stcms) since thc tone on I ow tone verhn is identical.
50. Present Negative, Main Clause a. Short vowel slcm No ohject marker si m bon hr a nl u hon 6r a nl a bon hr a One object marker si n ki hon 6r a nti u ki hon 6r a nl a ki hon kr a Two object markers: 3rd person nt a ki mu bon cr a Reflexive object marker: 3rd person nli y ii h6n er a b. Long vowel stern (all 3rd pcrson forms): Nn O M: ni a leek hr a One OM: nl a mu rehk cr a Two Ohls '. nl a ki mu leek er a Reflex: nli y ii IEek er a

nti tu hon 6r a nti mu hon 4r a nli ba bon cr a

nli tu ki bon br a nti mu ki bon 6r a nli ba ki bon br a

The othcr characteristic ol nti- that must bc considered is the nature of the boundary separating it from the I'ollowing verb. It can be clearly shown that in the Present and the Far Past negativcs, it is separated from the stem by the equivalent of a word houndary, while in the Recent Past the naturc ol' the boundary cannot bc determined directly. As we shall sce in a moment, there are reasons to believe that thc nti- forms a phonological word with the following verb as if, so lo speak, only a simple morpheme boundary separated the two in thc Rccent Past. The evidence concerning the boundary between the nti- and what follows comes from vowel coalescence, in thc case of thc Present tcnsc, and trom tone in the case of the Far Past. We will consider these in turn. A sequence of vowels in contact across truc word-boundaries is realizcd as a single short vowel, always the second of thc two adjacent vowels, although a high tone that might have occurred on the llrst vowel is prcscrvcd on the remaining short vowel. When vowels about inside a word (but outside the stcm, f'or strictly within the stem wc find a diffcrent pattern, discussed below), thc resulting vowel is always long, As we have seen, the tonc of the resulting long vowel is falling if the first mora of the input had a High tone; it is rising if any later mora had a High tonc. The sequencc of nli- plus the subjcct marker a- in the present tense negative rcsults in a short vowel, clcar evidence that a wo rd boundary is prcsent (see 50). Similarly, the High tone on thc Far Past lerlse marker has shlfted to thc right in the singular forms in 52, exactly parallel to lhc alfirmative main clause and subordinatc Far Past forms. 1his results, as we havc already deterrnined, from the cxtratonality of the initial mora (via rule Ig, In i tial Extratonality); thus the subject marker must perceive a word-boundary to its lcft as well. Only in the Rccent Past is there no segmental evidence regarding the nature of the houndary hetween the negative nti- and the following subject marker.

nti ba ki mu hon cr a
nli h ii h6n cr a
nii ha leek f.r a nii ha rnu rehk er a nti ba ki mu leek er a nli b ii trrek cr a

51. Rccent Past Negalivc, Main Clause (3rd Pcrson only) a, Short vowcl stem No OM: nti y a hon er a nti ba a bon cr a Onc OM: nti y a mu bon er a nti ba a mu hon cr a Two OMs: nii y a ki mu bon er a n ii h a a ki mu bon er a Reflex: nlr y ii bon er a nti b ii bon er a h, Long vowel stem Nn O M: nti y a teek er a nti ba a lcek er a Onc Oh(: nli y a rnu teek er a nti ba a mu leek cr a Two OMs: nti y a ki mu leek cr a n l i b a a ki mu tcek er a Rcflcx OM: nii y ii leek cr a nli h ii tcek er a

52. Far Past Negative, Main Clause (3rd Persononly) a. Short vowel stem No O hf: nti y a bon er a ntr ba k hon cr a O nc OM : nti y a mu hnn er a rrlr ba 6 mrr hon cr a Two OM : nii y a k i m u hnn cr a nli ba 6 ki rnu hon er a R eflex ; nli y ii hnn er a nli b ii bon er a h. l.ong vowel stem No O M: nti y a icek er a One OM : nti y a r nrr lcek er a 'lwo Dhfs . ' nti y a ki mu tcek cr a Rcflex: nli y ii teek er a

MOD&I.I-:S I'.N I'ONOLOGIE

THF. KIRUNDI VI :IIII

The tonal structures following the negative prefix nti- in each casc look likc the subordinate forms of the affirmative. In a nalyzing those subordinate Iorms, we appcaled to thc principle that the leftmost High tone wins (42). The fact that there is independent segmental reason to analyze the ncgative verbs with a word boundary as in (53) makes it clcar that we can continue to maintain this principle within the word-level domain, Thus the High tone on the negative prefix in the Present and Far Past, being separated from the following word by a word houndary, does not cause lowcring of a following High tone on the verb. 5 3. a, Preseni: nti x S M + O M + S t em h . Far Past; nti x S M + a + O M + S ( e m In thc Recent Past negative, however, we find no High tones at all on the verb itself, As we shall see when we look at thc subordinate negative form, it i s n ecessary to posit a special morphologically conditioncd rule that lowers all High tones in negative Recent past forms, a rule given in 54."'
54.

characteristic ccrtainly does not prima facic providc an argument for syntactic autonomy for nti-. Regarding the second, we shall sce that in Kirundi (as in many Bantu languagcs) thc negativc marker follows thc subject marker in subordinate clauses; but there is no compclling, or even plausible, basis for a syntactic analysis of these differing negative forms, The two morphemes are phonologically quiic different (nti versus ta), and thc negativc markcr in thc subordinatc vcrb is clcarly a simple morphological prcfix (sec below), not a possiblc target for a syntactic rule of permutation with the subject markcr in the matrix clause (as one conceivable analysis might have it). 6.2 What do the non-focusing forms have in common? Lct us address briefly the question of what it may be that the main clause negative forms have in c o mmon with th e subordinate affirmative forms. Whatever the answer turns out to hc, it is something shared as well, as we shall see in a moment, with suhjunctives and imperatives. The main clause affirmative stands alone against the negative, the subordinate, the suhjunctivc, and the imperativc in thrcc ways: on ly in the main clause affirmativc is the focus marker possihlc (i.c., a scgmcntal focus/nnn-I'ocus contrast is found); only in the main clause affirmative are two High tones possible in a single verb (elsewhere the First High Wins principle reigns); and in the negative, the suhordinnte, the subjunctive, and the imperative, a High tonc is assigncd to the second mora, obliterating the lexical tone of the verb radical. Furthermore, for r easons that may b e v i cwed cithcr synchronically o r diachronically, the Pullback of the radical tone onto an Object Marker is found only in the main clause affirmative, a fact that has in certain rcspccts bccn accounted I'or in our formulation of the rule. Of Il of these observations, which is the most fundamental 7 We would suggest that the most important contrast involves f'ocus. Evcn in thc main clause affirmative, we note, it is true that only one High tone may appcar in thc vcrb is in a non-focus form. Comparc rule 26, which lowers all High tones in a non-fncus main clause, except a possihle High tonc on thc Far Past tcnsc marker. Again, as we shall observe later, in those special tenses where no I o cus contrast is possible, the first High Wins principle quite generally holds in each phonological word. We suggcst thcrcforc, that the crucial cnntrast is bctwccn (hc class of [+I iocusablc] and t-Focusablc] tenses. It is this I'eaiure which primarily distinguishes the Prcscnt, Rccent Past, and Far Past main clause affirmative tenses froin the other tenses. 6.3 Subordinate Negative (indicativc) The negative indicative I'orms f'ound in suborilinate clauses present several interesting characteristics. As is commonly found in Bantu languagcs, the negative rnarker in subordinate Iornis follows the subjcct markcr, rather than preceding it as wc f'ound in main clauses.

H~L/
Recent Past Negative

Thus it docs appcar that the First I<igh wins principle is acting across the entire word in the Negative Reccnt Past, but independent evidence, as we shall scc, suggests that S4 rather than that principle is responsihle for Ihc lowering of tone throughout thc Recent Past Negative paradigm. 'Ilic suggestion has been made that the negative prefix nti- is in some sense a syntactic clitic, i.e., a syntactically indcpcndent formative undergoing a post-lexical attachment to the following verh. We prcfcr the (in ccrtain respects) more straightforward morphological analysis in which nti- is a prelix in much the samc sense that the subject marker is a prefix, differing however in the type of houndary that separates it (or, in the terrninology of lexical phonology that we will use below in Section 6.8, differing in thc layer on which it is attached). There are, after all, only two things that make thc ncgativc prefix nti- odd : Iirst, the pholological boundary that separates i( froni the following subject marker, and second, thc Iact that it is used tn mark the ncgativc only in ma i n c l a uscs and n ot i n s u b ordinate clauses. Thc f irst

8. This rule is generalized in Kinyarv;anda in sll Recent past forms; cr. Furere
and Itiallaiid 1984, (or exsmple.

50

MOOf LES EN 'IONOLOeltE

' I'HF KI R U N D I V E R B

5I

This subordinate negative marker shows no sign of coming wilh strong "word-houndaries" around it. Consider the forms below. As we would expect, the lexical contrast between Low and Iligh tone radicals is lost in thcsc tenses.
55. Ncgative Subordinate: P resent Tense 8. Short stem -rim- "cultivate"
rt ta rim Ii tu ta rim rr u ta r i m 8 mu ta rim a 8 la rim k ba ta rim It h. Short stem -tem- "to cut" (begins with voicclcss consonant; shows effccts of Daht's Law) rr ta tcm k tu da lem tt u da tcm 8 mu da tcm 6 a datem k ba da tcm tt One objcet rnarkcr n ta gi tcm 6 tu ta gi lcm k u la gi tem 0 mu ta gi tem 6 a ta gi tern 6 ba ta gi lcm k Two objccl markers n ta ki m r'i icm cr 8 tu ta ki mu tem er a, etc.

56. Recent Past negative, subordinatc

a. Short sicm
fl ta i l t C rll 8

lu ta a tem 8
lllu la a 't cnl 8

U la i l a tc m a

aia a

tem 8

ba ta a tem a
tu ta 8 gl t em 8, e tc.

One object marker n ta a gi tem a Two object markers n ta a ki rnu tem er a Rellexive n t i i te m c r a

tu ta a ki mu tem er 8, ctc.
tu t ii tem er 8

In the Far Past, one and only one High tone surfaces, that of the Far Past tense marker. Bear in mind that this is the source ol the High tone in n t ii tem a (from n ta a ii tc m a).
57. Far PaSt negatrve. Suburdlnate

No object markers n ta t i t e m a
Orlc object marker n ta 6 gi t crir 8

tu ta Ir tcrn a
tu lil k gl t C J ll 8

Rcflcxivc OM
n t ii t hrrr cr 8

tu t ii thm er 8

b. I.oirg sicm -raab- "look at"

Two object markers n ta 6 ki mu lc m er 8 Rcll c xi vc

n I lr t c l n 8

tu ta h ki m u tc m er 8

tu I ii tcm a

n ia raah k tu ta raab 6 u t raab 8 mu ta raab h a ta raab 4 ba ta raab 0 One Objcct Marker n ta ki rahb 8 tu ta ki raab a, etc. Two Object Markers n ia ki m u r aab ir 8 tu ta ki mu raab ir 8

We scc in 55 t hat th e negative marker -ta- (called by M e cussen a "post-initial") has no lligh tone, and that lhe tone pattern of the entire w ord follows I'rom the pri ncip]cs we have established so far. Th e voicing of - ta- to - da- i s th e result of D a h l's Law, w hich voices a voiceless obstruent in 8 morpheme immediately preceding a verb stem that begins with a voiceless obstruent. In K i r undi, unlike a numbcr of other Northeastern Barltu languages, only a stem-initial voiccless consonant triggers Ihe elTect ol Dahl's Law. Wc will return helow to lhc thcoretical signiflcancc of the failure nl the rule to apply when a nasal (1st sg. subject inarkcr) prccedcs. The Rcccnt Past negative is everywhere Low in tone, as can be sccn in 56. This clearly motivates the general Lowering rule given above in 54, (he rule that lowers all Hig>h tones in Recent Past negative verhs.

6.4 Subjunctive 6 .4.I AITirmnaiie. The alTirtnativc subjunctive form i s s i m ple a n d s traightforward, and virtually identical to the subordinate form in i t s t onal pattern. Th c F i na l V o w el, a s w e h a v c noted, is - e in t h e subjunctivc, and there is an odd irregularity in the Ist person singular form of th e v crb w hcn a c o nsonant-initial object marker (i,e., any object markcr other than the reflexive) lollows, In all such cases, the m arker -ra- (surfacing as -da- al'ter a n asal) appears hetween th e subjecl marker n- and the object markcr. See 58.
58. Subjunctive
tu rim c mu rirn 6 a rim s ba rim 5 One objcct rnarker (-ha-, loeativc) n da ha rim b tu ha rim 6 u ha rim 0 rnu ha rim 0 8 ha fr fll ba ha rim C
ir frlll

a. Short stem (-rim-) ndim 6

Two object markers (-ha.mu-) n da ha m6 rim ir e tu ha mu rim ir e, etc. Rcllcxivc n ii rim ire lw ii rlm ir c

52

MODELES EN TONOLOG!E

Tlt E K1 RUNBI VERlt

53

Object rnarker plus reflexlve n da h ii rim ir e tu h i i

r im i r e

6.4.2 Negative subjunctive . Th e N egative Subjunctive introduces a point nf some interest concerning the nature of the tone-bearing unit i n Kirundi. In the contrast between the forms si n dim e and si n i i rim ir c, for example, we see that the 1st singular marker n- is tone hearing in the l i rst case but not the second, in each casc appearing after the negative marker which places a High tone on the following mora or tone-hearing unit.
59. Negative Subjunctive (stem -rim-) si 'n dim e nti tu tim e nt u tim e ntr mu rrm e nti bh rim e nt fr rim e One object marker si'nda ha rim a nti tu ha rim e nt u ha rim e nti mu ha rim e nt 6 ha rim e nti bfr ha rim e Two object markers (ist person only) si nda ha ki ri m e nti ru ha ki rim e Ref!exive (1st person only) si n ii rim e n ti tw rl n m e

a verb stem. If t h cre arc no object markers, the Final Vowel is t h e n eutral aspect -a ; o t herwise, it t a kes thc F i nal V o wel - e . ' Ihis i s illustrated in 60.
60. Imperative a . kor a "wo r k ! " b . mu kor er e "work for him ! "

There is never a focus marker in the Imperative, and in gcneral the tone pattern of the Impcrative is that of a non-focuscd vcrb forrn. In the examples seen in 60, there are no High tones, apparently: howcver, the first vowel of t h e i m perative is assigned by gencral rule, w h at Mecussen calls an "unstable High tone", discussed earlicr in Section 6.1. As noted earlier, this unstable High tone has the characteristic that i t i s p r o n o u nced o n l y w h c n i t i s n ot ph r a se-initial; thus phrase-initial impcrativcs are all Low, as in 60, but whcn another word precedes, as in 6l , t h c f i rst (or second, in the casc of vowel-initial stcms) syllable of the imperative verb is Il i gh.
6 1. a. umugorb muraabe '! "look at the woman ! " b . Ynhartni r~aaba! "Jotrn, took ! " c, Yohahni andika! "3ohn, write1" d . Andika "write ! "

We sce immediately that all of' these forms begin with the melody LH (imposed, clearly, by the negative prefix), and this High tone created or the second mora eliminatcs all following High tones by the First lligh Wins rule. Ihe contrast in the tone-bearing status of the marker n- in the examplcs si n dime and si n ii rim ir e shows that the rules of syllable structure assignment must apply before tone assignment. The difference in the syllabic status of the -n- cannot be due to tonal c onsidcrations, since tone i s essentially kept f i xed across the t w o exarnples; hence the difl'erence in the syllabic structure of the material following the n- m ust be the cause of the d i l ference in its syllabic s tatus. When a c o n sonant f o l l ows, the n - c a n not b e p a r t o f t h e following syllabe (recall that these conclusions are, strictly speaking, valid only during the lexical phonology; permissible syllahle structure at the post-lexical, phrasc levcl may bc (must be) morc libcral). It thcn f orms either part o f t h e r h yme o f t h e p receding syllabe, if t hat i s possiblc (as it is in si n dime) or is marked as a syllable nucleus itself. Irr either case it i s p o sitionally tone-bearing, and subscquent tone mapping will assign the High tone from the negative prefix to the nprefix in the appropriate cases. 6.5 Imperatives and extratonality.The Irnperative in Kirundi is simple in structure, Iooking like the unit we have called the suprastem; that is, it bcgins optionally with one or morc object markers, lollowed by

We noted above that the imperative form provides us with evidence that the unstable High tone is an extrapositionally marked High tone. 'I'he evidence that we interpret in this way derives from the behavior of imperative forms of vowel-initial stems. We notcd earlier that there is a word-level rule marking word-initial vowels as extratonal. We saw there that a High tone that would otherwise be assigned to the f i rst vowel of a word will be assigned to the second vowel when that first vowel has been marked as extratonal, Indeed, a vowel-initial im p er ative that is phrase-medial will have its unstablc High tone on t h e s econd vowel, as in til c . 13ut if o n e o f t h e se v owel-initial i m p eratives should a p p ear phrase-initially, as in 6ld, thc unstable High disappears, a result that would bc incxplicable if it werc a special property of the initial vowel o r syllablc that were playing the central role in the phenomenon A l l ol' these facts are neatly accounted for, i n s um, i f w e a n alyze the unstable Iligh tone as one which is marked as being extrapositional a t the phrase level, which means that when it i s p c r i pheral at t h e phrase level (i.e., phrase-initial), it cannot associate with any vowel. When it i s n o t p h r ase-initial, i t w i l l a s sociate w it h t h e I e f t m ost

54

M()Df LES EN I D N OLOGIE

TIIL KJIIIJNDI VE II II

avaiiah'le vowel within its word, which will hc cither the first vowel, or, in the case of a vowel-initial word, the second vowel. [Fn.9] 6.6 ka, racaa, and participial forms. In this section we shall bricfiy sketch the relevant facts concerning three other forms. The -kaconsecutive tcnsc behaves tonally quite like the Present focus forms, with the Strong Pullback rule applying (see 62), The racaa perstitive tense is straighforward, though the Strong Pullback rule does not apply here, The Tense Marker is presumably composed ol -ra- plus -ki- plus eithcr -a- or -aa-, See 63.
62, -ka- tense. a. Low tone verb No object markers
nka som a

The participial form prescnts a fcsv interesting characteristics. First, subjcct markers are assigned a High Tone, rather than a Low tnne; thus, as we see in 64, a short Subject Marker such as a- will rejcct this High tone (by word-level cxtratonality), but a Subject Marker of the form CV will maintain it, 64. Participial: main clause
a s601 a

bli som a

a mu som er a bk mu s o mcr

a t6 n la b t It crn a

mu tcm cr a b6 mu Iem er a

The participial forms are obligatorily marked [-Focus] in the Present Tense (thus not h aving the F ocus M a rker, and u n dergoing thc N on-Focus rulc); i n I h c R c ccnt an d F a r P a st, th e c ontrast i s maintained between the [+Focus) and the [-Focus] forms, 65. Participial: Far I'ast (Focus) y ara som a yra mu somcr a
hara som ba

tu ka som a

One object marker

n ka gi sorrl a

tu ka gi som a 'Iwo ohject markers tu ka ki mu s6m cr a

bara mu som cr a

n ka ki mu s6m er a Kcflcx ivc

n k ii som cr a
n ka c I i snm er a

tu k ii som cr Object Marker + Rcflcxive tu ka c ii som er a

b. High tone verb

No Object Markcrs
n ka rknl a tu ka t6m a

Two Object Markers

tu ka gi tcm a Two Object Markers n ka ki mu t6m er a tu ka ki mu t6m cr Reflexive Marker tu k II tcnl cf a nk ii tcm er Object Marker + Rcflexivc n ki rnw ii itnl er a tu ki mw ii tem cr a n ka gi tem a 63, .rach- tense, No Object Marir.er
n da ca Ibm
tu r CII t6lrl a

Two Ohjcct Markcrs tu ra cAa ki mu tem cr n da cka ki mu tem er a Reflcxive tu ra c ii tem er a n d c ii tem er a Object Marker + Reflexive tu r cka c ii tem er a n da ch c ii tem cr a
9. We may remark on Ihe evident historical origin of the process Ihr has been grammalicalized as word-initial exlratonalily: il is the process of vowel-fusion at word-boundary rhat put a High Ione from the end of one word on the vowel of the following word. This process has nnw been, as we note, grammaticalized in the
word-level phenology itself,

6.7 Future tense and the organization of the morphology. The intcrnal morphological structure of the Future tense is quite different lrom thal of the Prcscnt or Past tcnscs. As we havc secn, these tenses are formed I'rom the suprastem hy the preflxation of subject ltrarkcrs, tcnsc and sometimes I'ocus markers. This prcfixation crcates a unit which is a word, and contains no words internaily, if wc judgc by the operation of the First High Wins rule. These are elementary observations, but important ones in the light of' the behavior of the Future tense, where nonc of these holds. We did note i n th c p reccding section that the main clause negative prefix nti - b ehaves like a prefix separated by a stronger word-boundary, What follows thc nti-, v c should emphasize, has the appearance of an cntire word, and can, indeed, stand as a word; it has a subject marker, a tense marker, and a vcrbal. We will see in this section Ihat thc futurc marker -zoo- is like ntiin attaching to a unit larger than the Base, even though tlris is not apparent in rnost cases, Wc can makc Kirundi morphology clearer w ith a d i agrarn like the f'ollowing Table I , a l a y ered model o f rnorphology reminisccnt of lexical phonology. Level l c r eates the derivational hase, cnnsisting of th e r adical und th e d erivational extensions. Srubscquently the Final Vowel, the Object Markcr prellxes, the Tense/Focus markcrs, the negative marker, and the Suhject Markers are added. Each ol thc two layers has distinct phonoiogical rules. Layer 2, for cxample, contains Dahl's Law, which is triggcrcd in Kirundi only by a radical-initial voiceless consonant (not, c.g., an Object Marker beginning with a voiceless consonant); see 66. 66. Dahl's Law [c l ~ ] + v oice] / - v +
-vnlcc

[ c] -voice

56

M O D N L ES LN TON O L O G I E

T HE KIRU ND I V E R B

57

Thc operation of t h is rule shows that D ahl's Law cannot apply cyclically with the addilion of each prefix, since it is blocked from applying to the subordinaie negativc -ta- just in casc a nasal Subject Marker prefix precedes it (cf. 55b ahove). This is a serious issue for the version ol Lcxical Phonology found, e.g., in Kiparsky 1982. We assume that a stem is endowed with inflectional fcatures, and that it can bc viewed as passing through the rules of the intlectional morphology, realizing each of the specified inflectional features. Thc stcm may be assigned the category Noun or Verb; if it is assigned the category Noun, it will undergo noun class prefixation (class l5) to hecomc an infinitive; thc fact that Dahl's Law affects noun class prclixes is one clue that noun class prelixcs are indeed added in Layer 2 (aiid thus onc cannot make a strict equation between Layer 2 and pure inflectional morphology). Many of thc other prelixation proccsses (prefixation of Subject Marker, for cxample) are restricted to verbs. ln the third layer, a h igh-level word can be created by thc attachment ol an outer layer of prcfixes, including nti- and, we will suggest, -zoo-, ihc f u ture inarker. This conclusion seems largely incscapable, though it leads to further questions that we will takc up bclow concerning thc mechanism by which Subjcct Markers can be attachcd Io the left of zoo. We shall now show why thc tuture marker -zbo- must be treated in this special way. Since the lexical contrast between High and Low toned radicals is neutralized in this tense, wc present only one example of each form. As abovc, we give 3rd person singular and plural forms.
67. Futurc iense, rnain clause affirmativc. No OM ' a z oo bon a ha zoo bon a I OM : a z o o mu bon a ba zoo mu bon a ha zoo ki mu hon er a 2 OM . 'a zo o ki mu bon cr a ba zoo kw ii bon a Reflex: a zoo kw ii bon a

Layer I
Radical

extension-suffixaiion

Base

Final Vowel

Layer 2
Rellexive OM

Object Markers
Tense-Focus (V)
Suhordinate negative

6S. Future tense, main clause negative No Om : nt(il a zbo hbn a I OM ; nt( i ) a z6o mu bbn a 2 Om: nt(i ) a z 6o ki mu bon er a Rellex : n t ( i ) a z6o kw ii hon a

ku-prefixation (N)
nti ba zbo b6n a nii ha zbo mu hon a n i i h a zbo ki mu honer a nii ba zbo kw li bon a

Subject Marker (Y)

69, Fuiure tense, suhordinate aflirmative No OM ; a z6o b6n a ha zbn hbn a ha z6o mu b6n a I Oht: a zb o m u b bn a 2 OM ; a zbo k i m u h o n er a ba zoo ki mu bon cr a

WORD

Layer 3 nii.prcfixaiion (V)

z6o.prcfixation (N)

Reflex . a zbn kw il bon a

ba zbo kw ii bon a

A number of characteristics distinguish this tense from all the others we have lookcd at so I'ar, We will discuss some of ihe tonal properties of thcse briefly and then return to the proposition that the future

Mnnf:.I.ES EN TON()t.nt-lE

TIIE KI RU ND I V E RB

59

marker -z6o- is a L a y er 3 p r efix, unlike the other tense markers. Notice, first, that in thc subordinatc forrn, there is a High tone on both t hc tcnsc marker and o n t h e stem; thus here the First H ig h W i n s principle is superficially violated. This is the result, as we have hinted, of the affixation of z6o at a level following the level at which the First Iligh Wins principle applies. The High tone that appears on the first syllable of the stem is morphologically conditioned; both H igh and Low stems bcar that High tone in the Future : We must therefore have a Layer 3 rule that assigns High tone to the stem in the Futurc, a rule that could be made part and parcel of the z*o prefixation procedure. It is worth noting that such a procedure would, however, violate the Bracket Erasure Convention of Ki parsky (1982) and others, since the rule would have to know where the stem bcgins. 7 0. [C o V I
H ~ I n s ert

The second indication that zoo is attachcd tn an in f initive concerne t he Final Vowel. We remarked in Section 2 abovc that in all o t h e r tcnses, the Final Vowel may be either the perfective -ye or the neutral -a, semantics permitting. In the Future, we find only the Final Vowel -a, and never -ye; thus we have a-zoo-bon-a, hut no ia-zoo-bon-ye, This follows immediately if the future is based on the infinitive, which always has the Final Vowel -a. The third argument involves a small number of stems which have no inflnitival form, such as Vzi 'know' (cf. ' k u -uzi, 'azooki, etc,) and - ri, one o f t h e c o p u las, used, fo r e x ample, i n t h e m ost c o mm o n continuous tense, as in 74. 74. ba ri k6 ba ra nm SM copula complementizer SM F O C c u l tivate lit., "they are that they cultivate"; i.e.,
"thev are cultivating"

a FV

Condition Future tense

stem I

In the main clause affirmative form, the Iligh tones assigned by rule 70 and the underlying High tone of thc markcr -z*o- are lost, by another, later I.evel 3 rule, 71.
7I. H L / M ai n c lause Future

If there is no infinitive, we would predict that there is equally no future tense for such radicals T his i s corrcct, and t o f or m th e f u ture continuous, another copular verb is used, a radical (-b) which has an infinitivc kuba) The fourth argument is based on the fact that in addition to the simple infinitive as in 8, there is also a future inlinitivc ol' thc form ku-z6o(ku)-suprastem, (e.g. ku-zoo-rima, 'to be going to cultivate'). No other tense or tense marker has such a special infinitive, and we would expect none such, The only grounds on which we would expect a future infinitive is precisely if -zoo- is a verb, in which case we would expcct it to have an infinitivc and thcse are precisely to so-callcd future infinitives. We arc thus lcd to thc conclusion that -z6o- is radical that takes infinitival complements, and phonological evidence suggests as well that it is added on a l ater layer; (Scc Sadock 1984 for a discussion o f autolexical syntax, a f r a mework i n w h i c h s uc h c ases can b e i nsightfully d ealt w i th ; se e also L i eber 198I f o r a d i s cussion o f category/subcategorization frames I'or morphemes,) We may sumtnarize the properties of' -z6o-, and of the main clause negative nti- as well, as in 75. 75. a. zoo [ V/N N f nti[ ) V V [

Let us rcturn now to the placement of -z6o- in Layer 3. Our argumcnt is, in essence, that the futurc is formed by attaching z6o not to the verbal, hut to the infinitive; that this.ipso facto distinguishcs it from all other tense markcrs, which attach to verbals; and that it attaches, like the other Layer 3 prefix, to an cxisting word, both from a tonal and a lexical point of view, If z6o attaches to infinitives, as we suggest, one would cxpect to find the infinitivc markcr ku- inside the future verb. The careful reader will have notcd that we f in d p recisely that when a reflexive object marker is present. In general the ku- prefix is indeed found inside the future verb whenever the verbal begins with a vowel, a circumstance t hat will arise eithcr if a reflexive object marker is present, or if t hc verb stcm begins with a vowel and there are no object markers. This second case is illustrated in 72.
72. a. stem: an dik -w rite" b. n zoo kw andik a " I w ill write" c. n zoo mw andik ir a "I w ill v rite for him"

1]

b,

To put the matter the other way around, the infinitival prefix ku- is indeed present in thc futurc verb, but is deleted when followed by a consonant in Layer 3, by rule 73.
7 3. ku o / zoo C

We see in 75a that -zoo-, like any " verbal" r adical, can be marked either as a noun or a verb, and the choice of one feature or the other will determine which inflectional prefixes are attached to the stem, No such choice is available with nti-.

MODF.t.F.S F.lsl 'I'OhtOLOGIE

T KF K l R t INIJI V F R R

61

Following this prefixation, it must be possible for the "routing" of the morphology to go hack up to the l.ayer Two morphology, up a i least three rulcs' worth (as i s shown in T a ble I ) , s i ncc the subordinate Future form inay have the subordinate negative marker -ia- as well as either the Subject Marker or the nominal ku- prefix. The precise nature of this non-linear routing in the morphology remains an open question, and one whose resolution we hope this inquiry will stimulate. Wc should observe that it d oes not seem possible to maintain tlie straightforward proposal that the output of Layer 3 might be able to loop backwards (using a term reminiscent of Mohanan's 19$2 proposal) only by going all the way back through the Layer 2 morphology. We note that no object markers, for example, or second Final vowel, may be added on the second time through the Layer 2 morphology, and wc also note that it i s possible to propose (an othcrwise plausible sounding) principle to block rcdundant double spelling-out of inorphological teatures, since it is in fact possible to find verhal forms with two copies of tlie Subject Marker, each one having beer) spelled out on a pass througli the I.ayer 2 morphology (this is found in the Prescnt Continous ncgative in Killunde, another Group J language).

VI I. HISTORICAL- COM P A R A T 1VE OBSERVATION S

In this final section, we would likc to d iscuss briefly three observations that bear on the analyses that we have developed in this paper. In each case, similarities and diffcrcnces in nearby Bantu languagcs appear to shed some light on the phenomena that we have described in Kirundi. 7.1 Focus Markers, As we observed above, the Kirundi verb has a Focus Marker position immediately following in the Tense Marker. In the Prescnt and Far Past tenscs, this Focus Marker is -ra-; in the Recent Past, it is -a-, While it is not unusual for Bantu languages to have focus systcms with parallcl functions, it is unusual to havc an overt I'ocus Marker inside the verb, Where did it come from, and why is there such allomorphy 7 If' we look at thc formal mcans by focus systems in other Bantu languages, we generally find two kinds (which can, to be sure, coexist),

thc typc of phonological boundary present between the verh and its I'ollowing object, In general, the stronger boundary (word-boundary) expresses focus on the verb, and phonological merger of the verb and its object expresses focus on a later constituent or on the VP as a whole. If we further recognize that -ra- was originally the Present Tensc marker, then we may inf'er that the focus/non-focus contrast was originally expressed in a pre-Kirundi stage segmentally only in the Present Tense (similar to Zulu, though in Zulu a contrast is exprcsscd in the Past Tense by contrasting the Final Vowels -ile [verb focus[ and -e [non-focus], When the Present Tense marker was omitted, focus was off the verb, and -ra- was naturally reanalyzed as a Focus Marker. The extension of the use of -ra- as a Focus Marker to the Far Past, as we find today, was then only natural. ra- is not used, however, in the Recent Past, where a long vowel -aa- (verb focus) contrasts with a short vowel -a- (non-focus). This may well be a reflection of an earlier state ol affairs, L, Hyman (personal communication) has pointed out that somc of thc odd accentual behavior of the Tonga Recent Past rnarker -a-, presently a short vowel, can rcceivc a historical explanation il -a- were originally a long vowel with a High tonc on the Cirst mora; this oddity, crucially, disappears in thc non-focus form, suggesting that at an earlier stage the focusl non-focus contrast aITected vowel length of the Recent Past marker cf. Goldsmith l984, in press a). This reconstructed phonetic structure in just what is found in Kirundi today. 7.2 Two Object Markers, Eastern Bantu languages vary considerably in their tolerance for multiple Object Markers. Some, like KiHundc, permit only animate Object Markers (while Killunde even disallorns thc third pcrson plural Object Markcr). In many othcr languages, such as Swahili, only one Object Marker is permitted; still others permit two, as in CiRuri: and linally, therc are languages like a KiRuridi where even more Object Markers are possible under conditions of extrcme syntactic durcss. Our knowledge of the evolution of these structures is not suflicien to determine which type most closely represents the earlier stage of the languages. It is therefore interesting to note that in more than nne language, special tone rules comc into thc picture whcn two or more Objoct Markers are present, In Cikuri, for exarnple, all Object Markers are High in tone, and when two precede the verb stem, a special rule steps in and places a High tone on the Final Vowel (cl; Massamha 1982, 1984, Goldsmith l982). But this is surely not the original state of the rule. There is in the language a later rule (restricted in the language today to the Present Continuous and the Subjunctive, it is true) that copies any High tone on the Final Vowel back io the first vowel of

ln Zulu and Tonga, for example, we find a verb focus/non-focus

contrast expressed in the present tense by the presence or absence of the Tense Marker (ya and ia, rcspectively). In such languages as Tonga and LuGanda, we find a difference in focus expressed by changing

MODE,I.Fg EN T()NO(.OCIF.

. It is highly likely that this rule is in part an inversion he verb oC the earlier situation, a stage at which a high tone was specia y plac(ed oa tbe first vowel of the stem whcn two Objcct Markers led Ihes lem, preciscly as in Kirundi. This rule, in turn, can be pfuca(( undcrstood as the result of the result of Meeussen's Rule shifting from a poat~ to a cyclic rule, in fact, When Meeusscn's Rule operated lss leAco-right manner (as Odden l98l suggcsts, for example, for $bola) on two High toned Objcct Markers in front of a High toned ates, il would lcave behind a rhythmic High-I~w-High result. When Maeslsea's Rule became lexical (cf. Goldsmith in press b), it may have clically, appoed cyc lca and thus it would have applied outward from the radical, appearing to apply from r ight to l et t across the Object Markers. A rule inserting a high tonc on a stem preceded by two Object Markers would then reinstate the pre-change surface facts. 73 Long vowels. Thc treatment of vowels in KiRundi is especially latrjcate, as we have seen. We have appealed to metrical structure in scveral places to account for this bchavior. It is therefore of some ' t rest to notc that in Shi (Polak-Bynnn l97S), another Group J f. alls language, certain similar effects can hc found. When a High tone on Ihe second mora of a long radical vowcl (though in Shi this necd not be in a subordinate or ncgative vcrb form), it is shiftcd to the following syllable, just as in KiRundi. In Shi, however, the conditioning factor is diffcrent; what must occur for this shift to apply is not a metrical condition, but rather a high toned Object marker or Tense Marker immediately before the stem. If the Shi system and the KiRundi'rule are indecd descendents of Ihe same principle, thcn it is likely that the Shi formulation is more conservativc and that it may enable us to gain some insight into the origin ol' the rule. In KiRundi, we have suggested that the shift of the High vowel from the second mora of a l ong vowel occurs under conditions of mctrical weakness, given a metrical structure on thc suprastem. In S hi , t h e " w eak p osition" ol the stem i s t o nally determined; the stem vowel is weak, and thus incapable of bearing a Rising tone, if and only if it is preccded by a High tone. This is highly reminiscent of the situation in e arly G ermanic, as dcscribed by Ilollifield l983 : - I n S l evers' Law conditioning, the second syllablc i n appears to count only a s a single rnora (lamhic shortening), although it prescrves the opposition of long and short vowels." In conclusion, we hope that further comparative work of the type in this section will aid us in arriving at a deeper understanding ol Ihe modern-day gramlnar of KiRundi and the other Bantu languages.

TONS ET ALLOT O N S A PRO PO S DU N OM INAL RW A N D A IS

Francis 30UANNt;f

Le kinyarwanda et lc kirundi sont des langues dans lesquelles seuls les elhments vocaliques sont susccptibles de porter des unitEs suprasegmentales. I.".es EIEments vocaliques snnt dc deux types, brefs ou longs, ainsi quc I'attestent les exemples suivants: I g u haaga "Etre rassasiE" (I) guhaga "gonllcr" / g u sba s abscntcr gusiba "rayer" / g u suura "rcndre visite" gusura "pEter" kongera "rccommcncer" / kongeera "augmcnler"
gulola "etre mouillE"
/ g u t oota "harcclcr"

Lcs EIEments vocaliqucs longs se dEcomposent en deux EIEmcnts. De nature homophonc, ils sont analysEs commc une succession de deux unitEs brEves appclEcs morcs car chacune d'entre cllcs peut etre support dc ton ainsi que le montrent les exemples suivants; (2) guhiigk "chasser" (haut-bas); umushambi "grue couronnEe"(bashaut) Deux mores successives appartienncnt toujours A la mEme syllabe: -/v + v/- et non -/v/- + -/v/-; soit un phonEme (/v:/) dEcomposE en deux mores (v+ v) appartenant A une meme syllabe (- -) ct non dcux phnnEmes appartenant k dcux syllabes diffErentcs. La segmentation syllahiquc obEit h la rEgle suivante: (3) Tracer une frontiEre syllabique aprEs chaque centre de syllabe que cclui-ci soit gEminE ou non, Le mot "natte" se scgmentera dc la fa(;on suivantc 'u-mi(-shh-mbi (umushhmbi)"'
(I) L'iax(ruc(ion mirds(&rielle du gouveruemeol rwandais du 2 juillet (985 (na I3. (Fa/03. 02/003) ponan( xur Ia Iixa(ion de I'orthographe oWicielle du kinyarwanda cnnsacre deux uxagcs, I'un pour I'orthographe ordinaire qui n'ulilixe oi Ia notation dex (nns ni eelle des Iongueurx vnealiquex (ar(ieie 2) e( I'autrc pour I'uxage xoienlifique: "Article 24: Dans I'uxagc scien(ifique (daox I'enxeignemeo( el Ia recherche) la (onali(E ex( reprEsen(Ee par I'aceenl cirennAexe ( ) pour Ic (on haut et par I'abxenoa

de signe sur le lnn hax. I.a quan(i(E vocalique sl represen(ee par uue xeule voyelle sur Iex xyllabes brEves el par le redoublemenl de la voyelle xur Ies syllabes Ionguex."
Nous conservor(s ees conventions que nous avonx loujourx u(ilisEes a I'cxcep(ion du ton haut que nous no(nnx par I'aecen( aigu (') en accord avec lex normex in(erna(iona-

Ies.