Zambezia (1980), VIII (ii).NDEBELE AND ZULU: SOME PHONETIC AND TONALCOMPARISONSD. K. RYCROFTSchool of Oriental and African Studies, University of London1. INTRODUCTIONSince it was the Ndebele who asserted their independence in 1822 and broke awayfrom the Zulu, and not the other way round, one tends to take it for granted thatlinguistic divergence has been unilateral, Ndebele having changed, while Zulu hasnot But this may not be altogether true.The scope of this article is confined to pronunciation. No attention will be given tolexical or grammatical divergence, which is an absorbing topic in its own right, as isevident from the examples presented in an earlier paper by Fortune.1Findings presented here are rather general and tentative, since opinions differ as towhat constitutes the norm, both in Ndebele and in Zulu, as no systematic dialectresearch has yet been completed. Future investigations in that direction may wellexpose shortcomings in this brief article, but it may still serve a purpose, even if only inprovoking argument and refutation.1.1. Statements regarding Ndebele pronunciation are here based on sporadic andintermittent encounters with Ndebele-speaking informants and friends from variousareas over a number of years. Findings so far have not brought much evidence ofregional variation among Ndebele speakers, but rather of slight differences whichperhaps relate not so much to area as to ageŠ'older generation' or 'senior' speakers(say over 40 years) as against more junior ones.21.2. The existence of a 'standard form' of Zulu is nowadays generally taken forgranted, the so-called Natal dialect usually being implied. The Zululand dialect,spoken north of the Tugela River, does not differ greatly from this, but there are a fewsmall tonal differences.3 This needs to be borne in mind in our present discussion,because Mzilikazi came from Zululand, not from south of the Tugela.1.3. Modern Ndebele and modern Zulu are in many respects so similar that it isplausible (though probably somewhat simplistic) to postulate that they both derivefrom a single 'proto-Zulu' parent language spoken in Zululand before 1822. Absenceof reliable linguistic evidence from that period is of course a setback to reconstruction,1 G. Fortune, 'Differences between Zulu and Ndebele' (Bulawayo Teachers' Conference, 1964;copy kept in Salisbury, Univ. of Zimbabwe, Dep. of African Languages).2 The present findings are based largely on speech samples provided by the Revd Joshua Danisa(interviewed in London throughout 1961; born at Essexvale, descendant of Nsindwane Danisa who leftZululand with Mzilikazi in 1822; Mr Richard Mpande, of the Matopo Mission (tape by G. Fortune, 1958,kindly made available to me); Mr E. L. Bulle, from near Plumtree (interviews, 1970-7, at the University ofRhodesia); Mr S. J. Mhlabi, from Ntabazinduna area (interviews, 1975-7, at the University ofRhodesia) ;Mr S. Nondo, from Zimnyama area( interviews, 1970-80, at the University of London); and Mr.C. Sileya, from Ntabazinduna, later Essexvale area (1978-9, at the University of Rhodesia).1 Most published works on Zulu are based on the N atal dialect, but the Zululanddialect is covered inD. Rycroft and A. B. Ngcobo, Say It in Zulu | manual and tapes| (Univ. of London, School of Oriental andAfrican Studies, 1979).109110 NDEBELE AND ZULU: COMPARISONSbut valuable clues to pre-Shakan Zulu could be gleaned by examining what modernNdebele and modern Zulu have in common. Such an exercise is not pursued in anydepth in this paper, but it might be a fruitful area for future research.41.4. We shall here mainly be concerned, not with similarities, but with differences,regarding pronunciation, and we shall examine a sampling of synchronic evidence inthat connection. This is an essential first stage, before one proceeds to findexplanations for such differences, and ventures to attempt serious historical recon-struction. Occasional working hypotheses below, in the way of explanations forchanges, and speculative reconstruction, are premature and tentative. Far morespadework needs to be done regarding differences; and for explanation, rigorousattention would need to be paid to outside influences on Ndebele during the past 158years, besides influences on Zulu (including possible effects from orthographicconventions); but such matters are far beyond the scope of the present exploratorysketch.2. SEGMENTAL PHONETIC DIFFERENCES BETWEEN NDEBELEAND ZULU2.1. Vowels Regarding vowels there appears to be little or no observabledifference except perhaps in the following respects:(a) The quality of e and o in Ndebele seems to be less affected by thefollowing vowel than in Zulu.5 They appear to match cardinal vowels 3and6,[e] and [o], fairly closely, rather than numbers 2 and 7, [e] and[o],in most environments.(b) Word-finally, however, there is a tendency among some speakers6 torender e and o with closer articulation, as [e] and [o] or even closer,especially when preceded by a fully close front or back vowel, i or u.Also the Ndebele cognates for certain Zulu nouns with final -e (e.g. ithole, 'calf) seemto have optional variants with final -/; but how extensive this is has not yet beenestablished.4 Very little phonetic divergence between the two languages is recorded in C. M. Doke, The SouthernBantu Languages (London, Oxford Univ. Press for International African Institute, 1954), 94-118, or inL. W. L»nham, 'The Comparative Phonology of Ngunf (Johannesburg, Univ. of Witwatersrand, unpubLPh.D. thesis, 1960). Doke in his Bantu: Modern Grammatical Phonetical and Lexicographical Studiessince 1860 (London, International African Institute, 1945), 82, even expressed the view that: 'It is veryquestionable whether separate linguistic and literary work should be continued in a dialect so little differentfrom Zulu'. Nevertheless see W. A. Elliott, Notes for a Sindebele Dictionary and Grammar (Bristol,Sindebele Publishing, 1911); J. O'NeiL Grammar of the Sindebele Dialect "of Zulu (Bulawayo, EllisAlien, 1912); J. Pelling, Practical Ndebele Dictionary (Salisbury, Longman, 1966); and J. and P. Pelling,Lessons in Ndebele (Salisbury, Longman, 1974).Historical details concerning the Ndebele and their ancestry appear, inter alia, in A. T. Bryant,Olden Times in Zululand and Natal (London, Longmans, 1929), 417-45, and A. J.-B. Hughes, Kin, Casteand Nation among the Rhodesian Ndebele (Manchester, Manchester Univ. Press for Rhodes-LivingstoneInstitute, 1956). Arising, no doubt, from Mzilikazi's sojourn in the Transvaal until 1838, some vestiges oflinguistic influence from Northern Sotho are certainly apparent in Ndebele; but Bryant's statement seems fartoo extreme when he claims that, even before 1830, 'the whole Mzilikazi horde had by this time become aprofoundly Sutuized community, not in name alone, butinmembers, habits, language and blood; for many ofits men were Sutu captives and practically all its females', Olden Times in Zululand and Natal, 425.s cf. C. M. Doke, The Phonetics of the Zulu Language (Johannesburg, Witwatersrand Univ. Press,1926), 16 ff.* Notably Mr R. Mpande; see above, fn.2.D.K. RYCROFT .,2.2. Consonants' There are several small but noticeable divergences betweenNdebele and Zulu. The consonants involved are Z>, h, k, ng, zw. Also, Ndebele uses/**in place of Zulu sh, and in some instances substitutes / for Zulu n.2.2.1. Ndebele b is not implosive (as it is in Zulu and Xhosa) among speakers so farencountered. It varies from a lenis bilabial stop with delayed voicing, when initial, to afully-voiced bilabial approximant, resembling w but without lip-rounding, which ismore common in non-initial positions. The divergence from the Zulu articulation mayhere be due to influence from Shona or from Northern Sotho. In certain Shonadialects (e.g. Karanga), however, the cognate bilabial sound is a voiced fricative, whileNdebele b has little or no fricative quality.2.2.2. The h sound in Ndebele may optionally be rendered as a glottal fricative (likeEnglish h) or as a velar fricative (as Scottish ch), the latter probably being morecommon. This alternation, though it is not widely found in Zulu, does occurdialectally. Moreover, documentary evidence suggests that the two varieties may havebeen distinctive in Zulu, in earlier times, as in Xhosa today.82.2.3. In both Ndebele and Zulu, k is rendered as an ejective velar stop when initialin a root(or as a reduplicated root initial). In other positions, Zulu allows ejective, orlenis voiced plosive, as free variant realizations, the latter being more common; whileNdebele allows these, plus a voiced approximant articulation as free variants (thelatter matching the approximant variety of b, but with velar articulation).2.2.4. The Ndebele use of / in place of Zulu n in certain prefixes and pronounsis probably an innovation, possibly under Northern Sotho influence (though this isconjectural). Notable instances are the following:(a) lind ('y°u'> plural, absolute pronoun). Zulu has nind; N. Sotho haslend. It is interesting, however, that the addition of prefixal formativesrequires a reversion to ni, as in kini, Idni, ngdni, as in Zulu. Also thecognate pronominal stem is -fnu, as in Zulu, not *ilu.(b) //- ('y°u'> plural, subject concord for the second person plural).Zulu has ni-; N. Sotho has le-.(c) la- ('and' or 'with', conjunctive extra prefix used with nouns orpronouns). Zulu has na- (or net-, in Zululand dialect); N. Sotho has le-.2.2.5. In 'standard Zulu', ng is rendered as [ng] (roughly as in English 'finger') in allenvironments. In Ndebele this rendering occurs only root-initially. Elsewhere, ng =[nj, or sometimes [rj] (as in English 'singer5). This conditioned alternation, betweenpfenasalized g and the velar nasal sonorant, does occur similarly in some Zulu'For Zulu consonants, see Dolce, The Phonetics of the Zulu Language, 4 Iff; Lanham, 'TheComparativePhonologyofNguni\46ff.(alsoNdebele);A.TCo^-ZuluPhonoIo^,TonoIogyandTonaIGrammar" (Durban, Univ. of Natal, unpubl. Ph. D. thesis, 1966), 19ff.; Rycroft and Ngcobo, Say It in Zulu,Ap^ndix B. Brief notes on Ndebele consonants in P. Ladefoged, Preliminaries to Linguistic Phonetics(Chicago, Univ. of Chicago Press, 1971), 14, contain some typographical errors (for comment see, D. K.diaeresis *£**£*^ZTe^ZuiuDiZ^^, The Commercial Printing Co.,1923), ^MStaSS offheiulu Language, tZnoXzs that 'they seem to be phonenucallydistinct, though possibly not so with all speakers .112 NDEBELE AND ZULU: COMPARISONSdialects, as also in Swati. The Ndebele may therefore have brought this with themfrom Zululand, rather than having introduced it subsequently.2.2.6. The ejective palato-alveolar affricate [tf\ occurs in both Ndebele and Zuluin words such as -tshela (tell). But Ndebele additionally employs the aspirated variety[t/h| (also written as tsh). This is found wherever Zulu has [ f] (sh), e.g. Zulu: -shaya(hit) = Ndebele: -tshaya. The sound sh is absent in Ndebele except in one or two loanwords.The use of [t/h] in Ndebele may possibly represent a retention of an earlier proto-Zulu pronunciation, while Zulu has in fact changed. Many Zulu words nowadays speltwith sh appear, in nineteenth-century publications, with a tsh or ty spelling instead, oras an alternative; and the earlier spelling of the name Shaka, as Chaka, may well besignificant here.*2.2.7. With some Ndebele speakers,l0 the post-labialized clusterzw is rendered as alabialized z, with a 'whistling fricative' quality, when followed by / ore. This possiblyrepresents Shona influence but it seems not to be universal among Ndebele speakers.3. SUPRASEGMENTAL FEATURESA few minor differences have so far been observed, though their universal validity hasyet to be tested.3.1. Imposed lengthening of the penultimate syllable, as an utterance-terminal orpre-pausal marker, operates indentically in Ndebele and Zulu, its suppression (linkedwith lack of'final low* pitch; see 6.3.1) serving to indicate interrogation. But what isusually referred to as 'inherent extra length', pertaining to certain Nguni formatives,is less evident in Ndebele than in Zulu, e.g. in the class 2a noun prefix and in thecontracted forms of the class 5, 10 and 11 prefixes. Also the Remote Past tenseprefixal formative -&-, which has extra length in Zulu examples like walimd (heploughed), seems to lack such extra length in Ndebele, e.g. walimd.n3.2. In Ndebele, the initial vowel in a noun prefix possibly has a greater propensityfor dynamic stress than in Zulu.12 This might perhaps account for, or be somehow' Dote, The Phonetics of the Zulu Language, 112, regards the aspirated affricate when used inZulu, as being 'only found as the emphasized form of the fricative sh'; but in C. M. Doke and B. W. Vilakazi,Zulu-English Dictionary (Johannesburg, Witwatersrand Univ. Press, 1948), 818, it is stated as occurringalso 'as an alternative pronunciation, used individually and dialectally, to sh'. In most earlier dictionariesthere is much overlapping of entries undersA tmdtsh, but a few words occur under only sh, or only £sA; see,inter alia. Bishop J. W. Colenso, Zulu-English Dictionary (Pietermaritzburg, P. Davies, 1861), A. T.Bryant, A Zulu-English Dictionary (Pinetown, Mariannhill Mission Press, 1905), and R. C. Samuelson,The King Cetywayo Zulu Dictionary. From my personal recollections of elderly Zulu speakers, in the1930s, I recali hearing tsh used rather than sh in many cases. The late James Stuart (1868Š1942), whoseZulu was impeccable, used tsh frequently in his sound recordings (on Zonophone, 1927 to 1930) ; also mymaternal grandfather, A. W. Baker (1856-1952), born in Pietermaritzburg and a fluent Zulu speaker sinceboyhood, always used tshisa, tshiya and Tshaka, rather than shisa, shaya and Shaka." Notably Mr R. Mpande ; see above, fa 2." Surface suppression of underlying extra length might possibly be attributable to 'speech rhythm'requirements, in D. Abercrombie's sense (Studies in Phonetics and Linguistics (London, Oxford Univ.Press, 1965), 16-34), if Ndebele speakers in fact employ 'syllable timing'Šas against 'stress timing* whichseems to apply in Zulu (see Rycroft and Ngcobo, Say It in Zulu, Appendix B, 7-8); but this is uncertain yet12 Otherwise, as stated for Zulu in Rycroft and Ngcobo, Say It in Zulu, Appendix B, 7, stress inNdebele seems to be potentially associated with the first syllable of each root or stem, and with certainformatives. In practice, stress may be suppressed on certain stressable syllables in favour of stronger claimsfor its realization on a neighbouring syllable, or when, as in rapid speech, the number of stresses becomesreducedD.K. RYCROFT113connected with, both the realization of initial high tone in type-A nouns in Ndebele,as compared with Zulu (see 4.1), and also the fact that an abutting final depressor inthe previous word does not affect them in the same way (see 3.4).3.3. Tonally, almost identical rules appear to operate (see 6.0), although initial hightones are more prevalent in Ndebele (e.g. in nouns and verbs of tonal type A; see4.1)and, conversely, final high tones and penultimate falling tones are less prevalent (see4.2.-3).3.4. The effects of depression upon tone (see 6.3.4) are almost identical in Ndebeleand Zulu (while they are less extensive in Xhosa), but whereas in Zulu an abuttingfinal depressor in the previous word displaces prefix tone in examples likengithdnd' inyama ('I like meat', < inyama), this has not been observed in Ndebele.4. TONE-PATTERN DIFFERENCES BETWEEN NDEBELE AND ZULU134.0.1. Tone-marking employed in this paper is as follows:The acute accent, e.g. a, denotes high tone. The circumflex accent, e.g. a,denotes falling tone (high + low). Low tone remains unmarked.(For tonal realization, see 6.0 below).Depressor consonants, which in all Nguni languages have a pitch-lowering effect,are marked with a subscript diaeresis sign, e.g. v. The diaeresis is placed under thevowel, e.g. u, if the depression feature is not conditioned by a depressor consonant(see 6.2).Although a rising-pitch on-glide to high or falling tone is predictable from thesymbol combinations, as in yd and yd, or a and a , the rising-pitch and rising-fallingpitch signs [-] and [-] are used here, to mark it more clearly, e.g. yd, ya.In certain Zulu examples (in 4.2.3 ff) the tonal downstep sign [!] is employed. Itdoes not seem necessary in Ndebele.14It should be noted that the tone-marking conventions employed in this paper differslightly from the system used for Swati and for Zulu in several previous publications,the difference being that all surface high tones are here marked as 'high', even if theyare in fact predictable from their context, i.e. when they result from assimilation to aprevious high tone.154.0.2. Tonal Typology: In Nguni languages in general, monosyllabic-stem nounscan be assorted into one or other of two tonal types, A and B, with low and with high'' Terms used here for "parts of speech' follow those for Zulu in C. M. Doke, Textbook of ZuluGrammar (Johannesburg, Univ. of Witwatersrand, 1927). If the following observations seem obscure atsome points, readers are advised to refer to either D. K. Rycroft, "Tonal formulae for Nguni, LI MI (1979).VII, 20, or Rycroft and Ngcobo, Say It in Zulu, where the present approach to Nguni tone is covered ingreater detail. See also D. K. Rycroft, "Tone patterns in Zimbabwean Ndebele', Bulletin of the Schoolof Oriental and African Studies (forthcoming).14 See below, fn.32.' 'Most medial high tones are unmarked in D. K. Rycroft, Say It in siSwati [manual and tapes] (Univ.of London, School of Oriental and African Studies, 1976; Mbabane, Websters, 2nd edn, 1979), and inRycroft and Ngcobo, Say It in Zulu. An explanation of that simplified, but more abstract tone-markingsystem appears in Rycroft, 'Tonal formulae for Nguni', 20. Such a system is indeed applicable to Ndebelealso, because in most Ndebele words with more than two high or falling surface tones it can be demonstrated(as in Swati or Zulu) that the medial ones are attributable to tonal assimilation. This can be establishedthrough a process of substitution, choosing words which are comparable, but which have depressorconsonants in various positions. For the present paper, however, a more direct marking of surface tonesseems preferable.114 NDEBELE AND ZULU: COMPARISONStone on the stem, respectively. Polysyllabic-stem nouns are mostly divisible into fourtonal types. In several earlier publications,16 these have been termed A, B, C and D,to avoid stating actual tone patterns, because some tonal types have several variantpatterns. Monosyllabic stems are labelled as Al and Bl; disyllabic stems as A2, B2,C2 and D2; trisyllabic as A3, B3, and so on. Stems retain their basic tonal type whensuffixally extended: Al stems when extended take the same patterns as A2 and A3stems; Bl stems graduate to B2 and B3, and so on. Adjective and relative stems canbe similarly grouped.Verbs are of two tonal types in most Nguni languages. But Xhosa, and theZululand dialect of Zulu, employ three types, in polysyllabic verbs; the distinctionbetween types B and C appears to match the vowel-length distinctions in CommonBantu starred forms.174.0.3. While Ndebele largely conforms to general Nguni practice regarding tonaltypology, certain deviations in tone-pattern are apparent. Notably, Ndebele employsinitial high tone in all nouns of tonal type A,18 while most other Nguni languages(except the Gcaleka-Rharhabe or 'Ciskeian' dialect of Xhosa19) do not.20 A similardifference is apparent with Ndebele verbs of tonal type A, in certain usages.4.0.4. While favouring initial high tone in such cases, Ndebele conversely tends todiscard final high tones and penultimate falling tones quite frequently, as comparedwith other Nguni languages. In some cases this has led to an apparent fusion betweentwo, and in some instances three, tonal types, among nouns in particular; and thispresents certain classificatory problems if one attempts to deal with Ndebele tone-patterns synchronically, without considering their Zulu counterparts.4.1. Prefixal high tones in Ndebele'type-A'nouns and verbs In Table I, it willbe seen that the Ndebele nouns have initial high tones where their Zulu cognates havelow. As Ndebele derives originally from Zulu, one might expect that these initialsyllables must have become raised, for some reason, after the Ndebele exodus fromZululand in 1822.Table IZULU NDEBELEabdntu dbdntu (people)abdntwana dbdntwana (children)abantwdnyana dbdnfwdnyana (small children)" See, inter alia, Rycroft, 'Tonal formulae for Nguni', 9-23."See D. K. Rycroft, "Nguni tonal typology and Common Bantu', African Language Studies (1980),'" Tonaltype A', fornouns in most Nguni languages, implies'thosenounswhichbearnothingbutlowtones in vocative usage', e.g. bantu! (O people!);bantwana! (O children!). But in Ndebele, certain 'pseudotype-A nouns' need to be excluded (see 4.2.3.ff)." Regarding Xhosa dialects, see A. S. Davey, 'Some Aspects of the Phonology of the Noun inXhosa' (Edinburgh Univ., unpubl. M.Litt thesis, 1975); H. W. Pahl, 'The distribution and functional rolesof certain significant tones and tonal sequences in Xhosa', LIMI (1977), V, 18-35 J A. Louw 'Someremarks on Nguni tone", LIMI (1979), VII, 45-6."An exception here is monosyllabic-prefix nounsoftype A with CV or CVCV stem. These do takeinitial high tone in all Nguni languages.D.K. RYCROFT 1154.1.1. Upon deeper investigation, however, this assumption becomes untenable.There are stronger grounds for believing the converse to be true, namely that it is infact Zulu that has changed since 1822, having discarded high tone from initialsyllables, while Ndebele has retained what was an earlier Zulu practice. Theargument in support of this view has been expounded in an earlier paper,21 where itwas proposed that the presence of certain high tones in Nguni nouns, sometimesrealized non-initially, could be ascribed to underlying +H pertaining to the initialprefix vowel, which has 'spread' from that vowel up to a predetermined later syllable.4.1.2. In Table II, the second line of examples shows three Ndebele nouns, tonally'derived' in this way from the underlying constructs in line 1. Besides Ndebele, theGcaleka-Rharhabe (or' Ciskeian') dialect of Xhosa employs these tone patterns also.On the other hand, Zulu, like most other Nguni dialects, employs the patterns shownin line 3 of Table II. It seems most likely that Ndebele-type patterns represent anearlier stage than the latter: that is, that the apparent 'delayed realization', as found inZulu, has come about first through tonal spreading, and then subsequently throughanterior deletion of all high surface tones except the last one.Table II'Proto-Nguni'underlying form * dbantu * dbantwana * dbantwanyanaby H-spreading(Rule 1 )* N. dbantu N. dbantwana N. dbantwanyanaby AnteriorH-deletion (Rule 2) t Z. abdntu TL. abdntwana Z. abantwdnyana*H-spreading rule: ŠH Š +H / +H (subject to stem-type constraints: i.e. for stemsof Tone Group Al, the last syllable is exempt: otherwise the last two syllables are exempt (i.e. forTone Groups A2 and longer).t Anterior H-deletion rule: +H Š -H L +H for Type-A nouns.4.1.3. Depressor-Conditioned Variant Patterns: In type-A nouns which containa depressor consonant, Ndebele and Zulu respond identically to depressor-conditioned displacement of high tone (as formulated in the 'Depressor H-displacement rule' under 6.3.4.2). However, they differ again regarding initial hightones. The effect of depressors on type-A nouns with two or more stem syllables is asfollows (being merely a particular application of the H-displacement rule).If the third-last syllable has a depressor consonant, and is followed by a non-depressor, high tone moves from that syllable to the next, the second-last syllable.There it is realized as a falling tone, if the word is utterance-final (but as high tone ifanother word follows). Table III shows utterance-final examples with two, three andfour stem syllables, for Zulu and Ndebele. As with the Table II examples, the tonepatterns of these words can be 'derived' from underlying initial high tone, but in thiscase, the Ndebele patterns require insertion of the Depressor H-displacement rule (orits specific abbreviation stated above), after the H-spreading rule; while the Zulupatterns require, after this, the Anterior H-deletion rule, as before.21 Rycroft, "Tonal formulae for Ngunf. 6-9.116 NDEBELE AND ZULU: COMPARISONSTable III'Proto-Nguni'underlying form: * izihlahla * imvumelo * uphondwanyanaby H-spreading (izihlahla) (imvumelo) (uphondwanyana)by DepressorH-displacement N. izihlahla N. imvumelo N. uphondwanyanaby AnteriorH-deletion Z. izihlahla Z. imvumelo Z. uphondwanyanaNote: The meanings of these words are: "bushes', 'permission', and 'tiny horn'.In passing, it may be of interest to note that the intermediate patterns in the secondrow are in fact used in the 'Ciskeian' dialect of Xhosa, neither of the subsequent rulesbeing applicable there. In Transkeian Xhosa, Anterior H-deletion is added, however,but not Depressor H-displacement, thus yielding izihlahla, imvumelo anduphondwanyana.224.1.4. Type-A Verbs: Tone-patterns vary considerably in different tenses, but incertain instances exactly the same differences between Ndebele and Zulu regardinghigh initial tones as those noted above are also found with type-A verbs.4.1.4.1. With the infinitive prefix uku-, verbs such as -Iwa (fight), -lima (plough), and-hlakula (weed) take identical patterns to those shown in Table I, above, for Zulu andNdebele respectively. This would be expected, of course, since verbs in this form alsoqualify as nouns of class 15.4.1.4.2. Depressor-conditioned variant patterns also occur. The verb -vumela(consent), with prefix uku-, for example, has the following patterns:Ndebele: ukuvumela; Zulu: ukuyumela (to consent).It will be seen that these tones tally with those of the second noun in Table III.4.1.4.3. Among other usages where type-A verbs are tonally comparable with type-A nouns is the Present Indicative tense, in its 'long' form with infix -ya-, whenoccurring with a third person subject concord, but with no object concord. Table IVshows examples.ZULUbaydlwabaydlimabayahldkulabayayumelaTable IVNDEBELEbaydlwabaydlimabayahldkulabdydvumela(They are fighting)(They are ploughing)(They are weeding)(They consent)4.2. Loss of Falling Tone Quite frequently, where Zulu has falling tone, Ndebelehas low tone, but there are exceptions to this. The matter will not be treatedexhaustively here but some notable instances will be cited, with reference to" Ibid., 8-9.D.K. RYCROFT117possessive pronouns, locative demonstratives, nouns, adjective and relative stems,and verbs.4.2.1. Possessive Pronouns: Zulu and Ndebele use identical high + low tonepatterns in possessive pronouns of the first person singular, as in indlu ydmi (myhouse), or indlu ydkho (your house); but for all other persons and classes, whereZulu has falling tone, as in yakhe ('his/her', class 1), or yabo ('their', class 2),Ndebele has low + low tone instead, e.g. yakhe, yabo, etc.4.2.2. Locative Demonstratives: These provide an exception, because Ndebeledoes employ initial falling tone in the disyllabic forms, when utterance-final, as innangu ('here he is', class 1), oxnampa ('here they are', class 2). It should be noted,however, that this falling + low pattern is not used in Zulu. Zulu has high + low, forclasses 1, 4 and 9, e.g. nangu (class 1), and falling + high, with a «o«-depressorinitial V, for all other classes, e.g. nampd (class 2). Ndebele appears unique amongNguni languages in using one tone pattern for all classes; Xhosa and Swati makesimilar class distinctions to Zulu in this respect4.2.3. Nouns: Nouns of tonal type C2 in the other Nguni languages bear falling+ low tone on the stem when pre-pausal or final.23 Most cognate nouns in Ndebelebear low + low stem tones instead, and are tonally indistinguishable from type-Anouns, in such usage, e.g.Cf. TYPE A2 EXAMPLESNDEBELE/ZULUintaba (mountain)indoda (man)indaba (matter)isi'hlahla (bush)(Zulu: isi'hlahla)iziziba (pools)(Zulu: iziziba)* -> a i in which the4.J.3.I. An exception occurs, however, with those type-C2 ncmrtS '" f Hergthird-last consonant is a depressor, and the second-last is a noo-4epreS ffl 'tones:Ndebele employs the same tone pattern as Zulu, with falling + l°w sZULU NDEBELEizikhwama (bags) izikhwamayinkabi (it is an ox) yinkabittern does notNote, however, that if the second-last consonant is a depressor, this P31 aJ agajnstoccur in Ndebele, e.g. izigodo cited above, or yimbuzi (it is a tPai''Zulu: vim'buzi.'" " " . ,uch situations4.2.3.2. Although this falling + low stem pattern, when occurring 'n h:s js in factin Ndebele, appears to represent a retention of the Zulu C2 patter0' \^ing + lowquestionable. As previously demonstrated under 4.1.3 above, fu~ "<"°eZULU (C2)inkabi (ox)im'buzi (goat)indawo (place)isi'khwama (bag)izi'godo (logs)NDEBELE (=A2)inkabiimbuziindawoisikhwamaizisodo"Ibid.. 13ft118 NDEBELE AND ZULU: COMPARISONStones occur with normal type-A Ndebele nouns, as a variant pattern conditioned bythe same consonantal sequence as our present falling + low examples; and in suchnouns the falling tone was ascribable to depressor-displaced prefixal high tone. Afurther test shows that the assumption that our'exceptional nouns' had, like Zulu C2nouns, inherent falling tone on the stem, was misguided. When such nouns occurwithout an initial prefix vowel, as in vocative or in post-axiomatic-negative usage, theZulu and Ndebele patterns differ radically; the stem pattern is then low + low inNdebele, which is the normal type-A2 pattern, while the Zulu cognates retain theirinherent falling + low type-C2 tones:ZULU NDEBELEakiina zikhwama akula zikhwama (cf. zihlahla(there are no bags) type A2, 'trees')asithengi nkabi asithengi nkabi(we buy no ox)4.2.3.3. We must conclude, I think, that when falling tone does in fact occur inthese 'quasi-type-C Ndebele nouns, it represents consonantally-displaced high tone,as in type-A nouns in Table III above, from which these nouns are indistinguishablein Ndebele. It does not represent inherent falling tone pertaining to the stem (which isthe case in Zulu. Xhosa and Swati, for genuine type-C2 nouns), because it disappearswhen the initial prefix vowel is elided.4.2.3.4. To say that Ndebele cognates of Zulu C2 nouns are tonally in-distinguishable from those of type A2 requires further qualification, however. This istrue only when their stems remain uninflected. If the stems are extended, by suffixes,the tone patterns of such Ndebele nouns do in fact differ from those of true type-Anouns: Their stem patterns tally with those of Zulu type-C nouns when similarlyextended:ZULU NDEBELEType-C derivatives:enkabini (< in'kdbi) enkabini (< i nkabi)(on the ox)isikhwamana (< isi'khwama) isikhwamana (< isikhwama)(small bag)imbu'zana (< im'buzi) imbuzana (< imbuzi)(small goat)enda'weni (< indawo) endaweni (< indawo)(in the place)Type-A derivatives:entdbeni (< intaba) entdbeni (< intaba)(on the hill)Depressor-conditioned variants:endabeni (< indaba) endabeni (< indaba)(in the matter)D.K. RYCROFT 1194.2.3.5. The fact that, unless extended, Ndebele cognates of Zulu C2 nouns aretonally indistinguishable from type-A2 nouns raises certain classificatory problemsnot found in other Nguni languages, since it is in the citation forms that neutralizationhas taken place, between two tonal types, while differentiation has been retained onlyin the derivative forms.4.2.3.6. Though we have dealt here only with disyllabic-stem nouns, loss of fallingtone is also apparent in longer nouns. This point will not be pursued here, beyondciting a few examples:ZULU NDEBELEinhlabathi (soil) inhlabathiithendele (partridge) hhendeleintombdzane (young girl) intombazaneA sequence of three or more low tones on the stem, as in the last two Ncteteleexamples, never occurs in uninflected Zulu nouns.4.2.3.7. Falling tone does occur in certain Ndebele polysyllabic nouns, ho*ever>when it represents displaced high tone from a depressed syllable (as in the disyl'abic~stem nouns covered in 4.2.3.1), e.g.indatshana (small matter) cp. intdtshana (smallmountain)dmddodakazi (daughters) cp. amdkhosikazi (queens)4.2.3.8. With monosyllabic noun stems, falling tone occurs on certain Psyllables exactly as it does in Zulu, Le. with the following nouns of tonal tyP6when pre-pausal or final:(a) Nouns with class 9 prefix: e.g.into (thing); imvu (sheep)(b) Nouns with a disyllabic prefix (other than classes 8 or 10 whichcontain a depressor) when preceded by depression in a prefixalformative, which conditions the displacement of its initial high tone>e.g.yikiidla (< ukudla)(it is food) (food)It should perhaps be noted, however, that in Zulu these three examples have h'#\ ^ 3on the stem, not low. Loss of final high tones in Ndebele is discussed u°"below.u/ con>4.2.4. Adjectives and Relative Stems: In attributive usage these are close 7 tonesparable with nouns, in tone. Those of tonal type C2 in Zulu, bearing falling + '° e g.on the stem, take low + low in Ndebele, just like the nouns cited in 4.2--''ZULU NDEBELEdbd'khulu ('big ones', class 2) dbdkhuluemanzi ('a wet one' class 9) emanzi ,4.2.4.1. Unlike comparable type-C2 nouns (cf. 4.2.3.1), Ndebele adje^ .materelative stems do not seem to accept displaced prefixal tone, realized as p*olJ120 NDEBELE AND ZULU: COMPARISONSfalling tone on the stem, when the prefix contains a depressor; the second prefixsyllable just loses its high tone altogether, e.g.ezinkulu ('big ones', class 10)4.2.4.2. In adverbial usage, type-C2 stems in Ndebele (also in Swati) takehigh + low stem tone while Zulu has falling + low, e.g.ZULU NDEBELEkakhitlu (greatly) kakhulu4.2.4.3. Polysyllabic stems in Ndebele lose penultimate falling tone in the sameway as nouns (cf. 4.2.3.6), eg.ZULU NDEBELEeluhlaza ('a green one', class 9) eluhlaza4.2.4.4. With monosyllabic stems of tonal type Bl, falling tone occurs on a class 9prefix, just as with Bl nouns (cf. 4.2.3.8), e.g.entsha ('a new one', class 9) cf. Zulu: entsha4.2.5. Verbs: In Ndebele, the characteristic penultimate falling tone of type-C2verbs in the kwaZulu dialect and in Xhosa (in certain usages) has largelydisappeared. In the kwaZulu dialect and in Xhosa, three tonal types aredistinguishable among CVC-root verbs.24 In infinitive usage (i.e. with noun class 15prefix, and final -a) they resemble nouns of tonal types A2, B2, and C2 (withpenultimate falling tone when pre-pausal or final), e.g.Type A: ukulima (to plough)Type B: iikufiinda (to read/learn)Type C: ukufuna (to want)It should be noted, however, that verb tone patterns vary considerably in differenttenses, often obscuring typological distinctions.4.2.5.1. Ndebele cognates of type-C2 verbs in the kwaZulu dialect differ from themtonally in some respects, but in addition to this, tone patterns tend to vary withdifferent Ndebele speakers. Whether such variation has a regional basis has not yetbeen established.Tentative findings have been that speakers of the 'older generation' (say, 40 yearsand over) still distinguish type-C2 verbs from the other tonal types, but that theyounger speakers tend not to do so. The label 'Ndebele I' will here denote tonalrenderings by several informants drawnfromthe 'older generation', while Ndebele II'applies to those of younger informants. The following examples show tones used byolder, and by younger Ndebele informants consulted, for cognates of verbs which areof type-C2 in the kwaZulu dialect, in infinitive, imperative and present tenseindicative usage:Unfortunately, Zulu verbs of the third tonal type, type C, are indistinguishable from those of type Bin the standard dictionaries and grammatical works, since these favour the Natal dialect; but see Rycroft andNgcobo, Say It in Zulu, Appendix E; and also Rycroft, 'Nguni tonal typology and Common Bantu',Appendix 1.D.K. RYCROFT121KWAZULU DIALECTType-C2 patternsukufiina(to want)fima(want!)ngiyafuna(I want)NDEBELE Iukufuna(= type-A2 pattern)(as Zulu: type-C2pattern)ngiyafuna(= type-B2 pattern)NDEBELE IIukufuna(=type-B2 pattern)funa(=type-B2 pattern)ngiyafuna(Ltype_B2 pattern)4.2.5.2. With 'Ndebele I' speakers, type-C2 verbs tested seemed to retain theirdistinctive falling tone in imperative usage, but to assimilate with either type-A2 orB2 in other usages. Type-C2 verbs tend to remain distinct in present participial usage,however, taking an 'all low' tone pattern, e.g.nxd efuna (if he wants) [cf. kwaZulu dialect: nxd efuna]in contrast to 'low-high-low' for types A2 and B2, e.g.A2: nxd elima (if he ploughs);B2: nxd efunda (if he reads).4.2.5.3. With 'Ndebele II' speakers, total assimilation, mainly with tonal typeB2, seems to occur, and this tallies with Natal Zulu and Swati practice; but a smallproportion of type-C2 verbs seem to have joined type A2 instead, e.g- 'P.hala (wnte)'-hleba (slander); -linga (tempt). Further research may find differences amongindividual speakers here, however.4.3. Loss of Final High Tone Lack of high tone on word-final ^^prevalent in Ndebele than in other Nguni languages, though some elderlyspeakers follow Zulu practice to some extent4.3.1. Monosyllabic-Stem Nouns: With many younger speakers.identical in pairs of nouns such as the following. When used without anvowel they are tonally indistinguishable in Ndebele, while in Zulu the ^and high tone, respectively (representing tonal types Al and Bl):NDEBELE(akula) ndlu (there is no) house(akula) nja (there is no) dogPlural:(akula) zindlu (there are no) houses(akula) zinfa (there are no) dogsare tonally4.3.1.1. In normal use, however, with a full prefix, such no*0* wae8 differdistinguishable through their prefix tones, in Ndebele. In Zulu, the si60al bt i Ndbl l dii tfZULu(akuna) ndlu(akuna) nja(akuna) zindlu(akuna) zinjage through their prefix tones, in Ndebele. In Zulu, the s fl yardedalso, but in Ndebele, tonal distinction on the stem itself seems to have tf fix yy^dd *, stinction on the stem itself seems to havas redundant, since in normal usage an adequate clue is provided bypattern.!22 NDEBELE AND ZULU: COMPARISONSNDEBELE ZULUAl indlu (house) indluBl inja (dog) injaPlural:Al izindju (houses) izindluBl izinja (dogs) izinja4.3.1.2. Monosyllabic adjective stems in Ndebele similarly show loss of high stemtone and /eliance on prefix-tone distinctions, e.g.AlBlNDEBELEende ('a longembi ('a badPlural:AlBlezinde ('longezimbi ('badone',one',ones'ones'class 9)class 9)', class 10), class 10)ZULUendeembiezindeezimbi4.3.1.3. When Ndebele nouns of tonal types Al and Bl are extended by suffix,however, the stem tones do differ, taking the same patterns as disyllabic-stem nouns oftonal types A2 and B2, low + low and high + low, respectively, e.g.DERIVED STEMS NON-DERIVED STEMSA2 isimbana (small mattock) cf. isigodo (tree stump)B2 isidlwdna (small feast) cf. isigubhu (drum)Plural: _A2 izimbana (small mattocks) cf. izigodo (tree stumps)B2 izidlwana (small feasts) cf. izigubhu (drums)4.3.2. CVCV-Stem Nouns: In most Nguni languages, disyllabic-stem nouns ofone tonal type, D2, take low + high stem tones. Though elderly Ndebele speakerssometimes retain the Zulu pattern, such nouns are mostly rendered with low + lowstem tones in Ndebele, or falling + low if prefixal high tone is displaced by a depressor,as in the last example:NDEBELEinkomo (head of cattle)intombi (marriageable girl)isikhathi (time)izikhathi (times)ZULUinkomointombiisikhathiizikhathi4.3.2.1. Adjective and relative stems which take type-D2 tones in other Ngunilanguages similarly show loss of final high tone in Ndebele (with many speakers):NDEBELE ZULUababili ('two', class 2) ababilieii?/wvu('aredone', class 9) ebomvuD.K. RYCROFT 1234.3.2.2. Whereas prefix tones give a clue to tonal type, for monosyllabic-stemnouns, adjectives or relative stems in Ndebele, they do not, in the case of disyllabicwords like those above. In their normal usage, such words are tonally identical toones from tonal type A2, and also to ones which are cognates of type C2, with falling+ low stem tones, in other Nguni languages. For example, the following threeNdebele nouns, with identical tone patterns, have three different tone patterns inZulu, representing tonal types A2, C2 and D2, respectively:NDEBELE ZULUintaba (mountain) intaba type A2inkabi (ox) inkdbi type C2inkomo (bovine) inkomo type D24.3.2.3. As demonstrated previously, in 4.2.3.4, suffixally extended derivativesfrom Ndebele types A2 and C2, like intaba and inkabi, are in fact tonallydistinguishable from each other, taking penultimate low and high tone, respectively.Type-D2-derived items, like inkomo when similarly extended, differ from A2derivatives, but share the same pattern as C2 derivatives:< A2 < C2 or D2entabeni (on the mountain) enkdbini (on the ox)enkomeni (on the bovine)Depressor-conditioned variants:endabeni (in the matter) endaweni (in the place)4.3.3. Verbs: For Ndebele verbs, some similar instances of the loss of final hightone could be cited, e.g. with type-Bl verbs, in some tenses, as in ngiyadla. I ameating' (cf. Zulu: ngiyadla).4.3.3.1. Besides cases of final high tone loss, one finds that earlier high tones havedisappeared in certain tenses, as in asihlakulu 'we are not weeding (cf. Zulu:asihlakuli). In this instance there has been a neutralization of tonal contrast betweenverbs of types A and B, in Ndebele. Zulu uses the -penultimate high only pattern tortype-B (or-C) verbs only, in this tense, e.g. asisebenzj (yrt are not working): and the'high except first and last' pattern, for type-A verbs. Ndebele uses "penultimate highfor either type.4.4. Overall picture of tonal changes To sum up our findings presented ir.thissection, concerning tone-pattern differences, I think the general p.cture that emergesis that Ndebele tends to favour early rather than late realization of high.tone whi eZulu tends to prefer the reverse. An exception occurs, of course, m the case^c ed ,n4.3.3.1 where Ndebele discards all high tones except on the Penult1Ž* sflab£compared with Zulu: but this arises from a standardization of patterns or that tensjdistinctiveness between tonal types A and B being neutralized Th» >"*">Žtendency is evident in certain other instances, in Ndebele, as for exarnpkrw.thlocative demonstrative tone patterns (4.2.2). and in the fusion of tonal types B and Camong verbs (4.2.5).5. TONAL FORMULAE .As a convenient means of referring to Nguni tone patterns, a concise 'tonal formulae124NDEBELE AND ZULU: COMPARISONSsystem has been outlined in an earlier paper.25 It involves symbolizing stem syllablesby counting backwards from the end and labelling them as Z, Y, X, etc. Thus allNdebele 'penultimate high' patterns, as in asilimi or asihlakuli (we are notploughing; we are not weeding) can be represented as Y; and all 'antepenultimatehigh' patterns, as in siyayihlakula or siyayilimisisa (we are weeding; we areploughing it intensively) as X, and so oa For additional, earlier high tones, syllablesare numbered from the beginning, as 1, 2, etc. But conditioned, or assimilated hightones remain unmarked: e.g. isikhwamana and izikhwamdna (small bag; smallbags) are both symbolized as l'Y, since the presence or absence of intervening hightones is predictable (in terms of the Assimilation rule; see 6.3.3, iii and iv).5.1. The tone patterns found in most Zulu and Ndebele nouns, for example, can berepresented by the formulae shown in Table V. The symbol Z in parenthesis forNdebele Bl and D2 nouns denotes final high tone, which seems to be unrealizedexcept by some 'older generation' speakers. As they stand, these formulae refer tonormal 'full prefix' forms with an initial vowel. Tone-patterns for the corresponding'short' forms, without initial vowel, can be directly derived from these, by just deletingthe first term in each of the formulae; but for Ndebele type-A and type-D2 nouns, thefirst two terms need to be deleted. The footnotes marked as t and * within the Tablemerely cite various realizations of the 'depressor H-displacement' rule, stated in theAppendix to this paper.26Table VTONE-PATTERN FORMULAE FOR NOUNSOF TONAL TYPES A, B, C & DNo. ofof stem(a) ZULUsyllables1234AYX*"B1Z++l'Yix*ciztl'YIX*DtitlYix*(b)AiYix*"NDEBELEBi(z++)1Yix*"cix*l'Yix*Dix*(Z)l'Yix*t 1/depr non-depr;3 Š 3 /3 = pre-pausal Y++ Pre-pausally: 1t Pre-pausally: Z* X -» Y /depr_1 /monosyllabic prefixYnon-depr; Y -* Y /pre-pausalKey:1 & 2 = prefix syllables (with monosyllabic prefix, 2=1).3 = next syllable.Z, Y & X = last, second-last, and third-last syllable.25 Rycroft, 'Tonal formulae for Nguni'.26 See ibid, 23-5, for fuller details of this formula system, plus examples of nouns of each tonal type-D.K. RYCROFT1255.2. A rather different and somewhat problematical tonal typology for Ndebelenouns emerges if no account is taken of the cognate Zulu forms. For ex ample, there isno way of distinguishing type- A2 nouns, by their surface tones, fron*type-C2 (or, wid*younger speakers, from type-D2 nouns either). Yet it is desirable to identify them in
order to be able to predict their differing A3 and C3 /D3 tone patterns when such nounsare suffixally extended.5.3. Conclusion This paper gives only a tentative sketch of certain phonetic andtonal differences between Ndebele and Zulu. Divergences concerning vowel andconsonant pronunciation are minimal, and in some cases the evidence suggests that itis in fact Zulu that has changed, since 1822, while Ndebele has retained earlier forms
of pronunciation, intact Similarly with tone, the fact that Ndebele uses initial hightone instead of low in certain nouns may well represent a survival of proto-Ngunipractice no longer retained in Zulu. On the other hand certain tonal innovations arealso apparent in Ndebele, especially among the younger generation: there is a definitetendency to discard final high tones and, in certain instances, penultimate fallingtones, in words in which these occur in all other Nguni languages. Also there are someinstances of standardization (or generalization) of tone patterns in Ndebele, in somecases involving a reduction of distinctiveness between 'tonal types' among lexicalitems.For the future, I think Nguni comparative studies, particularly concerning tone,offer a rewarding field for systematic exploration.6. APPENDIX: TONAL REALIZATION IN NDEBELE276.1. Tone At a phonetic level of description, many fine gradations of absolutepitch occur in Ndebele, as in Zulu.28 At a phonological level these can be systema-tized into three significant categories of tone (or tonemes),29 their actual pitchrealization being subject to modification by three additional factors: downdrift,
assimilation, and depression, as explained below.6.2. Depression30 In all Nguni languages, so-called depressor consonants con-
dition initial breathy-voice phonation, in place of normal voicing, in the followingvowel, and this has a pitch-lowering effect The following Ndebele consonantsare depressors: bh, d, dl, g, gc, gq, gx, j, mb, my, nd, ndl, ng, ngc, ngq, ngx, ry, nz,v, z. Also the following, if marked with a subscript diaeresis, in this paper:h, I, m, n. Most of these can also be followed by w. All other Ndebele consonantsare non-depressors.27 For Zulu, a statement which is closely similar to what is given below appears in Rycroft andNgcobo, Say It in Zulu, Appendix B, 8-24.28 cf. Doke, The Phonetics of the Zulu Language, 199ff., regarding 'nine different pitches' in Zulu.29 Regarding Xhosa tonemes, see D. M. Beach, "The science of tonetics and its application to Bantulanguages', Bantu Studies (1924), II, 75-106; AN. Tucker, 'Sotho-Nguni orthography and tone-marking',Bulletin of the School of Oriental and African Studies (1949), XIII, 200-24; Lanham, 'The ComparativePhonology of Nguni'. For Zulu, see Cope, 'Zulu Phonology, Tonology, and Tonal Grammar1, and 'Zulu-tonal morphology', Journal of African Languages, 1970), IX, 111-52; D. K. Rycrdft, 'Tone in Zulu
nouns', African Language Studies (1963), IV, 46-68, and'Tonal formulae for Nguni'; and Rycroft andNgcobo, Say It in Zulu.30 A fuller account of this feature appears in Rycroft, The Depression Feature in Nguni Languagesand Its Interaction with Tone.126NDEBELE AND ZULU: COMPARISONSIn a few instances, breathy voicing occurs without a preceding depressorconsonant, notably in subject concords of the first and second person in certain tenses,and initially in nouns under copulative inflection, e.g. umuntu ('it is a person',<iimuntu, person).31 In such cases it is here referred to as the 'depression feature',and the diaeresis sign appears under the vowel instead of under a consonant6.3. Tonal Realization The effects of downdrift, assimilation and depressionupon high, falling and low tones in Ndebele are as follows:6.3.1. Downdrift: Except in questions, successive low tones, high tones, and theonset pitches of falling tones, respectively, become progressively lower in actual pitchtowards the end of of an utterance, as may be seen in examples (a) and (b) below.'Final low' pitch appears to serve as an utterance-terminal marker. In terms of whatmight be called the 'Final low' rule: Lowest pitch must be reached on either or both ofthe last two syllables.326.3.2. For early unmarked syllables, depression conditions low-mid pitch realiza-tion instead of mid pitch: see example (g), as compared with (e).6.3.3. Assimilation:(i) Low tone on a penultimate syllable (after H) begins with a falling-pitch onsetglide (see example (c));33 but(ii) No falling onset occurs if that syllable commences with a depressor consonant(see example (d)).(iii) Low syllables that precede a penultimate or earlier high tone in a word,assimilate to high pitch if another H tone precedes them (see examples (e) and(f));34 but(iv) No assimilation occurs if such syllables follow a depressor (see examples (g)and (h)).6.3.4. Depression: This feature, conditioned either by a preceding depressorconsonant or, more rarely, by imposed depression, affects tonal realization asfollows:356.3.4.1. High tone is lost or displaced, in certain contexts. Conditions for this arethat the affected syllable must:(i) be earlier than penultimate in the word;(ii) not be followed immediately by a depressor,(iii) not have extra length.31 An alternative interpretation could posit an unwritten breathy-voiced ft here, as an 'invisible'initial depressor. But this is not feasible when there is an initial non-depressor consonant, as intnttshani ('it is grass') < iitshani (< ubti-, class 14).32 This "final low pitch' rule applies in all Nguni languages except Xhosa. In most Nguni languages,besides 'final lowering', a drop in pitch between successive high tones generally occurs, between words, andalso within words, throughout the utterance, thereby delimiting the separate domains of underlyingmorphemic high tones. In Ndebele, high-pitch down-step within words seems rare (except utterance-finally).When it does occur it appears to be optional and arbitrary." Note that a penultimate syllable takes extra length (see 3.1).34 These syllables are marked as 'high' under our present system, employed in this article; but theywould be unmarked under the system cited above, fa 15.J! In Ndebele, depressor H-displacement seems to observe word-boundaries; see 3.4.NDEBELE AND ZULU: COMPARISONS127Provided that these conditions are fulfilled, the affected syllable then takes low toneinstead of high (see example (i)). Furthermore, the next syllable, if normally low, thenadopts high tone instead or, if it bears imposed extra length (through being utterance-penultimate), adopts falling tone (see examples (j) and (k)).6.3.4.2. This 'Depressor H-displacement' rule, which is relevant in all Ngunilanguages except Xhosa, may be expressed as follows:' CV "+ H+ depr1" CV 1Š depr(v).C1-H2+ HItems 1 and 2 must have depression and lack depression, respectively, and item 4 isobligatory. Item 3 represents an optional increment of imposed extra length, extendingitem 2 if the word is final or pre-pausal. This increment, when present, does not acceptdisplaced high tone; so [+H] and 3, together, represent falling tone (unless item 2already bears non-displaced high tone, in which case 2 + 3 = high tone).6.3.4.3. If the above conditions for loss or displacement do not apply, then high orfalling tone on a depressed syllable commences with a rising-pitch on-glide and takesslightly lower pitch (see examples (1) and (m)).6.3.4.4. The question of what constitutes a 'high tone' is relevant here. Syllablesonly realised at high pitch due to assimilation are exempt: they merely lose that highpitch, after a depressor (see 6.3.3, iii-iv).EXAMPLES(a) dbdlimi abalimi (the farmers do not plough)[Š-----I(b) abalimi bahle (the fanners are handsome)(c) dbdlimi (farmers) (d) dmddoda (men)r'"«--] I":--](e) sisebenzi (O servant!) > (f) isisebenzi (aservant)["--"-] [---I(g) zisebenzi (O servants!) > (h) izisebenzi["---"-I I"---'-](i) (v)inqola (it is a wagon) < inqola (wagon)128NDEBELE AND ZULU: COMPARISONS(j) yikiidla kwdmi (it is my food) < ukudla (food)(k) yikiidla (it is food)r- \-i(1) (y)indoda (it is a man) < indoda (man)[>--] [Ł--](m) nampa (here they are)