silewp2016 001.
DigitalResources
Electronic Working Paper 2016-001
New perspectives on the
genetic classification of
Manda (Bantu N.11)
Hazel Gray and Tim Roth
New perspectives on the genetic classification of
Manda (Bantu N.11)
Hazel Gray and Tim Roth
SIL International®
2016
SIL Electronic Working Paper
2016-001
2016 SIL International®
©
ISSN 1087-9250
Fair Use Policy
Documents published in the Electronic Working Paper series are intended for scholarly research and
educational use. You may make copies of these publications for research or instructional purposes
(under fair use guidelines) free of charge and without further permission. Republication or commercial
use of an Electronic Working Paper or the documents contained therein is expressly prohibited without
the written consent of the copyright holder(s).
Managing Editor
Eric Kindberg
Series Editor
Lana Martens
Content/Copy Editor
Mary Huttar
Compositor
Bonnie Waswick
Abstract
Manda (N.11) is an under-documented Bantu language in southwestern Tanzania, with most mention of
its closest genetic affiliation in the work of Nurse (1988). Nurse concludes that Manda belongs with the
Southern Tanzania Highlands (SH) subgroup based primarily on phonological evidence. This paper uses
new data from workshops and surveys conducted by SIL International to show that Nurse’s conclusion
does not take into account the necessary dialectal information, namely that the Matumba dialect (which
Nurse uses) is phonologically distinct from the remaining Manda dialects. Lexicostatistical, phonological,
and sociohistorical evidence is taken into account. Further, in light of this new evidence, we propose two
likely historical scenarios, both of which posit that Manda is most closely genetically affiliated with the
Rufiji subgroup.
Contents
1
2
3
Introduction and background
Dialectometry and lexicostatistical evidence
Phonological evidence
3.1 Dahl’s Law
3.2 Spirant weakening
3.3 *NC̥> NC or N
4 Sociohistorical evidence
5 Synthesis and conclusion
Appendix A: Lexicostatistics
Appendix B: 296-item wordlist for corpus languages
References
iv
1
Introduction and background
Manda [ISO 639-3 code: mgs] is a Bantu language (N.11) spoken by the Manda and Matumba language
communities located in the area between Lake Nyasa and the Livingstone Mountains (Maho 2009). The
language area straddles two administrative regions in Tanzania: Njombe Region, north of the Ruhuhu
River, and Ruvuma Region, south of the river. The area is bordered to the north by Kisi (G.67), 1 to the
east by Pangwa (G.64) and Ngoni (N.12), and to the south by Matengo (N.13) and Mpoto (N.14) (Maho
2009). Lewis et al. (2015) report the Manda (and Matumba) population at 22,000. A previous SIL survey
puts the estimate even higher, between 25,000 and 40,000 (Anderson et al. 2003a). This paper uses
survey data from four other locations in addition to the Matumba variety in Luilo: Iwela, Lituhi, Litumba
Kuhamba and Nsungu (see map below). We examine three streams of evidence (lexicostatistical,
phonological, and sociohistorical) in this study to determine the closest genetic relatives of Manda (and
Matumba). New data is put forward from survey work conducted by SIL personnel in 2013 and
subsequent additional linguistic research (Gray, forthcoming; Gray and Mitterhoffer 2016).
1
The Guthrie codes for Bantu are referential, reflecting geography and not genetic relationship (see Schadeberg
2003: 146).
1
2
3
The Matumba consider themselves a separate ethnic group from the Manda, but still regard their
language to be essentially the same as Manda. The Matumba themselves claim that they were once
Manda who moved from the shores of Lake Malawi up into the mountains (Anderson et al. 2003a). The
prestige dialect (even according to the Matumba) of Manda is spoken in the area near Ilela and Nsungu,
villages on the lakeshore (Gray and Mitterhoffer 2016). Various comparative linguistic studies have
included Manda wordlists (Guthrie 1971, Nurse and Philippson 1975) and mention of Manda is made in
diachronic studies by Nurse and Philippson (1980) and Nurse (1988, 1999). None of these studies
mention the Matumba, so it is quite likely that the Matumba were unintentionally grouped with the
Manda in previous studies. As we will see in section 3, this lack of previous dialect research has given
rise to some misconceptions about the innovations that Manda shares with its neighbors.
Of those who have studied the Manda language with the aim of positing its closest genetic
affiliation, Nurse and Philippson (1980) and Nurse (1988, 1999) are in conflict with Ehret (1999, 2009).
Ehret (2009:17) puts Manda alongside Ngoni in the Rufiji-Ruvuma (RR) subgroup, whereas Nurse
classifies Manda within his Southern (Tanzania) Highlands (SH) subgroup. These are the main
hypotheses we evaluate in light of the new data which reflects a better understanding of Manda
dialectology. RR consists of two sub-branches. We are concerned primarily with Rufiji, which includes
Manda’s immediate neighbors: Ngoni, Matengo, and Mpoto. The SH subgroup includes the G.60
languages, of which we are primarily concerned with Kisi and Pangwa for the same reason.
This paper mainly focuses on the arguments put forward by Nurse (1988), since he deals with the
question of Manda’s genetic affiliation in the most depth. Ehret primarily relies on stem-morpheme
innovations (lexical/semantic evidence) in his (1999) work, and does not offer the same depth of
interaction with the corpus most like an SH language that Nurse does. From the Manda data Nurse uses,
Manda appears to behave phonologically mostly like an SH language. However, we argue that the
Manda data Nurse uses appears to be the Matumba dialect, which phonologically is quite different from
the other Manda dialects. In section 3 we see how the features of Dahl’s Law, spirant weakening, and
NC̥>NC or N cast doubt on Nurse’s argument for Manda as SH. We offer three scenarios regarding the
genetic history of Manda and Matumba in section 5.
In section 2 we present initial lexicostatistical evidence using dialectometry. Manda and Matumba
varieties are compared to the other corpus languages: Kisi, Matengo, Mpoto, Ngoni, and Pangwa. Section
3 examines the phonological evidence for the corpus languages, while section 4 briefly examines the
sociohistorical evidence. Section 5 concludes the paper with a synthesis, preliminary conclusions, and
possibilities for future research.
2
Dialectometry and lexicostatistical evidence
Dialectometry is essentially quantitative dialectology. Distance-based networks are examples of such
quantitative data explorations. A distance-based network analysis offers ‘‘an introductory visual means of
data exploration’’ (Pelkey 2011:279). Subsumed under the rising field of dialectometry are several
distance-based algorithmic applications that aim to help researchers explore language variation and/or
change, while “making it possible to show more than one evolutionary pathway on a single graph”
(Holden and Gray 2006:24). To create a distance-based network the opposite values of regular
lexicostatistical percentages are used within a standard matrix (e.g., the Kisi and Pangwa languages in
figure 2 are 0.59 similar, but 0.41 dissimilar; see Appendix A). The resulting distance matrix is then
subjected to the Neighbor-Net algorithm, as developed by Bryant and Moulton (2004) and implemented
within the Splits Tree 4 (4.11.3) software program (see Huson and Bryant 2010). If the lexical
relationships are ambiguous, the length of each branch indicates the confidence level of the split. Figure
1 shows such a network using four Bantu lects from southwest Tanzania (Roth 2011:43).
4
Nyika
Nyiha
Safwa
Malila
Figure 1. Sample network demonstrating ambiguity (adapted from Roth 2011:43).
In figure 1, the middle box indicates the ambiguous relationship: Split A—Nyika and Nyiha/ Safwa
and Malila versus Split B—Nyika and Safwa/ Nyiha and Malila. In figure 1, Split A is more likely. Holden
and Gray discuss other patterns and their meaning within the network diagram:
Rapid radiation may be inferred from a lack of phylogenetic signal, i.e. a rake- or star-shaped
phylogeny, whereas reticulation would indicate possible borrowing. Reticulations can also
pinpoint those languages which may have been involved in borrowing. Complex chains of
conflicting relationships involving numerous languages may indicate that borrowing occurred in
the context of dialect chains. (2006:24)
If the lexical relationships are more clear, the splits-graph looks more like a regular tree diagram
(see Holden and Gray (2006) for further explanation of the Neighbor-Net algorithm, and the unique
historical relationships between other Bantu languages).
Figure 2 below is based on the lexicostatistical data in Appendix A. The data includes lexical
percentages from 313-item wordlists elicited in five locations (Iwela, Lituhi, Litumba Kuhamba, Luilo
and Nsungu) during the 2013 dialect survey (Gray and Mitterhoffer 2016). This 313-item wordlist was
based on the 100-item Leipzig-Jakarta wordlist (Tadmor 2009). Two hundred ninety-six of the lexical
items 2 in the 313-item wordlist were compared to data from SIL Fieldworks Language Explorer (FLEx)
2
Some lexical items were omitted due to duplications in the data.
5
databases for Kisi and Pangwa, and CBOLD data for Mpoto, Matengo and Ngoni (supplemented by data
from Ngonyani 2003 and Yoneda 2006) using the comparative method as the basis for determining
cognacy. The lexical data (the 296-item wordlist) can be found in Appendix B. Consider the distancebased network below in Figure 2.
Matengo
Pangwa
Kisi
RR
SH
Mpoto
Ngoni
Lituhi
Litumba
Kuhamba
Nsungu
Luilo
(Matumba)
Iwela
Manda
Figure 2. Distance-based network for SH, Rufiji and Manda corpus languages.
From the diagram in figure 2, we can clearly see the split between the branch containing the SH
languages (Pangwa and Kisi) on the left and two of the Rufiji languages (Matengo and Mpoto) on the
right. We can also see that the Manda dialects, while lexically related to SH and Rufiji, are at the same
time lexically distinct, i.e., the Manda dialects seem to have a split-lexicon. Ngoni, while lexically more
like Manda, is still split between Manda and the SH languages. (Ngoni is generally classified as RR, but
its situation is similar to Manda’s, with the added complication of having an unclear relationship with
the South African Nguni, see section 4).
In sum, the lexicostatistical evidence for Manda’s closest genetic affiliation is ambiguous as Manda
has a split-lexicon. However, we can evaluate Nurse’s claim that Manda is “lexicostatistically a Rufiji
6
language” (1988:70) and say that this more recent wordlist data does not corroborate that claim. Section
3 proceeds with evaluating the phonological evidence that ties Manda to either SH or Rufiji (or both).
3
Phonological evidence
As we discuss in section 1, Nurse (1988) provides the most detailed argument regarding the genetic
affiliation of Manda. Table 1 (adapted from Nurse 1988:47) 3 is a comparison of what Nurse considers the
most relevant phonological features between SH, Rufiji, and Manda, in combination with conclusions
from our newer data in italics.
Table 1. Comparison of Manda phonological features with SH and Rufiji (adapted from Nurse 1988:47)
Feature
SH
Manda
Rufiji
Spirant-devoicing
(Proto-SH) yes
yes
yes (Proto-R)
Class 5 /li-/
(Proto-SH) yes
yes
yes (Proto-R)
NC̥
(Proto-SH) N(h)
N and NC
NC (Proto-R)
Dahl’s Law
(Proto-SH) yes
yes (minor)
no
Spirant weakening
no
no (yes)
yes (Proto-R)
Loss of preprefix
no
yes
yes (Proto-R)
Nurse concludes that Manda is an SH language primarily based on the phonological evidence
represented in this table. Nurse’s argument is as follows: (1) that SH and Rufiji differ in at least four
categories (lines 3–6); (2) of those four categories, Manda matches SH in two cases, maybe three (line 5
is ambiguous); (3) thus, if Manda is SH, the only line which needs explanation is line 6 (the loss of the
preprefix, or augment); whereas if Manda is Rufiji, lines 3–5 all need explanation. Nurse summarizes as
follows:
It is therefore simplest to posit Manda, not as an original member of Rufiji, but rather as an
original member of SH, which in recent times has undergone lexical change, and loss of preprefix,
inherited vowel length (and possibly tones) under the influence of Rufiji, which presumably means
Ngoni or Matengo (1988:48).
In this paper, we seek as much as possible to distinguish genetic inheritance from contact/areal
diffusion, as well as shared innovations from shared retentions. For example, the class 5 *li- feature is a
retention (compare Nurse 1988:30 and Nurse 1999:23), and a feature such as the loss of the preprefix
(augment) is historically unclear and more likely to have been spread by contact (see Nurse 1988:47).
Both are excluded from consideration here on this basis. Furthermore, the spirant-devoicing feature is
shared by both subgroups (and most likely necessarily preceded spirant weakening), and so does not
factor into the discussion (Nurse 1988:40, 43–44). That leaves Dahl’s Law, spirant weakening, and
NC̥>NC or N. Each of them are innovations, but they each also have their own unique considerations
regarding the possibility of contact influence.
As we see in the remainder of this section, the Manda data Nurse uses appears to be the Matumba
dialect. Even though Manda and Matumba are lexically similar (see section 2), Matumba is
phonologically quite different from the other Manda lects. This section provides the bulk of the evidence
for Manda and Matumba as members of either SH or Rufiji. We examine Dahl’s Law, spirant weakening,
and NC̥>NC or N in sections 3.1, 3.2 and 3.3, respectively.
3
The table is duplicated with the exception of the feature “nasal and spirant”. This category is excluded from this
study for two main reasons: (1) it is not a proto-SH or proto-Rufiji feature, and (2) it is already sufficiently covered
under the “spirant weakening” proto-feature (Nurse 1988: 46–47).
7
3.1
Dahl’s Law
Dahl’s Law is a voicing dissimilation in many Bantu languages in which the first of two voiceless plosives
becomes voiced, either within the root or across morpheme boundaries. Table 2 shows the examples of
Dahl’s Law in several of the corpus languages. These are the only examples found in the Manda corpus.
Nurse reports that there is no Dahl’s Law in Rufiji except for traces in Ngoni and Mbunga (P.15)
(1988:103). The examples in italics are the lexical items showing Dahl’s Law. The BLR3 Proto-Bantu
form is included in the table as well for comparison (Bastin et al. 2003). The lexical items not in italics
are various stems where we might expect Dahl’s Law to operate but where no evidence of Dahl’s Law is
found, despite both Proto-Bantu stem consonants being voiceless plosives.
Table 2. Dahl’s Law examples in the corpus languages
BLR3
*kàtɪ ́
*kèt
*kɪ ́t
*kópè
*kʊ̀ʊ̀kʊ́
*pàk
*pít
*tákò
*tátʊ̀
*tétè
Manda
pagati
kuketa
kukɪta
lukopi
kʊkʊ
kubaka
kubɪta
lidakʊ
-datu
lidete
Kisi
n/a
kukeːta
kukɪta
lukopi
kʊːkʊ
kubakala
kubɪta
lidaːku
-datu
lidete
Pangwa
pakatinakati
xuxeːta
xuxita
n/a
xuxu
xubaxa
xubita
n/a
xidatu
lidete
Ngoni
paɡatipaɡati
kuketa
kukita
(nɡ’opi)
n/a
kubakala
n/a
lidaku
-datu
lidete
English
middle
to shave
to do
eyelash
grandparent
to smear
to go
buttock
three
reed
The Manda dialects (including Matumba) share the same expression of Dahl’s Law: traces in the vast
majority of the same lexical items and not across morpheme boundaries. Of the SH languages, only
Hehe, Bena and Kinga have Dahl’s Law in all stems. Kisi and Pangwa show traces of it as does
Manda/Matumba, while Sangu and Vwanji show no evidence of it at all. Based on this geographic
distribution, it appears as though those languages with little or no trace of Dahl’s Law were on the outer
edge of an earlier dialect continuum where Dahl’s Law was present. Regardless, Kisi, Pangwa,
Manda/Matumba and Ngoni all share the same traces of Dahl’s Law.
There are two main possibilities that might explain the pattern of Dahl’s Law traces in
Manda/Matumba: (1) genetic relationship with SH with (a) inconsistent application and/or (b) early
phonological reversal (Nurse 1999:28), or (2) contact. Bantu languages can have inconsistent application
of Dahl’s Law due to when the feature became inactive (Batibo 2000). A phonological feature becoming
inactive can also be due to contact influences, as Batibo relates for the Sukuma/Nyamwezi languages
(F.20) in western Tanzania:
The major reason for this inactivity, and therefore incompleteness, may have been the instability
that both Sukuma and Nyamwezi experienced in their early years of resettlement due to the influx
of intruding groups…The influx of the intruding groups meant the influx of new lexical items by
speakers who did not have these rules in their language (2000:25).
Dahl’s Law is generally considered to be diagnostic of phylogenetic relationship. However, Nurse
says that “I would not claim that these features [Dahl’s Law, Bantu spirantization, *p-lenition] never
cross language boundaries but rather that they are more likely to be inherited in our languages from an
early stage of common development and thus historically diagnostic for the early period” (1999:20).
Thus, it is possible to conclude that traces of Dahl’s Law are due to borrowing. Considered by itself, what
the presence of Dahl’s Law traces means for the genetic affiliation (SH or RR) of Manda/Matumba is
inconclusive.
8
3.2
Spirant weakening
Spirant weakening, or lenition, is a process where spirants, such as /s/ and /f/, become weakened to /h/
or are lost entirely. Manda is again characterized by ambiguous data: a large number of lexical items still
contain /s/ and /f/, though some have weakened to /h/. Table 3 shows all the relevant examples from
the corpus of spirant versus /h/ forms in Matengo, Mpoto and Ngoni (RR), Kisi and Pangwa (SH), as well
as Manda.
Table 3. Spirant vs /h/ in corpus languages
English
gloss
fingernail
tear
egg
milk
name
leaf
grass
knot
cloud
dust
fish
ashes
to die
to choose
to get lost
to remain
to work
to pull
Rufiji-Ruvuma
Matengo
Mpoto
sô̠bu/hyô̠bu
chobu
lîːholi
maholi
lihóm
̠ bi
lihombi
másiba
lihiba
lihina
lihina
lihamba
lihamba
linyahi
linyai
lihundu
lihundwa
lihundi
maundi
luhô̠mbi
kuhombi
homba
homba
ilîhu
iliu
-ha
-wa
-hagula
-hagula
-hoba
-hoba
-higala
-higalila
-henga
-henga
-huta
-huta
Ngoni
luhyuwu
maholi
n/c
n/c
lihina
lihamba
linyahi
lihundu
lihundi
luhumbi
somba
n/c
-fwa
-hagula
-hova
-higala
-henga
n/c
Manda
luhʲuwu/lufʲuwu
liholi/ lisoli
lihombi
lisiva
lihina
lihamba
liɲahi / liɲasi
lihundu/lifundu
lihundi / lifundi
luhumbi/lufumbi
somba
lifu
-hwa/ -fa/ -fwa
-hagula
-hova /-sova
-higasa / -sigasa
-henga
-huta / -futa
Southern Highlands
Kisi
Pangwa
n/c
n/c
n/c
n/c
lifʊmbi
n/c
n/c
n/c
n/c
n/c
n/c
n/c
linyasi
n/c
(liputu)
lifundo
n/c
n/c
(lifu)
mafumbi
somba
somba
n/c
n/c
-fwa
-fwa
-haghula
n/c
n/c
n/c
-sighala
-sikala
n/c
n/d
n/c
n/c
Of the corpus languages, Matengo, Mpoto and Ngoni (all RR) generally use the weakened /h/ forms,
while Kisi, Pangwa (SH) and the Manda lects of Iwela, Nsungu, and Luilo (Matumba) generally preserve
/s, f/. The Lituhi and Litumba Kuhamba lects are in between, both geographically and in the type of
forms they use. The Ruhuhu River separates these Manda dialects from the others.
The modern-day distribution of spirant weakening among the corpus languages fits that of a wave
(see Mühlhäusler (1996:10) for a summary of Bailey’s dynamic wave model in particular). Thus, the
variation within Manda (and to some extent its neighbors) is not entirely unexpected. The distribution of
spirant weakening could be explained by one or both of two factors: (1) inconsistent application of the
phonological change due to the varied timing of its inactivity, and (2) historical contact. As we saw in
section 3.1, contact influences can cause a phonological innovation to become inactive, which is how
both inconsistent application and historical contact might have come into play together. If (1) were true,
Proto-RR could not have already undergone spirant weakening, as inconsistent application across some
lects would presuppose its daughter languages had already started to expand and diversify. Apart from
any other evidence, it would also mean that in all likelihood Manda’s closest genetic affiliation would be
the Rufiji subgroup due to its shared spirant weakening innovation with Matengo, Mpoto, and other
Rufiji. However, there is no way to eliminate the possibility of contact/areal diffusion. In regard to (2),
because of the variation in his own Manda data regarding spirant weakening, Nurse (1988:47) says that
it “could be explained in either direction by simple borrowing of lexis carrying either spirant or /h/”. In
9
sum, spirant weakening in the corpus languages remains inconclusive in terms of helping to determine
the closest genetic affiliation for Manda.
3.3
*NC̥> NC or N
The phonological process whereby the historical sequence of a nasal followed by a voiceless homorganic
plosive (*NC̥) has become either a nasal followed by a voiced plosive (NC) or a nasal by itself (N, in
which the consonant has deleted altogether) is represented here as *NC̥> NC or N. Once again, the
problem is that we need to try to distinguish inheritance/innovation from contact. Nurse explains the
historical process as follows:
We assume here that inherited sequences of nasal plus voiceless homorganic stop have undergone
one of two major processes: either the stop is voiced…or the sequence is maintained, which,
however, can lead to devoicing of the nasal, which in turn can lead to N, Nʰ, C̥ʰ or C̥. Only the
proto-languages for [RR] and Nyakyusa/Ndali show the stop voicing, whereas all the others show
one or other form of the second process. (Nurse, 1988:31)
These processes constitute two very different pathways, and the issue is that we generally see
*NC̥>NC in Manda and *NC̥>N in Matumba. Kisi and Pangwa have N like Matumba, while Matengo
and Mpoto have NC like the rest of Manda.
This is also an area where Nurse (1988) seems to use the Matumba dialect as normative for the
entire language. In table 4, this is represented with the CBOLD label, which is the data that Nurse and
Philippson collected. Table 4 shows all of the relevant reflexes of *NC̥ across Manda and Matumba from
the corpus, and it also includes the CBOLD data and BLR3 references.
Table 4. Class 9/10 across Manda/Matumba dialects
PB stem
(BLR3)
English
gloss
Lituhi
Litumba
Kuhamba
Nsungu
Iwela
CBOLD
Manda
ŋguku
mbepo
Luilo
(Matumba)
ŋuku
mepu
*-kʊ́kʊ́
*-pépò
‘chicken’
‘cold’
ŋguku
mbepʊ
*-kópè
*-kómb?
‘eyelash’
‘finger’
ŋgoɲɟi
n/c
n/c
*-ntù
*-ntù
‘goat’
‘lightning’
‘person’
‘thing’
mene
mbamba
mundu
kindu
*-jónk-
‘suck’
-joŋga
ŋguku
mepʊ/
mbepʊ
ŋoɲɟi/
ŋgoɲɟi
mene
mamba
mundu
kindu /
kinu
-ɲoŋga
ŋguku
mbepu
ng’oko
mepu
ŋgopi
(fiŋgoɲɟi)
ŋopi
ŋgoɲɟi
ŋgopi
ŋgoɲɟi
(lukopi)
(lukonji)
mene
mbamba
mundu
kindu
mene
mamba
munu
kinu
mene
mbamba
mundu
kinu
mene
mamba
munu
kɪnu
-joŋga
-ɲoŋga
/ -ɲoŋa
-joŋga
-ɲoŋa
We can clearly see from table 4 that the CBOLD data patterns most closely with the Luilo
(Matumba) dialect. To a certain extent there are examples of both N and NC in each dialect (often in the
form of lexical doublets), but the overall pattern remains.
Some of the examples from table 4 could possibly be looked at as the after-effects of either Dahl’s
Law or Meinhof’s Rule/Law (Ganda Rule/Law). Since positing Dahl’s Law in class 9/10 roots is not
supported by evidence from any other noun or verb roots, it is very unlikely that Dahl’s Law is what is at
play here. Meinhof’s Rule would be more likely as it resembles the process *NC>N or NC. Meinhof’s
Rule is a process in which a prenasalized voiced consonant elides when followed by a nasal or
10
prenasalized consonant in the subsequent syllable, leaving behind either a simple or geminate nasal.
Gray (forthcoming) describes in her Manda phonology sketch that it is not Ganda Law at work, but
rather the post-nasal consonant is deleted across the board consistently for classes 1 and 3. There are
many words where there is no NCVNC sequence, and yet C1 is deleted (for class 1 and 3), such as *ndala/nala, and *n-gosi/ŋosi, etc. There are also lots of NCVN(C) sequences that do not become NVNC:
ndemba (hen), ngondo (quarrel/war), mbanda (slave), ndongo (relative), ndomondo (hippo).
The *NC̥> NC or N evidence in this section is not the silver bullet of clarity to Manda’s genetic
relationship to either SH or RR. However, this new research into the *NC̥ process does pose a serious
difficulty in any argument for Manda as a SH language. Possible rejoinders to the argument that Manda
is not SH on the basis of *NC̥ reflexes are that (1) perhaps Manda and Matumba should not be
considered a single historical language variety, in which case Matumba could certainly be SH, or (2)
perhaps Manda/Matumba is SH and all the Manda dialects except for Matumba were in contact with RR.
We explore these possibilities in our synthesis and conclusion in section 5, after we examine the
sociohistorical evidence in section 4.
4
Sociohistorical evidence
While oral traditions do not necessarily prove historical origin, it is worth taking into account what the
Manda and Matumba say about their own historical origins. Data gleaned from group questionnaires
taken during the 2013 SIL dialect survey of the Manda/Matumba area reveal that the Manda and
Matumba have mixed views of their origins. Several groups claim the Manda are partly descended from
the Nguni tribe of South Africa; others claim that they were from Malawi or from the Songea region. 4
Some claim they are partly descended from the Pangwa and partly from the Matengo language
communities. We will briefly comment on these claims.
Ngonyani (2003:1), in his explanation of the origins of the Tanzanian Ngoni people, states that the
Ngoni people incorporated many indigenous inhabitants of the area from different language groups when
they moved into the highlands east of Lake Nyasa. Ngonyani includes Manda in the list of language
communities that the Ngoni incorporated. This statement implies there was at least a group called
Manda living there at the time the Ngoni invaded, which seems to run counter to claims that the Manda
are descended from the Ngoni. Linguistically, Nurse (1999:13) claims that the Ngoni themselves
abandoned the Nguni language in favour of the local languages in the nineteenth century when they
invaded from the south. Nurse (1988) also made this claim based on the absence of connection between
Tanzanian Ngoni and South African Nguni languages. It would seem that the invasion left few linguistic
traces from Nguni (Nurse 1988:48), and that the language of the invaders themselves, now known as
Tanzanian Ngoni, was more affected by the invasion than the neighboring groups were. However,
according to the original SIL sociolinguistic survey (Anderson et al. 2003a), the Manda/Matumba
themselves claim to understand Ngoni very well. This holds true even for those villages furthest from the
Ngoni language area, with some saying that even children can understand Ngoni.
In regard to the claims about the historical relationship to Pangwa, Anderson et al. (2003a:9) state
that “historically, people of various ethnic groups (mostly Pangwa) migrated from various areas into the
region along the coast (of the lake)”. According to this history, the Manda would be most related to the
Pangwa; however, the people’s own perception of the relationship between the languages contradicts
these origins. The Pangwa survey report (Anderson et al. 2003b) confirms that currently at least, the
Manda and Pangwa people groups have little contact, and that the Pangwa understand very little of the
Manda language. The Manda/Matumba themselves claim that there is “no language relationship
whatsoever between Manda and Pangwa” (Anderson et al. 2003a:8), even for those villages closest to the
Pangwa language area. The intelligibility of Kisi seems more disputed; some villages claimed that there
was little resemblance, but some (Matumba villages) claimed to understand Kisi relatively well. All
groups that were interviewed during the survey claimed that the lexicon is similar and the difference lies
4
This may also refer to Nguni origins since that is where the Tanzanian Ngoni settled.
11
in pronunciation. Interestingly, the Manda village called Iwela, which has had more contact with Kisi
than most of the rest of the Manda area, is considered by some other Manda (Matumba) villages as a
place where the people are not ethnically Manda and do not speak the Manda language. The people of
Iwela themselves claim to be Manda, but many features of their dialect show the influence of the Kisi
language on it. Nurse (1988:70) claims that Kisi has been heavily affected by the N.10 languages, which
may be a factor influencing the higher intelligibility with Kisi than with Pangwa.
Regarding the possible connection with Matengo, the findings in the 2013 dialect survey were that
for those villages south of the Ruhuhu River in the Ruvuma region there is considered to be little
intelligibility between Mpoto and Manda and even less between Matengo and Manda.
In summary, the sociohistory of the Manda language community is still unclear. General feeling
among the Manda/Matumba would connect them most closely to the Ngoni. However, Ngoni appears to
have had merely a superstratum influence on Manda and other language communities in the area, and so
this would not be an indicator of close genetic affiliation. For Pangwa, one would expect the
comprehension with Manda to be higher if the Manda had indeed descended from Pangwa as reported;
however, comprehension is quite low as the Livingstone Mountains create a barrier between the Manda
and the Pangwa language areas. As an added note, the accessibility of the Kisi area by canoe gives the
Manda/Matumba more opportunity for contact with Kisi.
We now turn to synthesizing the lexical, phonological, and sociohistorical evidence, and developing
some conclusions, specifically in interaction with the results regarding Manda’s historical relationships
from Nurse (1988). We will also briefly explore ideas for further research.
5
Synthesis and conclusion
So far, this paper has considered three main streams of evidence (lexicostatistical, phonological, and
sociohistorical) in working towards determining whether Manda is most closely genetically affiliated
with SH or RR. In isolation, all of the streams have been ambiguous and so far inconclusive. In this
section, we seek to synthesize these streams of evidence and arrive at some tentative conclusions.
In section 3.3 we encountered evidence that led to the possibility that we should consider Manda
and Matumba historically separate languages. The *NC̥> NC or N feature was found to be the significant
phonological difference between the Matumba/Luilo dialect (*NC̥>N) and the remaining Manda dialects
(*NC̥>NC). In this view, Manda could be historically RR while Matumba could be SH, solving the most
pressing difficulty. There are at least two arguments for the historical unity of Manda and Matumba as
dialects of one language: (1) the lexicostatistics bear this out (see section 2 and Appendix A), and (2) the
testimony of Manda and Matumba native speakers themselves.
Frankly, however, this is not enough evidence to eliminate the option that Manda and Matumba
were historically separate languages that originally came from different genetic subgroups. If this were
indeed the case, it would not change the fact that Matumba can certainly be considered a modern-day
dialect of Manda. In regard to (1), as important as lexicostatistics can be, they “can only describe and
extend relatedness but cannot establish it” (Nichols 1996:64). The same can be said of native speaker
testimony (see Hinnebusch 1999:179). The gold standard of evidence for genetic relatedness is shared
innovations, specifically what Nichols (1996) calls ‘individual-identifying’ evidence. The only such
evidence that could be considered ‘individual-identifying’ in this paper is the *NC̥ feature, which
happens to cut across Manda and Matumba. Part of the reason for the lack of shared innovations could
be rooted in the history of these Bantu subgroups. Nurse says of SH that there is “a relative lack of really
distinctive innovations. None of the innovations preceding […] is unique, all being shared with some
combination of surrounding groups, which raises the possibility that the innovations might be the result
of areal spread, or that the Proto-SH period was short, not allowing time for innovation” (1988:40).
Rufiji is in a similar situation (see Nurse 1999:31).
Normally with such a lack of shared innovations, we could rely even more on paradigmatic
grammatical evidence (see Nichols 1996), especially for establishing Matumba historically as a dialect of
12
Manda. But again, the evidence remains ambiguous and inconclusive. We can see this in the clear
differences between the personal pronoun set(s) in Manda and Matumba in table 5.
Table 5. Personal pronouns in Manda and Matumba (Gray 2016:148)
Person
1SG
2SG
3SG
1PL
2PL
3PL
Manda (Nsungu)
nenga
wenga
jʊla
tenga
mwenga
vala
Matumba
nene
veve
mwene
tete
nyenye/mwemwe
vene
Obviously, this is just one example of a paradigmatic grammatical set in which Manda and
Matumba are in disagreement. More research is needed on the grammar of the Matumba dialect in
particular. What we do know is that Manda and Matumba indeed share much in common phonologically
and grammatically. That Matumba is a modern-day dialect of Manda is not in dispute. The issue is that
we still cannot dismiss the possibility that historically Manda and Matumba did not come from the same
recent ancestor—that Manda belonged to the Proto-Rufiji subgroup, while Matumba was a member of
Proto-SH. This would entail that Matumba underwent massive relexification, to the point where today
the lexicostatistics are indistinguishable. Matumba must have also adopted many grammatical elements
from Manda. Under a long period of extreme contact, none of this is out of the question in a
geographical area where SH and Rufiji collide. What of the other possibilities?
If we set aside the option that Manda and Matumba come from different subgroups historically, we
are left with Manda/Matumba as either SH or Rufiji. We saw in section 2 that Manda/Matumba
essentially has a split-lexicon between SH and Rufiji, which does not support one subgroup over another.
The argument for Manda/Matumba as SH centers around the traces of Dahl’s Law. Under this scenario,
spirant weakening is adequately explained due to contact with Rufiji languages to the south. The *NC̥
feature is harder to explain, but not impossible. Its distribution too would have to be the result of
contact: *NC̥>N in Matumba represents the original SH feature, while *NC̥>NC in the rest of Manda
would be due to contact with Rufiji. The question under this scenario: Why didn’t Matumba adopt NC like
the rest of Manda? It would have to have been in contact with Rufiji languages (or the other Manda
dialects with NC) to explain the spirant weakening pattern.
The argument for Manda/Matumba as Rufiji centers around the majority of Manda dialects showing
*NC̥>NC and the spirant weakening pattern representing inconsistent application of the innovation.
Dahl’s Law traces are explained by the borrowing of individual lexical items from SH. Given the small
amount of the Manda lexicon that has been affected by Dahl’s Law and those traces being in common
with the neighboring SH languages, contact influence is reasonable despite Dahl’s Law normally being
diagnostic in other branches of Bantu. *NC̥>N in Matumba is due to contact with SH. In this scenario,
this borrowing (or reversal) only happens in one dialect instead of several. Dahl’s Law lexical items
would have diffused much earlier. Nurse also mentions for Rufiji, “an apparently unique set of
allomorphs for the /-ile/ suffix” (1988:45). Manda appears to have these allomorphs (Gray 2016:108),
but although their geographical distribution has been clarified (e.g. Nurse 2008:267), it is still unclear
whether these /-ile/ allomorphs truly would distinguish Rufiji from SH, but needs to be explored further.
Of these three different options, it is difficult to choose between the first scenario of Manda and
Matumba as historically different languages and the third scenario of a united Manda/Matumba as
Rufiji. Both are not just possible but plausible, even though they have their drawbacks and relatively
unusual historical journeys. The second scenario with a united Manda/Matumba as SH seems much more
unlikely. Further research is needed on morphosyntactic innovations (e.g. tense/aspect) and tonal data
for the corpus languages, especially detailing any further differences in the Matumba lect.
13
In this paper we have explored lexicostatistical, phonological, and sociohistorical evidence in the
goal of determining the closest genetic affiliation for Manda (SH or RR). Much of the evidence was
ambiguous and inconclusive, but in this section we were able to put together three viable scenarios. Two
of those scenarios we found much more likely than the remaining option, and both refine our
understanding of Manda’s history and dialectology. We primarily interacted with Nurse (1988) who had
given the most in-depth previous account. Most crucially, it appears that the dialect used by Nurse for
Manda is actually Matumba, which is phonologically different from the Manda dialects, especially in
regard to the reflexes of *NC̥. Regardless which one of the likely options correspond to historical reality,
both posit the majority of Manda dialects as historically Rufiji, not SH.
Appendix A: Lexicostatistics
Lituhi
75
Litumba
Kuhamba
77
67
Nsungu
73
76
71
Luilo
75
73
77
78
Iwela
53
54
53
58
58
Kisi
43
44
45
50
48
59
Pangwa
65
65
60
67
63
52
45
Ngoni
47
44
48
43
48
37
26
43
Matengo
53
45
51
42
48
33
26
43
71
14
Mpoto
Appendix B: 296-item wordlist for corpus languages
The wordlist data transcriptions are replicated from the original source databases (see §2), except for the Manda and Matumba varieties which
are in IPA. Tone is generally not included due both to the nature of the source databases and the rapid survey word-collection conducted by SIL.
1
English
Gloss
eye
2
eyelid
3
ear
4
mouth
5
*jaw
6
nose
7
*chin
8
beard
9
tooth
10
Lituhi
líhu - míhu
Litumba
Kuhamba
liho - mihu
Nsungu
Luilo
Iwela
Kisi
Pangwa
Ngoni
Matengo
Mpoto
lihu - mihu
lihu - mihu
lihu - mihu
liːhu
liho
lîhu - mîhu
liu
ŋgopi
ŋopi
ŋgopi
lukopi
ng'opi
ingopi
ingopi
likutu makutu
ndomo milomo
likutu makutu
ndomo milomo
likutu makutu
ndomo milomo
ng'ope cha
mumiho
lihu
mbʊlʊkʊtʊ
mbulukhutu
likutu; ɲɟɛvɛ
ndomo
mlomo
m̩ lɔmɔ
fíkupalɪlɪ kíkʊpalɪlɪ
líkutu mákutu
ndomo milomo
fikupulila kikupulila
líkutu makutu
ndomo milomo
ɲɟeɟe
ɲɟeɟe
ɲɟege
tili
taʝa
lucheeche
njeje
mbúnu mbunu
kiɲɟwemba fiɲɟwemba
mbunu /
meŋelu
litama matama
mbunu
mbunu
mbunu
mheŋelu
meng'elo
mɛŋɛlu
khilefu
ciɲɟwɛmba
kíleu - íleu
kileu - ileu
lúnde(b)u índe(b)u
indeo
lîno - mîno
lino
lúlimi - ínimi
lulimi
kiɲɟwemba - kiɲɟwemba - kiɲɟwemba - kiɲɟwemba kidefu
fiɲɟwemba
fiɲɟwemba fiɲɟwemba fiɲɟwemba
mlefu
maɲɟwɛmba
linu - minu
liɲɟwemba maɲɟwemba ndefu
maɲɟwemba
ndefu
línu - mínu
liɲɟwemba maɲɟwemba
linʊ - minʊ
linu - minu
linu - minu
liːnu
lino
linu
tongue
lulimi
lulími
lulimi
lulimi
lulimi
lulimi
lulimi
lulimi
11
head
mútu - mítu
mútu - mítu
mutu - mitu mʊtu - mitu mutu - mitu mutu
mutwe
mutu
12
hair (of
head)
lijúɲɟu majúɲɟu
lijuɲɟu majuɲɟu
ljuɲɟu majuɲɟu
ɲɟwili
njwili
ɲɟwili
liléme maléme
ŋkúfu mikúfu
liléme maléme
nkufu mikufu
lileme maleme
ŋkufu mikufu
siŋgu
singo
siŋgu
13
neck
14
stomach
15
navel
liɲɟwémba maɟwémba
siŋgu
siŋgʊ
siŋgʊ
lijuɲɟu
siŋgʊ
lileme
ŋkufu
lijuɲɟu majuɲɟu
siŋgʊ
lileme maleme
ŋkufu mikufu
15
lileme maleme
lileme maleme
nkʊfʊ
mdungu
lileme maleme
lukufu
likûtu makûtu
ndomo mílomo (lip)
lugômu –
ingômu
ímbulu ímbulu
umûtu mimûtu
likutu
kukano
lugomo
imbulu
mmutu
líjunzu májunzu
lijunju
hîngu-hîngu
hingo
lutumbo
lutumbu
nkuhu míkuhu
nkuu
16
16
English
Gloss
back (of
body)
Lituhi
ŋoŋgo migoŋgo
17
*knee
lifugumíɾu mafugumíɾu
18
leg
kigʊlʊ magʊlʊ
19
foot
20
arm
21
22
palm (of
hand)
*finger
nail
Litumba
Kuhamba
ŋoŋgo migoŋgo
lifugumilu –
mafugumilu /
ligoti
kigʊlʊ magʊlʊ
lwaju ɲaju
kiwóko mawóko
kiwóko mawóko
luhjuwʊ hjowʊ
lúkoɲɟi ŋgóɲɟi
lwaju ɲaju
kiwóko mawóko
lukoɲɟi ŋgoɲɟi
Nsungu
Luilo
Iwela
Kisi
Pangwa
Ngoni
Matengo
Mpoto
ŋoŋgo
ŋoŋgo
ŋoŋgo migoŋgo
ɲuma
mkongo
mgɔŋgɔ; msana
ngongu mígongu
nngongo
lifugumɪlʊ - lifugumɪlʊ - lifugumɪlu goti
mafugumɪlʊ mafugumɪlʊ mafugumɪlu
lifukamilo
lifugamilu
ligóti - mágǒti lijugwa
kigʊlʊ figʊlʊ
kigʊlʊ
kigʊlʊ figʊlʊ
kigʊlu figʊlu
kilʊndi /
kiɣʊlʊ
likulu
cigendelu
luɲajo
kiɣʊlʊ /
kilʊːndi
likulu
cigɛndɛlu
kiwoko mawoko
kiβoko
livokho
ciwɔkɔ
kigaɲɟa
kiɣaɲɟa
khikanja
cigaɲɟa
lukoɲɟi
lukhonji
luhyuwu
kikoɲɟi kja
kiβoko
khivokho
lukɔɲɟi
ngwapanilu
kiwoko mawoko
kigaɲɟa figaɲɟa
lwaju ɲaju
kiwoko mawoko
kigaɲɟa figaɲɟa
fjuhʊ
fjʊgʊ
ŋgoɲɟi
lukoɲɟi ŋoɲɟi
kikoɲɟi fiŋgoɲɟi
lukoɲɟi fikoɲɟi
lukoɲɟi ŋgoɲɟi
ŋgoɲɟi fikoɲɟi
ŋgwápa
ŋgwapa
ŋgwapa
ŋgwapa
ŋgwapa
ŋ'hwapa
mkhwapa
likúmba mákumba
lihúpa mahúpa
likúmba makúmba
lifupa mafupa
likumba makumba
liɟege maɟege
likumba makumba
lifupa mafupa
ŋgosi ja
mundu
sonda ya munu cikumba
lifupa
licheke
liʄɛgɛ
mwasi
mwasi
mwasi
danda
ŋasi
mwâhi
maholi
mahosi
luhosɪ
linyochi
mahɔli
lî:holi - máholi maholi
lifuki mafuki
kibindi fibindi
kuhʊma
lifuki
-huma ijashu
mfutumalu
lihógatela máhogatela
-
mtima
cibindi
lî:mani - ímani imani
lilɔvi
lîlobi - málobi
liŋɛlu
ling’élo
23
finger
24
armpit
25
skin
26
bone
27
blood
mwasi
28
tear
maholi
29
sweat
30
liver
31
voice
32
horn
lihuki mahúki
kibíndi fibindi
liŋelu maŋelu
lwaju
likumba makumba
lifupa mafupa
mwasi mwasi - mjasi
mjasi
lihõli lisoli maholi
masoli
lihũki lifuki
mahúki
liini ntima maini
mitima
ɲɟwaŋgu /
ɲɟwaʋila
liŋélʊ maŋélʊ
lifuki
kibindi fibindi
ɲɟoʋelu
liŋelu maŋelu
liːsu
liŋelʊ maŋelʊ
liŋelu maŋelu
lupembe
lipembe /
malamala
lwâgi - hyâgi lwâju - hyâju
kúboku máboku
kíganza íganza
ligolo
sôbu - hyôbu
lúgonzi ngônzi
ngwâpa ngwâpa
híkanda ngânda
lî:hupa máhupa
ligolo
lib'oko mabhoko
kig'anja
chobu
lukonji
ingwapa
likambila
kihupa
mwai
lihogutela
liɲelu
17
English
Gloss
Lituhi
Litumba
Kuhamba
lipapanílʊ mapapanílu
33
wing
lipapanílu mapapanílu
34
feather
liŋgóma maŋgóma
35
tail
ŋkɪ ́la - mikɪ ́la
36
egg
lihʊ́mbi mahʊ́mbi
37
honey
38
milk
lukáma
lukáma
lisiʋa
39
meat
ɲama
ɲama
40
person
múndu ʋándu
mundu ʋandu
41
man
ŋósi - ʋagósi
ŋosi - ʋagosi ŋosi - ʋagosi ŋosi - ʋagosi ŋosi - ʋagosi munagoːsi
42
husband
ŋgwana
waŋgu
43
woman
nala ʋadala
44
wife
muhana
waŋgu
45
father
(my)
dádi ʋadádi
dadi ʋadadi
46
mother
(my)
mawu ʋamawu
47
elder
brother
kaka
mawu ʋamawu
ndʊmbʊkaja ʋalʊmbʊkaja
/ kaka ʋakaka
liŋgóma maŋgóma
Nsungu
Luilo
Iwela
Kisi
Pangwa
Ngoni
Matengo
lipapatilo
ligwaba;
lipapanilu
kípapatila
liʝoʝa
likala
liŋgɔma
lingôma
mangoma
nkɪla
mkhila
mkila
ń̩ke̠la
nkela
lihʊmbi
lifʊmbi
likhaang'a
likaŋa
lihó̠mbi
lihombi
wuki
βwʊki
vuukhi
wuci
búsi
busi
lisiʋa
lisiʋa
lukama
machiva
lukama
másiba
lihiba lihibha
njama
njama
ɲama
ɲama
nyama
ɲama
ínyama
inyana
mundu ʋandu
munu - ʋanu
mundu ʋandu
mundu
munu
mundu
mûndu - bându mundu
mwadaada
m̩ gɔsi
mwánalomi ákanálomi
lipapanílu - lipapanilu - lipapanilu kibabatilu
mapapanílu mapapanilu mapapanilu
liŋgoma maŋgoma
ŋkila ŋkɪla - mikɪla
mikila
lihʊ́mbi lihʊmbi mahʊ́mbi
mahʊmbi
liŋgoma maŋgoma
ŋkɪla mɪkɪla
lihʊmbi mahʊmbi
wutʃi
ŋgwana
ʋaŋgu ʋaŋgwana
ʋaŋgu
nala ʋadala
ŋgwana
waŋgu ʋaŋgwana
ʋaŋgu
nala ʋadala
muhana
ŋʊli ʋaŋgu - waŋgu ʋagʊli ʋaŋgu ʋahana
ʋaŋgu
liŋgoma maŋgoma
ŋkila mikila
Mpoto
mwanalomi
ŋgwana
ʋaŋgu
ŋgwana
waŋgu
ŋgoːsi
mkoosi
mŋgwana
ńdomi - álomi,
ndomi
akaálomi
nala ʋadala
nala ʋadala
munadala
mwayuva /
mkhidala
mdala
mbômba - áka
mbomba
mbômba
ŋoli ʋaŋgu
ŋoli waŋgu
ndala
mdala
mdala
ńhanu, áhanu uhwano
ákaáhanu
dadi ʋadadi
dadi
dadi
daːdi
daadi
dadi
atâti - ákatâti áhěngolo atati b'ango
ákahéngolo
mau ʋamau
mau
mawu
maːβu
yuuva
mawu
amâbu ákamâbu
muhatʃa ʋahatʃa
kaka
kaka
mama
mhaacha
kaka; mkulu
mbeli - ábeli nkoano
ákaábeli
amabo
18
English
Gloss
Lituhi
Litumba
Kuhamba
Nsungu
Luilo
Iwela
Kisi
Pangwa
48
mother’s
brother
mɟomba
mwipa ʋipa
mwipa ʋipa
jaja
jaja
ʝaːʝa
muyaaya
49
child (my) mwána - ʋána
mwana ʋana
nseja ʋaseja
mwana
mwâna - bâna mwana
nséja ʋaséja
mwana ʋana
nseja ʋaseja
umwana
elder
mwana ʋana
ŋgogolo ʋagogolo
mwana
50
mwana ʋana
nseja ʋaseja
ŋgogolo
sehe
mkokolo /
mkoyo
vaɲaluhala;
gɔgɔlɔ
nsêja - asêja
51
grand
parent
hukulu
ŋgeɲɟi
mkenji
mgɛni; mhɛɲɟa;
ńgeni - ágeni
m'yehi
53
chief
mbaha witu mfalme
mutwa
n'kosi
54
god
tʃapáŋga
Muŋgu
muŋgu
muluŋgu
muŋgu
linguluvi
cimluŋgu
55
sickness
ilwala
ntamu
tamu
wound
utamu ʋatamu
kilonda filonda
tamu
56
utámu mátamu
kilónda filónda
sʊkʊlʊ ʋasukulu
wakuhitʃa ʋakuhitʃa
mutwa ʋatwa
muluŋgu /
kjuta
utamu matamu
kilonda filonda
khukhu
guest
sʊkʊlʊ ʋasʊkʊlʊ
ŋgene ʋajene
ɟumbi ʋaɟumbi
kʊːkʊ
52
húkuɾu ʋahukulu
ŋgéne ʋajene
mútwa ʋátwa
kilonda filonda
kilonda
kilonda
khivamba /
khikhoong'a
cilɔnda
57
medicine
mitiʃamba
ŋoda
ŋgoda
mkoda
mtɛla / migɔda ńtěla - mítěla
58
body
mbɪlɪ
mbili
mβɪlɪ
mvili
mhwili; mvili
hyêga - hyêga yega
59
name
lihina mahina
lihina mahina
litaːβu
litaawa
lihina
lihína - máhĭna lihina
60
lihowo mahowo
lihowo mahowo
lififi
lififi
lihɔgɔ
lipúndwa
lipundwa
lipusi mapusi
lisuʋi masuʋi
lihimba mahimba
lipusi mapusi
suʋi masuʋi
lihimba mahimba
liɲalu
limisi
kamlamu /
kapusi
pûsi mbunyâli
pusi mbunyau
sʊβi
liduuma
cihuvi
kího̠bi
kihobi
lisɪmba
libonji
lihimba
líhimba
lihimba
nténde miténde
mbɪlɪ miʋɪlɪ
lihína mahina
ŋoda migoda
mbɪlɪ miʋɪlɪ
lihína mahína
*hyena
lihówo mahowo
lihowo mahowo
61
cat
lipúsi mapúsi
62
leopard
63
lion
lipusi mapusi
kihʊʋi kihʊ́ʋi - fihʊ́ʋi
fihʊʋi
lihímba lihimba mahímba
mahimba
ntende mitende
mbɪlɪ miʋili
lihina mahina
lihowo mahowo /
libondela mabondela
limbuɲali mambuɲali
suʋi masuʋi
lihimba mahimba
kʊku
ŋgene ʋajene
sukulu /
kʊkʊ
ŋgene ʋajene
Ngoni
Matengo
Mpoto
ánijăja ákaníjăja,
ákiníjăja
nseja
ahokolo
bambu
nkolongu
ń̩do̠ngu
tʃapaŋga
utâmu
-lwala
kílonda ílonda
mitela
19
English
Gloss
Lituhi
Litumba
Kuhamba
64
*fruit-bat
liniminími maniminími
liniminimi maniminimi
65
termite
likele
66
termite
hill
67
Nsungu
Luilo
Iwela
Kisi
Pangwa
kiduli fiduli
ligeke mageke
kidúli fidúli
lindʊlindʊliliniminimi - liniminimi mandʊlindʊ lindʊlindʊli khindulinduli
maniminimi maniminimi
li
ligeke ligeke likele
muhwa
mmehe
mageke
mageke
kisʊɣulu kiduli kisugulu kihwa
khisukulu
fisuɣulu
fiduli
ʋisugulu
goat
mene
mene
mene
mene
mene
mhene
mene
68
cow
ŋómbi
ŋombi
ŋombi
ŋombi
ŋombi
ŋgolombi
69
bull
likambaku makambaku
likambaku makambaku
70
calf
litóli - matóli
likambako makambako
litoli matoli
likambaku likambaku
makambaku
litoli ŋombi
matoli
ndala
ŋguku
ŋuku
ŋguku
liɟogolo maɟogolo
kidege madege
lisatu masatu
liɟogolo maɟogolo
lidege madege
lisatu masatu
liɟogolo maɟogolo
lidege madege
lisatu masatu
somba
somba
ɲɟutʃi
lisosólo masosólo
liʋémbe house fly
maʋémbe
tree
71
72
73
74
75
76
77
78
79
louse (of
body)
ŋgʊ́ku
lisosolo masosolo
liʋembe maʋembe
ŋkoŋgo mikoŋgo /
ukoŋgo /
libihɪ - mabihi libɪhɪ mabɪhɪ
Matengo
cinimánima
kínǐma
likɛlɛ
ń̩keki
ciduli; cidulu
mɛnɛ
Mpoto
limea
kikula
imbuhi
imbui
isenga indala /
ŋɔmbɛ
mbukuma
ng’ômbi
ng'ombi
likida
likambaku
ng’ombi
lipôngu
ŋombi
khikwada
litoli
litoli
litoli
ŋhʊkʊ
ng'uukhu
ŋguku
íngo̠ku
ingoko
liɟogolo
lichokolo
liɟogolo
likóngobi
likoŋgobi
kiʝuni
khideke
cidɛgɛ
kíjuni
kijuni ijuni
lisatu
lihatu
lisatu
lǐhátu
lihatu
somba
somba
somba
sɔmba
hô̠mba
homba
ɲɟusi
ɲɟuki
luʝʊki
luyukhi
luɲɟuci
lijusi
injuchi
lisosolo masosolo
liʋembe maʋe
Electronic Working Paper 2016-001
New perspectives on the
genetic classification of
Manda (Bantu N.11)
Hazel Gray and Tim Roth
New perspectives on the genetic classification of
Manda (Bantu N.11)
Hazel Gray and Tim Roth
SIL International®
2016
SIL Electronic Working Paper
2016-001
2016 SIL International®
©
ISSN 1087-9250
Fair Use Policy
Documents published in the Electronic Working Paper series are intended for scholarly research and
educational use. You may make copies of these publications for research or instructional purposes
(under fair use guidelines) free of charge and without further permission. Republication or commercial
use of an Electronic Working Paper or the documents contained therein is expressly prohibited without
the written consent of the copyright holder(s).
Managing Editor
Eric Kindberg
Series Editor
Lana Martens
Content/Copy Editor
Mary Huttar
Compositor
Bonnie Waswick
Abstract
Manda (N.11) is an under-documented Bantu language in southwestern Tanzania, with most mention of
its closest genetic affiliation in the work of Nurse (1988). Nurse concludes that Manda belongs with the
Southern Tanzania Highlands (SH) subgroup based primarily on phonological evidence. This paper uses
new data from workshops and surveys conducted by SIL International to show that Nurse’s conclusion
does not take into account the necessary dialectal information, namely that the Matumba dialect (which
Nurse uses) is phonologically distinct from the remaining Manda dialects. Lexicostatistical, phonological,
and sociohistorical evidence is taken into account. Further, in light of this new evidence, we propose two
likely historical scenarios, both of which posit that Manda is most closely genetically affiliated with the
Rufiji subgroup.
Contents
1
2
3
Introduction and background
Dialectometry and lexicostatistical evidence
Phonological evidence
3.1 Dahl’s Law
3.2 Spirant weakening
3.3 *NC̥> NC or N
4 Sociohistorical evidence
5 Synthesis and conclusion
Appendix A: Lexicostatistics
Appendix B: 296-item wordlist for corpus languages
References
iv
1
Introduction and background
Manda [ISO 639-3 code: mgs] is a Bantu language (N.11) spoken by the Manda and Matumba language
communities located in the area between Lake Nyasa and the Livingstone Mountains (Maho 2009). The
language area straddles two administrative regions in Tanzania: Njombe Region, north of the Ruhuhu
River, and Ruvuma Region, south of the river. The area is bordered to the north by Kisi (G.67), 1 to the
east by Pangwa (G.64) and Ngoni (N.12), and to the south by Matengo (N.13) and Mpoto (N.14) (Maho
2009). Lewis et al. (2015) report the Manda (and Matumba) population at 22,000. A previous SIL survey
puts the estimate even higher, between 25,000 and 40,000 (Anderson et al. 2003a). This paper uses
survey data from four other locations in addition to the Matumba variety in Luilo: Iwela, Lituhi, Litumba
Kuhamba and Nsungu (see map below). We examine three streams of evidence (lexicostatistical,
phonological, and sociohistorical) in this study to determine the closest genetic relatives of Manda (and
Matumba). New data is put forward from survey work conducted by SIL personnel in 2013 and
subsequent additional linguistic research (Gray, forthcoming; Gray and Mitterhoffer 2016).
1
The Guthrie codes for Bantu are referential, reflecting geography and not genetic relationship (see Schadeberg
2003: 146).
1
2
3
The Matumba consider themselves a separate ethnic group from the Manda, but still regard their
language to be essentially the same as Manda. The Matumba themselves claim that they were once
Manda who moved from the shores of Lake Malawi up into the mountains (Anderson et al. 2003a). The
prestige dialect (even according to the Matumba) of Manda is spoken in the area near Ilela and Nsungu,
villages on the lakeshore (Gray and Mitterhoffer 2016). Various comparative linguistic studies have
included Manda wordlists (Guthrie 1971, Nurse and Philippson 1975) and mention of Manda is made in
diachronic studies by Nurse and Philippson (1980) and Nurse (1988, 1999). None of these studies
mention the Matumba, so it is quite likely that the Matumba were unintentionally grouped with the
Manda in previous studies. As we will see in section 3, this lack of previous dialect research has given
rise to some misconceptions about the innovations that Manda shares with its neighbors.
Of those who have studied the Manda language with the aim of positing its closest genetic
affiliation, Nurse and Philippson (1980) and Nurse (1988, 1999) are in conflict with Ehret (1999, 2009).
Ehret (2009:17) puts Manda alongside Ngoni in the Rufiji-Ruvuma (RR) subgroup, whereas Nurse
classifies Manda within his Southern (Tanzania) Highlands (SH) subgroup. These are the main
hypotheses we evaluate in light of the new data which reflects a better understanding of Manda
dialectology. RR consists of two sub-branches. We are concerned primarily with Rufiji, which includes
Manda’s immediate neighbors: Ngoni, Matengo, and Mpoto. The SH subgroup includes the G.60
languages, of which we are primarily concerned with Kisi and Pangwa for the same reason.
This paper mainly focuses on the arguments put forward by Nurse (1988), since he deals with the
question of Manda’s genetic affiliation in the most depth. Ehret primarily relies on stem-morpheme
innovations (lexical/semantic evidence) in his (1999) work, and does not offer the same depth of
interaction with the corpus most like an SH language that Nurse does. From the Manda data Nurse uses,
Manda appears to behave phonologically mostly like an SH language. However, we argue that the
Manda data Nurse uses appears to be the Matumba dialect, which phonologically is quite different from
the other Manda dialects. In section 3 we see how the features of Dahl’s Law, spirant weakening, and
NC̥>NC or N cast doubt on Nurse’s argument for Manda as SH. We offer three scenarios regarding the
genetic history of Manda and Matumba in section 5.
In section 2 we present initial lexicostatistical evidence using dialectometry. Manda and Matumba
varieties are compared to the other corpus languages: Kisi, Matengo, Mpoto, Ngoni, and Pangwa. Section
3 examines the phonological evidence for the corpus languages, while section 4 briefly examines the
sociohistorical evidence. Section 5 concludes the paper with a synthesis, preliminary conclusions, and
possibilities for future research.
2
Dialectometry and lexicostatistical evidence
Dialectometry is essentially quantitative dialectology. Distance-based networks are examples of such
quantitative data explorations. A distance-based network analysis offers ‘‘an introductory visual means of
data exploration’’ (Pelkey 2011:279). Subsumed under the rising field of dialectometry are several
distance-based algorithmic applications that aim to help researchers explore language variation and/or
change, while “making it possible to show more than one evolutionary pathway on a single graph”
(Holden and Gray 2006:24). To create a distance-based network the opposite values of regular
lexicostatistical percentages are used within a standard matrix (e.g., the Kisi and Pangwa languages in
figure 2 are 0.59 similar, but 0.41 dissimilar; see Appendix A). The resulting distance matrix is then
subjected to the Neighbor-Net algorithm, as developed by Bryant and Moulton (2004) and implemented
within the Splits Tree 4 (4.11.3) software program (see Huson and Bryant 2010). If the lexical
relationships are ambiguous, the length of each branch indicates the confidence level of the split. Figure
1 shows such a network using four Bantu lects from southwest Tanzania (Roth 2011:43).
4
Nyika
Nyiha
Safwa
Malila
Figure 1. Sample network demonstrating ambiguity (adapted from Roth 2011:43).
In figure 1, the middle box indicates the ambiguous relationship: Split A—Nyika and Nyiha/ Safwa
and Malila versus Split B—Nyika and Safwa/ Nyiha and Malila. In figure 1, Split A is more likely. Holden
and Gray discuss other patterns and their meaning within the network diagram:
Rapid radiation may be inferred from a lack of phylogenetic signal, i.e. a rake- or star-shaped
phylogeny, whereas reticulation would indicate possible borrowing. Reticulations can also
pinpoint those languages which may have been involved in borrowing. Complex chains of
conflicting relationships involving numerous languages may indicate that borrowing occurred in
the context of dialect chains. (2006:24)
If the lexical relationships are more clear, the splits-graph looks more like a regular tree diagram
(see Holden and Gray (2006) for further explanation of the Neighbor-Net algorithm, and the unique
historical relationships between other Bantu languages).
Figure 2 below is based on the lexicostatistical data in Appendix A. The data includes lexical
percentages from 313-item wordlists elicited in five locations (Iwela, Lituhi, Litumba Kuhamba, Luilo
and Nsungu) during the 2013 dialect survey (Gray and Mitterhoffer 2016). This 313-item wordlist was
based on the 100-item Leipzig-Jakarta wordlist (Tadmor 2009). Two hundred ninety-six of the lexical
items 2 in the 313-item wordlist were compared to data from SIL Fieldworks Language Explorer (FLEx)
2
Some lexical items were omitted due to duplications in the data.
5
databases for Kisi and Pangwa, and CBOLD data for Mpoto, Matengo and Ngoni (supplemented by data
from Ngonyani 2003 and Yoneda 2006) using the comparative method as the basis for determining
cognacy. The lexical data (the 296-item wordlist) can be found in Appendix B. Consider the distancebased network below in Figure 2.
Matengo
Pangwa
Kisi
RR
SH
Mpoto
Ngoni
Lituhi
Litumba
Kuhamba
Nsungu
Luilo
(Matumba)
Iwela
Manda
Figure 2. Distance-based network for SH, Rufiji and Manda corpus languages.
From the diagram in figure 2, we can clearly see the split between the branch containing the SH
languages (Pangwa and Kisi) on the left and two of the Rufiji languages (Matengo and Mpoto) on the
right. We can also see that the Manda dialects, while lexically related to SH and Rufiji, are at the same
time lexically distinct, i.e., the Manda dialects seem to have a split-lexicon. Ngoni, while lexically more
like Manda, is still split between Manda and the SH languages. (Ngoni is generally classified as RR, but
its situation is similar to Manda’s, with the added complication of having an unclear relationship with
the South African Nguni, see section 4).
In sum, the lexicostatistical evidence for Manda’s closest genetic affiliation is ambiguous as Manda
has a split-lexicon. However, we can evaluate Nurse’s claim that Manda is “lexicostatistically a Rufiji
6
language” (1988:70) and say that this more recent wordlist data does not corroborate that claim. Section
3 proceeds with evaluating the phonological evidence that ties Manda to either SH or Rufiji (or both).
3
Phonological evidence
As we discuss in section 1, Nurse (1988) provides the most detailed argument regarding the genetic
affiliation of Manda. Table 1 (adapted from Nurse 1988:47) 3 is a comparison of what Nurse considers the
most relevant phonological features between SH, Rufiji, and Manda, in combination with conclusions
from our newer data in italics.
Table 1. Comparison of Manda phonological features with SH and Rufiji (adapted from Nurse 1988:47)
Feature
SH
Manda
Rufiji
Spirant-devoicing
(Proto-SH) yes
yes
yes (Proto-R)
Class 5 /li-/
(Proto-SH) yes
yes
yes (Proto-R)
NC̥
(Proto-SH) N(h)
N and NC
NC (Proto-R)
Dahl’s Law
(Proto-SH) yes
yes (minor)
no
Spirant weakening
no
no (yes)
yes (Proto-R)
Loss of preprefix
no
yes
yes (Proto-R)
Nurse concludes that Manda is an SH language primarily based on the phonological evidence
represented in this table. Nurse’s argument is as follows: (1) that SH and Rufiji differ in at least four
categories (lines 3–6); (2) of those four categories, Manda matches SH in two cases, maybe three (line 5
is ambiguous); (3) thus, if Manda is SH, the only line which needs explanation is line 6 (the loss of the
preprefix, or augment); whereas if Manda is Rufiji, lines 3–5 all need explanation. Nurse summarizes as
follows:
It is therefore simplest to posit Manda, not as an original member of Rufiji, but rather as an
original member of SH, which in recent times has undergone lexical change, and loss of preprefix,
inherited vowel length (and possibly tones) under the influence of Rufiji, which presumably means
Ngoni or Matengo (1988:48).
In this paper, we seek as much as possible to distinguish genetic inheritance from contact/areal
diffusion, as well as shared innovations from shared retentions. For example, the class 5 *li- feature is a
retention (compare Nurse 1988:30 and Nurse 1999:23), and a feature such as the loss of the preprefix
(augment) is historically unclear and more likely to have been spread by contact (see Nurse 1988:47).
Both are excluded from consideration here on this basis. Furthermore, the spirant-devoicing feature is
shared by both subgroups (and most likely necessarily preceded spirant weakening), and so does not
factor into the discussion (Nurse 1988:40, 43–44). That leaves Dahl’s Law, spirant weakening, and
NC̥>NC or N. Each of them are innovations, but they each also have their own unique considerations
regarding the possibility of contact influence.
As we see in the remainder of this section, the Manda data Nurse uses appears to be the Matumba
dialect. Even though Manda and Matumba are lexically similar (see section 2), Matumba is
phonologically quite different from the other Manda lects. This section provides the bulk of the evidence
for Manda and Matumba as members of either SH or Rufiji. We examine Dahl’s Law, spirant weakening,
and NC̥>NC or N in sections 3.1, 3.2 and 3.3, respectively.
3
The table is duplicated with the exception of the feature “nasal and spirant”. This category is excluded from this
study for two main reasons: (1) it is not a proto-SH or proto-Rufiji feature, and (2) it is already sufficiently covered
under the “spirant weakening” proto-feature (Nurse 1988: 46–47).
7
3.1
Dahl’s Law
Dahl’s Law is a voicing dissimilation in many Bantu languages in which the first of two voiceless plosives
becomes voiced, either within the root or across morpheme boundaries. Table 2 shows the examples of
Dahl’s Law in several of the corpus languages. These are the only examples found in the Manda corpus.
Nurse reports that there is no Dahl’s Law in Rufiji except for traces in Ngoni and Mbunga (P.15)
(1988:103). The examples in italics are the lexical items showing Dahl’s Law. The BLR3 Proto-Bantu
form is included in the table as well for comparison (Bastin et al. 2003). The lexical items not in italics
are various stems where we might expect Dahl’s Law to operate but where no evidence of Dahl’s Law is
found, despite both Proto-Bantu stem consonants being voiceless plosives.
Table 2. Dahl’s Law examples in the corpus languages
BLR3
*kàtɪ ́
*kèt
*kɪ ́t
*kópè
*kʊ̀ʊ̀kʊ́
*pàk
*pít
*tákò
*tátʊ̀
*tétè
Manda
pagati
kuketa
kukɪta
lukopi
kʊkʊ
kubaka
kubɪta
lidakʊ
-datu
lidete
Kisi
n/a
kukeːta
kukɪta
lukopi
kʊːkʊ
kubakala
kubɪta
lidaːku
-datu
lidete
Pangwa
pakatinakati
xuxeːta
xuxita
n/a
xuxu
xubaxa
xubita
n/a
xidatu
lidete
Ngoni
paɡatipaɡati
kuketa
kukita
(nɡ’opi)
n/a
kubakala
n/a
lidaku
-datu
lidete
English
middle
to shave
to do
eyelash
grandparent
to smear
to go
buttock
three
reed
The Manda dialects (including Matumba) share the same expression of Dahl’s Law: traces in the vast
majority of the same lexical items and not across morpheme boundaries. Of the SH languages, only
Hehe, Bena and Kinga have Dahl’s Law in all stems. Kisi and Pangwa show traces of it as does
Manda/Matumba, while Sangu and Vwanji show no evidence of it at all. Based on this geographic
distribution, it appears as though those languages with little or no trace of Dahl’s Law were on the outer
edge of an earlier dialect continuum where Dahl’s Law was present. Regardless, Kisi, Pangwa,
Manda/Matumba and Ngoni all share the same traces of Dahl’s Law.
There are two main possibilities that might explain the pattern of Dahl’s Law traces in
Manda/Matumba: (1) genetic relationship with SH with (a) inconsistent application and/or (b) early
phonological reversal (Nurse 1999:28), or (2) contact. Bantu languages can have inconsistent application
of Dahl’s Law due to when the feature became inactive (Batibo 2000). A phonological feature becoming
inactive can also be due to contact influences, as Batibo relates for the Sukuma/Nyamwezi languages
(F.20) in western Tanzania:
The major reason for this inactivity, and therefore incompleteness, may have been the instability
that both Sukuma and Nyamwezi experienced in their early years of resettlement due to the influx
of intruding groups…The influx of the intruding groups meant the influx of new lexical items by
speakers who did not have these rules in their language (2000:25).
Dahl’s Law is generally considered to be diagnostic of phylogenetic relationship. However, Nurse
says that “I would not claim that these features [Dahl’s Law, Bantu spirantization, *p-lenition] never
cross language boundaries but rather that they are more likely to be inherited in our languages from an
early stage of common development and thus historically diagnostic for the early period” (1999:20).
Thus, it is possible to conclude that traces of Dahl’s Law are due to borrowing. Considered by itself, what
the presence of Dahl’s Law traces means for the genetic affiliation (SH or RR) of Manda/Matumba is
inconclusive.
8
3.2
Spirant weakening
Spirant weakening, or lenition, is a process where spirants, such as /s/ and /f/, become weakened to /h/
or are lost entirely. Manda is again characterized by ambiguous data: a large number of lexical items still
contain /s/ and /f/, though some have weakened to /h/. Table 3 shows all the relevant examples from
the corpus of spirant versus /h/ forms in Matengo, Mpoto and Ngoni (RR), Kisi and Pangwa (SH), as well
as Manda.
Table 3. Spirant vs /h/ in corpus languages
English
gloss
fingernail
tear
egg
milk
name
leaf
grass
knot
cloud
dust
fish
ashes
to die
to choose
to get lost
to remain
to work
to pull
Rufiji-Ruvuma
Matengo
Mpoto
sô̠bu/hyô̠bu
chobu
lîːholi
maholi
lihóm
̠ bi
lihombi
másiba
lihiba
lihina
lihina
lihamba
lihamba
linyahi
linyai
lihundu
lihundwa
lihundi
maundi
luhô̠mbi
kuhombi
homba
homba
ilîhu
iliu
-ha
-wa
-hagula
-hagula
-hoba
-hoba
-higala
-higalila
-henga
-henga
-huta
-huta
Ngoni
luhyuwu
maholi
n/c
n/c
lihina
lihamba
linyahi
lihundu
lihundi
luhumbi
somba
n/c
-fwa
-hagula
-hova
-higala
-henga
n/c
Manda
luhʲuwu/lufʲuwu
liholi/ lisoli
lihombi
lisiva
lihina
lihamba
liɲahi / liɲasi
lihundu/lifundu
lihundi / lifundi
luhumbi/lufumbi
somba
lifu
-hwa/ -fa/ -fwa
-hagula
-hova /-sova
-higasa / -sigasa
-henga
-huta / -futa
Southern Highlands
Kisi
Pangwa
n/c
n/c
n/c
n/c
lifʊmbi
n/c
n/c
n/c
n/c
n/c
n/c
n/c
linyasi
n/c
(liputu)
lifundo
n/c
n/c
(lifu)
mafumbi
somba
somba
n/c
n/c
-fwa
-fwa
-haghula
n/c
n/c
n/c
-sighala
-sikala
n/c
n/d
n/c
n/c
Of the corpus languages, Matengo, Mpoto and Ngoni (all RR) generally use the weakened /h/ forms,
while Kisi, Pangwa (SH) and the Manda lects of Iwela, Nsungu, and Luilo (Matumba) generally preserve
/s, f/. The Lituhi and Litumba Kuhamba lects are in between, both geographically and in the type of
forms they use. The Ruhuhu River separates these Manda dialects from the others.
The modern-day distribution of spirant weakening among the corpus languages fits that of a wave
(see Mühlhäusler (1996:10) for a summary of Bailey’s dynamic wave model in particular). Thus, the
variation within Manda (and to some extent its neighbors) is not entirely unexpected. The distribution of
spirant weakening could be explained by one or both of two factors: (1) inconsistent application of the
phonological change due to the varied timing of its inactivity, and (2) historical contact. As we saw in
section 3.1, contact influences can cause a phonological innovation to become inactive, which is how
both inconsistent application and historical contact might have come into play together. If (1) were true,
Proto-RR could not have already undergone spirant weakening, as inconsistent application across some
lects would presuppose its daughter languages had already started to expand and diversify. Apart from
any other evidence, it would also mean that in all likelihood Manda’s closest genetic affiliation would be
the Rufiji subgroup due to its shared spirant weakening innovation with Matengo, Mpoto, and other
Rufiji. However, there is no way to eliminate the possibility of contact/areal diffusion. In regard to (2),
because of the variation in his own Manda data regarding spirant weakening, Nurse (1988:47) says that
it “could be explained in either direction by simple borrowing of lexis carrying either spirant or /h/”. In
9
sum, spirant weakening in the corpus languages remains inconclusive in terms of helping to determine
the closest genetic affiliation for Manda.
3.3
*NC̥> NC or N
The phonological process whereby the historical sequence of a nasal followed by a voiceless homorganic
plosive (*NC̥) has become either a nasal followed by a voiced plosive (NC) or a nasal by itself (N, in
which the consonant has deleted altogether) is represented here as *NC̥> NC or N. Once again, the
problem is that we need to try to distinguish inheritance/innovation from contact. Nurse explains the
historical process as follows:
We assume here that inherited sequences of nasal plus voiceless homorganic stop have undergone
one of two major processes: either the stop is voiced…or the sequence is maintained, which,
however, can lead to devoicing of the nasal, which in turn can lead to N, Nʰ, C̥ʰ or C̥. Only the
proto-languages for [RR] and Nyakyusa/Ndali show the stop voicing, whereas all the others show
one or other form of the second process. (Nurse, 1988:31)
These processes constitute two very different pathways, and the issue is that we generally see
*NC̥>NC in Manda and *NC̥>N in Matumba. Kisi and Pangwa have N like Matumba, while Matengo
and Mpoto have NC like the rest of Manda.
This is also an area where Nurse (1988) seems to use the Matumba dialect as normative for the
entire language. In table 4, this is represented with the CBOLD label, which is the data that Nurse and
Philippson collected. Table 4 shows all of the relevant reflexes of *NC̥ across Manda and Matumba from
the corpus, and it also includes the CBOLD data and BLR3 references.
Table 4. Class 9/10 across Manda/Matumba dialects
PB stem
(BLR3)
English
gloss
Lituhi
Litumba
Kuhamba
Nsungu
Iwela
CBOLD
Manda
ŋguku
mbepo
Luilo
(Matumba)
ŋuku
mepu
*-kʊ́kʊ́
*-pépò
‘chicken’
‘cold’
ŋguku
mbepʊ
*-kópè
*-kómb?
‘eyelash’
‘finger’
ŋgoɲɟi
n/c
n/c
*-ntù
*-ntù
‘goat’
‘lightning’
‘person’
‘thing’
mene
mbamba
mundu
kindu
*-jónk-
‘suck’
-joŋga
ŋguku
mepʊ/
mbepʊ
ŋoɲɟi/
ŋgoɲɟi
mene
mamba
mundu
kindu /
kinu
-ɲoŋga
ŋguku
mbepu
ng’oko
mepu
ŋgopi
(fiŋgoɲɟi)
ŋopi
ŋgoɲɟi
ŋgopi
ŋgoɲɟi
(lukopi)
(lukonji)
mene
mbamba
mundu
kindu
mene
mamba
munu
kinu
mene
mbamba
mundu
kinu
mene
mamba
munu
kɪnu
-joŋga
-ɲoŋga
/ -ɲoŋa
-joŋga
-ɲoŋa
We can clearly see from table 4 that the CBOLD data patterns most closely with the Luilo
(Matumba) dialect. To a certain extent there are examples of both N and NC in each dialect (often in the
form of lexical doublets), but the overall pattern remains.
Some of the examples from table 4 could possibly be looked at as the after-effects of either Dahl’s
Law or Meinhof’s Rule/Law (Ganda Rule/Law). Since positing Dahl’s Law in class 9/10 roots is not
supported by evidence from any other noun or verb roots, it is very unlikely that Dahl’s Law is what is at
play here. Meinhof’s Rule would be more likely as it resembles the process *NC>N or NC. Meinhof’s
Rule is a process in which a prenasalized voiced consonant elides when followed by a nasal or
10
prenasalized consonant in the subsequent syllable, leaving behind either a simple or geminate nasal.
Gray (forthcoming) describes in her Manda phonology sketch that it is not Ganda Law at work, but
rather the post-nasal consonant is deleted across the board consistently for classes 1 and 3. There are
many words where there is no NCVNC sequence, and yet C1 is deleted (for class 1 and 3), such as *ndala/nala, and *n-gosi/ŋosi, etc. There are also lots of NCVN(C) sequences that do not become NVNC:
ndemba (hen), ngondo (quarrel/war), mbanda (slave), ndongo (relative), ndomondo (hippo).
The *NC̥> NC or N evidence in this section is not the silver bullet of clarity to Manda’s genetic
relationship to either SH or RR. However, this new research into the *NC̥ process does pose a serious
difficulty in any argument for Manda as a SH language. Possible rejoinders to the argument that Manda
is not SH on the basis of *NC̥ reflexes are that (1) perhaps Manda and Matumba should not be
considered a single historical language variety, in which case Matumba could certainly be SH, or (2)
perhaps Manda/Matumba is SH and all the Manda dialects except for Matumba were in contact with RR.
We explore these possibilities in our synthesis and conclusion in section 5, after we examine the
sociohistorical evidence in section 4.
4
Sociohistorical evidence
While oral traditions do not necessarily prove historical origin, it is worth taking into account what the
Manda and Matumba say about their own historical origins. Data gleaned from group questionnaires
taken during the 2013 SIL dialect survey of the Manda/Matumba area reveal that the Manda and
Matumba have mixed views of their origins. Several groups claim the Manda are partly descended from
the Nguni tribe of South Africa; others claim that they were from Malawi or from the Songea region. 4
Some claim they are partly descended from the Pangwa and partly from the Matengo language
communities. We will briefly comment on these claims.
Ngonyani (2003:1), in his explanation of the origins of the Tanzanian Ngoni people, states that the
Ngoni people incorporated many indigenous inhabitants of the area from different language groups when
they moved into the highlands east of Lake Nyasa. Ngonyani includes Manda in the list of language
communities that the Ngoni incorporated. This statement implies there was at least a group called
Manda living there at the time the Ngoni invaded, which seems to run counter to claims that the Manda
are descended from the Ngoni. Linguistically, Nurse (1999:13) claims that the Ngoni themselves
abandoned the Nguni language in favour of the local languages in the nineteenth century when they
invaded from the south. Nurse (1988) also made this claim based on the absence of connection between
Tanzanian Ngoni and South African Nguni languages. It would seem that the invasion left few linguistic
traces from Nguni (Nurse 1988:48), and that the language of the invaders themselves, now known as
Tanzanian Ngoni, was more affected by the invasion than the neighboring groups were. However,
according to the original SIL sociolinguistic survey (Anderson et al. 2003a), the Manda/Matumba
themselves claim to understand Ngoni very well. This holds true even for those villages furthest from the
Ngoni language area, with some saying that even children can understand Ngoni.
In regard to the claims about the historical relationship to Pangwa, Anderson et al. (2003a:9) state
that “historically, people of various ethnic groups (mostly Pangwa) migrated from various areas into the
region along the coast (of the lake)”. According to this history, the Manda would be most related to the
Pangwa; however, the people’s own perception of the relationship between the languages contradicts
these origins. The Pangwa survey report (Anderson et al. 2003b) confirms that currently at least, the
Manda and Pangwa people groups have little contact, and that the Pangwa understand very little of the
Manda language. The Manda/Matumba themselves claim that there is “no language relationship
whatsoever between Manda and Pangwa” (Anderson et al. 2003a:8), even for those villages closest to the
Pangwa language area. The intelligibility of Kisi seems more disputed; some villages claimed that there
was little resemblance, but some (Matumba villages) claimed to understand Kisi relatively well. All
groups that were interviewed during the survey claimed that the lexicon is similar and the difference lies
4
This may also refer to Nguni origins since that is where the Tanzanian Ngoni settled.
11
in pronunciation. Interestingly, the Manda village called Iwela, which has had more contact with Kisi
than most of the rest of the Manda area, is considered by some other Manda (Matumba) villages as a
place where the people are not ethnically Manda and do not speak the Manda language. The people of
Iwela themselves claim to be Manda, but many features of their dialect show the influence of the Kisi
language on it. Nurse (1988:70) claims that Kisi has been heavily affected by the N.10 languages, which
may be a factor influencing the higher intelligibility with Kisi than with Pangwa.
Regarding the possible connection with Matengo, the findings in the 2013 dialect survey were that
for those villages south of the Ruhuhu River in the Ruvuma region there is considered to be little
intelligibility between Mpoto and Manda and even less between Matengo and Manda.
In summary, the sociohistory of the Manda language community is still unclear. General feeling
among the Manda/Matumba would connect them most closely to the Ngoni. However, Ngoni appears to
have had merely a superstratum influence on Manda and other language communities in the area, and so
this would not be an indicator of close genetic affiliation. For Pangwa, one would expect the
comprehension with Manda to be higher if the Manda had indeed descended from Pangwa as reported;
however, comprehension is quite low as the Livingstone Mountains create a barrier between the Manda
and the Pangwa language areas. As an added note, the accessibility of the Kisi area by canoe gives the
Manda/Matumba more opportunity for contact with Kisi.
We now turn to synthesizing the lexical, phonological, and sociohistorical evidence, and developing
some conclusions, specifically in interaction with the results regarding Manda’s historical relationships
from Nurse (1988). We will also briefly explore ideas for further research.
5
Synthesis and conclusion
So far, this paper has considered three main streams of evidence (lexicostatistical, phonological, and
sociohistorical) in working towards determining whether Manda is most closely genetically affiliated
with SH or RR. In isolation, all of the streams have been ambiguous and so far inconclusive. In this
section, we seek to synthesize these streams of evidence and arrive at some tentative conclusions.
In section 3.3 we encountered evidence that led to the possibility that we should consider Manda
and Matumba historically separate languages. The *NC̥> NC or N feature was found to be the significant
phonological difference between the Matumba/Luilo dialect (*NC̥>N) and the remaining Manda dialects
(*NC̥>NC). In this view, Manda could be historically RR while Matumba could be SH, solving the most
pressing difficulty. There are at least two arguments for the historical unity of Manda and Matumba as
dialects of one language: (1) the lexicostatistics bear this out (see section 2 and Appendix A), and (2) the
testimony of Manda and Matumba native speakers themselves.
Frankly, however, this is not enough evidence to eliminate the option that Manda and Matumba
were historically separate languages that originally came from different genetic subgroups. If this were
indeed the case, it would not change the fact that Matumba can certainly be considered a modern-day
dialect of Manda. In regard to (1), as important as lexicostatistics can be, they “can only describe and
extend relatedness but cannot establish it” (Nichols 1996:64). The same can be said of native speaker
testimony (see Hinnebusch 1999:179). The gold standard of evidence for genetic relatedness is shared
innovations, specifically what Nichols (1996) calls ‘individual-identifying’ evidence. The only such
evidence that could be considered ‘individual-identifying’ in this paper is the *NC̥ feature, which
happens to cut across Manda and Matumba. Part of the reason for the lack of shared innovations could
be rooted in the history of these Bantu subgroups. Nurse says of SH that there is “a relative lack of really
distinctive innovations. None of the innovations preceding […] is unique, all being shared with some
combination of surrounding groups, which raises the possibility that the innovations might be the result
of areal spread, or that the Proto-SH period was short, not allowing time for innovation” (1988:40).
Rufiji is in a similar situation (see Nurse 1999:31).
Normally with such a lack of shared innovations, we could rely even more on paradigmatic
grammatical evidence (see Nichols 1996), especially for establishing Matumba historically as a dialect of
12
Manda. But again, the evidence remains ambiguous and inconclusive. We can see this in the clear
differences between the personal pronoun set(s) in Manda and Matumba in table 5.
Table 5. Personal pronouns in Manda and Matumba (Gray 2016:148)
Person
1SG
2SG
3SG
1PL
2PL
3PL
Manda (Nsungu)
nenga
wenga
jʊla
tenga
mwenga
vala
Matumba
nene
veve
mwene
tete
nyenye/mwemwe
vene
Obviously, this is just one example of a paradigmatic grammatical set in which Manda and
Matumba are in disagreement. More research is needed on the grammar of the Matumba dialect in
particular. What we do know is that Manda and Matumba indeed share much in common phonologically
and grammatically. That Matumba is a modern-day dialect of Manda is not in dispute. The issue is that
we still cannot dismiss the possibility that historically Manda and Matumba did not come from the same
recent ancestor—that Manda belonged to the Proto-Rufiji subgroup, while Matumba was a member of
Proto-SH. This would entail that Matumba underwent massive relexification, to the point where today
the lexicostatistics are indistinguishable. Matumba must have also adopted many grammatical elements
from Manda. Under a long period of extreme contact, none of this is out of the question in a
geographical area where SH and Rufiji collide. What of the other possibilities?
If we set aside the option that Manda and Matumba come from different subgroups historically, we
are left with Manda/Matumba as either SH or Rufiji. We saw in section 2 that Manda/Matumba
essentially has a split-lexicon between SH and Rufiji, which does not support one subgroup over another.
The argument for Manda/Matumba as SH centers around the traces of Dahl’s Law. Under this scenario,
spirant weakening is adequately explained due to contact with Rufiji languages to the south. The *NC̥
feature is harder to explain, but not impossible. Its distribution too would have to be the result of
contact: *NC̥>N in Matumba represents the original SH feature, while *NC̥>NC in the rest of Manda
would be due to contact with Rufiji. The question under this scenario: Why didn’t Matumba adopt NC like
the rest of Manda? It would have to have been in contact with Rufiji languages (or the other Manda
dialects with NC) to explain the spirant weakening pattern.
The argument for Manda/Matumba as Rufiji centers around the majority of Manda dialects showing
*NC̥>NC and the spirant weakening pattern representing inconsistent application of the innovation.
Dahl’s Law traces are explained by the borrowing of individual lexical items from SH. Given the small
amount of the Manda lexicon that has been affected by Dahl’s Law and those traces being in common
with the neighboring SH languages, contact influence is reasonable despite Dahl’s Law normally being
diagnostic in other branches of Bantu. *NC̥>N in Matumba is due to contact with SH. In this scenario,
this borrowing (or reversal) only happens in one dialect instead of several. Dahl’s Law lexical items
would have diffused much earlier. Nurse also mentions for Rufiji, “an apparently unique set of
allomorphs for the /-ile/ suffix” (1988:45). Manda appears to have these allomorphs (Gray 2016:108),
but although their geographical distribution has been clarified (e.g. Nurse 2008:267), it is still unclear
whether these /-ile/ allomorphs truly would distinguish Rufiji from SH, but needs to be explored further.
Of these three different options, it is difficult to choose between the first scenario of Manda and
Matumba as historically different languages and the third scenario of a united Manda/Matumba as
Rufiji. Both are not just possible but plausible, even though they have their drawbacks and relatively
unusual historical journeys. The second scenario with a united Manda/Matumba as SH seems much more
unlikely. Further research is needed on morphosyntactic innovations (e.g. tense/aspect) and tonal data
for the corpus languages, especially detailing any further differences in the Matumba lect.
13
In this paper we have explored lexicostatistical, phonological, and sociohistorical evidence in the
goal of determining the closest genetic affiliation for Manda (SH or RR). Much of the evidence was
ambiguous and inconclusive, but in this section we were able to put together three viable scenarios. Two
of those scenarios we found much more likely than the remaining option, and both refine our
understanding of Manda’s history and dialectology. We primarily interacted with Nurse (1988) who had
given the most in-depth previous account. Most crucially, it appears that the dialect used by Nurse for
Manda is actually Matumba, which is phonologically different from the Manda dialects, especially in
regard to the reflexes of *NC̥. Regardless which one of the likely options correspond to historical reality,
both posit the majority of Manda dialects as historically Rufiji, not SH.
Appendix A: Lexicostatistics
Lituhi
75
Litumba
Kuhamba
77
67
Nsungu
73
76
71
Luilo
75
73
77
78
Iwela
53
54
53
58
58
Kisi
43
44
45
50
48
59
Pangwa
65
65
60
67
63
52
45
Ngoni
47
44
48
43
48
37
26
43
Matengo
53
45
51
42
48
33
26
43
71
14
Mpoto
Appendix B: 296-item wordlist for corpus languages
The wordlist data transcriptions are replicated from the original source databases (see §2), except for the Manda and Matumba varieties which
are in IPA. Tone is generally not included due both to the nature of the source databases and the rapid survey word-collection conducted by SIL.
1
English
Gloss
eye
2
eyelid
3
ear
4
mouth
5
*jaw
6
nose
7
*chin
8
beard
9
tooth
10
Lituhi
líhu - míhu
Litumba
Kuhamba
liho - mihu
Nsungu
Luilo
Iwela
Kisi
Pangwa
Ngoni
Matengo
Mpoto
lihu - mihu
lihu - mihu
lihu - mihu
liːhu
liho
lîhu - mîhu
liu
ŋgopi
ŋopi
ŋgopi
lukopi
ng'opi
ingopi
ingopi
likutu makutu
ndomo milomo
likutu makutu
ndomo milomo
likutu makutu
ndomo milomo
ng'ope cha
mumiho
lihu
mbʊlʊkʊtʊ
mbulukhutu
likutu; ɲɟɛvɛ
ndomo
mlomo
m̩ lɔmɔ
fíkupalɪlɪ kíkʊpalɪlɪ
líkutu mákutu
ndomo milomo
fikupulila kikupulila
líkutu makutu
ndomo milomo
ɲɟeɟe
ɲɟeɟe
ɲɟege
tili
taʝa
lucheeche
njeje
mbúnu mbunu
kiɲɟwemba fiɲɟwemba
mbunu /
meŋelu
litama matama
mbunu
mbunu
mbunu
mheŋelu
meng'elo
mɛŋɛlu
khilefu
ciɲɟwɛmba
kíleu - íleu
kileu - ileu
lúnde(b)u índe(b)u
indeo
lîno - mîno
lino
lúlimi - ínimi
lulimi
kiɲɟwemba - kiɲɟwemba - kiɲɟwemba - kiɲɟwemba kidefu
fiɲɟwemba
fiɲɟwemba fiɲɟwemba fiɲɟwemba
mlefu
maɲɟwɛmba
linu - minu
liɲɟwemba maɲɟwemba ndefu
maɲɟwemba
ndefu
línu - mínu
liɲɟwemba maɲɟwemba
linʊ - minʊ
linu - minu
linu - minu
liːnu
lino
linu
tongue
lulimi
lulími
lulimi
lulimi
lulimi
lulimi
lulimi
lulimi
11
head
mútu - mítu
mútu - mítu
mutu - mitu mʊtu - mitu mutu - mitu mutu
mutwe
mutu
12
hair (of
head)
lijúɲɟu majúɲɟu
lijuɲɟu majuɲɟu
ljuɲɟu majuɲɟu
ɲɟwili
njwili
ɲɟwili
liléme maléme
ŋkúfu mikúfu
liléme maléme
nkufu mikufu
lileme maleme
ŋkufu mikufu
siŋgu
singo
siŋgu
13
neck
14
stomach
15
navel
liɲɟwémba maɟwémba
siŋgu
siŋgʊ
siŋgʊ
lijuɲɟu
siŋgʊ
lileme
ŋkufu
lijuɲɟu majuɲɟu
siŋgʊ
lileme maleme
ŋkufu mikufu
15
lileme maleme
lileme maleme
nkʊfʊ
mdungu
lileme maleme
lukufu
likûtu makûtu
ndomo mílomo (lip)
lugômu –
ingômu
ímbulu ímbulu
umûtu mimûtu
likutu
kukano
lugomo
imbulu
mmutu
líjunzu májunzu
lijunju
hîngu-hîngu
hingo
lutumbo
lutumbu
nkuhu míkuhu
nkuu
16
16
English
Gloss
back (of
body)
Lituhi
ŋoŋgo migoŋgo
17
*knee
lifugumíɾu mafugumíɾu
18
leg
kigʊlʊ magʊlʊ
19
foot
20
arm
21
22
palm (of
hand)
*finger
nail
Litumba
Kuhamba
ŋoŋgo migoŋgo
lifugumilu –
mafugumilu /
ligoti
kigʊlʊ magʊlʊ
lwaju ɲaju
kiwóko mawóko
kiwóko mawóko
luhjuwʊ hjowʊ
lúkoɲɟi ŋgóɲɟi
lwaju ɲaju
kiwóko mawóko
lukoɲɟi ŋgoɲɟi
Nsungu
Luilo
Iwela
Kisi
Pangwa
Ngoni
Matengo
Mpoto
ŋoŋgo
ŋoŋgo
ŋoŋgo migoŋgo
ɲuma
mkongo
mgɔŋgɔ; msana
ngongu mígongu
nngongo
lifugumɪlʊ - lifugumɪlʊ - lifugumɪlu goti
mafugumɪlʊ mafugumɪlʊ mafugumɪlu
lifukamilo
lifugamilu
ligóti - mágǒti lijugwa
kigʊlʊ figʊlʊ
kigʊlʊ
kigʊlʊ figʊlʊ
kigʊlu figʊlu
kilʊndi /
kiɣʊlʊ
likulu
cigendelu
luɲajo
kiɣʊlʊ /
kilʊːndi
likulu
cigɛndɛlu
kiwoko mawoko
kiβoko
livokho
ciwɔkɔ
kigaɲɟa
kiɣaɲɟa
khikanja
cigaɲɟa
lukoɲɟi
lukhonji
luhyuwu
kikoɲɟi kja
kiβoko
khivokho
lukɔɲɟi
ngwapanilu
kiwoko mawoko
kigaɲɟa figaɲɟa
lwaju ɲaju
kiwoko mawoko
kigaɲɟa figaɲɟa
fjuhʊ
fjʊgʊ
ŋgoɲɟi
lukoɲɟi ŋoɲɟi
kikoɲɟi fiŋgoɲɟi
lukoɲɟi fikoɲɟi
lukoɲɟi ŋgoɲɟi
ŋgoɲɟi fikoɲɟi
ŋgwápa
ŋgwapa
ŋgwapa
ŋgwapa
ŋgwapa
ŋ'hwapa
mkhwapa
likúmba mákumba
lihúpa mahúpa
likúmba makúmba
lifupa mafupa
likumba makumba
liɟege maɟege
likumba makumba
lifupa mafupa
ŋgosi ja
mundu
sonda ya munu cikumba
lifupa
licheke
liʄɛgɛ
mwasi
mwasi
mwasi
danda
ŋasi
mwâhi
maholi
mahosi
luhosɪ
linyochi
mahɔli
lî:holi - máholi maholi
lifuki mafuki
kibindi fibindi
kuhʊma
lifuki
-huma ijashu
mfutumalu
lihógatela máhogatela
-
mtima
cibindi
lî:mani - ímani imani
lilɔvi
lîlobi - málobi
liŋɛlu
ling’élo
23
finger
24
armpit
25
skin
26
bone
27
blood
mwasi
28
tear
maholi
29
sweat
30
liver
31
voice
32
horn
lihuki mahúki
kibíndi fibindi
liŋelu maŋelu
lwaju
likumba makumba
lifupa mafupa
mwasi mwasi - mjasi
mjasi
lihõli lisoli maholi
masoli
lihũki lifuki
mahúki
liini ntima maini
mitima
ɲɟwaŋgu /
ɲɟwaʋila
liŋélʊ maŋélʊ
lifuki
kibindi fibindi
ɲɟoʋelu
liŋelu maŋelu
liːsu
liŋelʊ maŋelʊ
liŋelu maŋelu
lupembe
lipembe /
malamala
lwâgi - hyâgi lwâju - hyâju
kúboku máboku
kíganza íganza
ligolo
sôbu - hyôbu
lúgonzi ngônzi
ngwâpa ngwâpa
híkanda ngânda
lî:hupa máhupa
ligolo
lib'oko mabhoko
kig'anja
chobu
lukonji
ingwapa
likambila
kihupa
mwai
lihogutela
liɲelu
17
English
Gloss
Lituhi
Litumba
Kuhamba
lipapanílʊ mapapanílu
33
wing
lipapanílu mapapanílu
34
feather
liŋgóma maŋgóma
35
tail
ŋkɪ ́la - mikɪ ́la
36
egg
lihʊ́mbi mahʊ́mbi
37
honey
38
milk
lukáma
lukáma
lisiʋa
39
meat
ɲama
ɲama
40
person
múndu ʋándu
mundu ʋandu
41
man
ŋósi - ʋagósi
ŋosi - ʋagosi ŋosi - ʋagosi ŋosi - ʋagosi ŋosi - ʋagosi munagoːsi
42
husband
ŋgwana
waŋgu
43
woman
nala ʋadala
44
wife
muhana
waŋgu
45
father
(my)
dádi ʋadádi
dadi ʋadadi
46
mother
(my)
mawu ʋamawu
47
elder
brother
kaka
mawu ʋamawu
ndʊmbʊkaja ʋalʊmbʊkaja
/ kaka ʋakaka
liŋgóma maŋgóma
Nsungu
Luilo
Iwela
Kisi
Pangwa
Ngoni
Matengo
lipapatilo
ligwaba;
lipapanilu
kípapatila
liʝoʝa
likala
liŋgɔma
lingôma
mangoma
nkɪla
mkhila
mkila
ń̩ke̠la
nkela
lihʊmbi
lifʊmbi
likhaang'a
likaŋa
lihó̠mbi
lihombi
wuki
βwʊki
vuukhi
wuci
búsi
busi
lisiʋa
lisiʋa
lukama
machiva
lukama
másiba
lihiba lihibha
njama
njama
ɲama
ɲama
nyama
ɲama
ínyama
inyana
mundu ʋandu
munu - ʋanu
mundu ʋandu
mundu
munu
mundu
mûndu - bându mundu
mwadaada
m̩ gɔsi
mwánalomi ákanálomi
lipapanílu - lipapanilu - lipapanilu kibabatilu
mapapanílu mapapanilu mapapanilu
liŋgoma maŋgoma
ŋkila ŋkɪla - mikɪla
mikila
lihʊ́mbi lihʊmbi mahʊ́mbi
mahʊmbi
liŋgoma maŋgoma
ŋkɪla mɪkɪla
lihʊmbi mahʊmbi
wutʃi
ŋgwana
ʋaŋgu ʋaŋgwana
ʋaŋgu
nala ʋadala
ŋgwana
waŋgu ʋaŋgwana
ʋaŋgu
nala ʋadala
muhana
ŋʊli ʋaŋgu - waŋgu ʋagʊli ʋaŋgu ʋahana
ʋaŋgu
liŋgoma maŋgoma
ŋkila mikila
Mpoto
mwanalomi
ŋgwana
ʋaŋgu
ŋgwana
waŋgu
ŋgoːsi
mkoosi
mŋgwana
ńdomi - álomi,
ndomi
akaálomi
nala ʋadala
nala ʋadala
munadala
mwayuva /
mkhidala
mdala
mbômba - áka
mbomba
mbômba
ŋoli ʋaŋgu
ŋoli waŋgu
ndala
mdala
mdala
ńhanu, áhanu uhwano
ákaáhanu
dadi ʋadadi
dadi
dadi
daːdi
daadi
dadi
atâti - ákatâti áhěngolo atati b'ango
ákahéngolo
mau ʋamau
mau
mawu
maːβu
yuuva
mawu
amâbu ákamâbu
muhatʃa ʋahatʃa
kaka
kaka
mama
mhaacha
kaka; mkulu
mbeli - ábeli nkoano
ákaábeli
amabo
18
English
Gloss
Lituhi
Litumba
Kuhamba
Nsungu
Luilo
Iwela
Kisi
Pangwa
48
mother’s
brother
mɟomba
mwipa ʋipa
mwipa ʋipa
jaja
jaja
ʝaːʝa
muyaaya
49
child (my) mwána - ʋána
mwana ʋana
nseja ʋaseja
mwana
mwâna - bâna mwana
nséja ʋaséja
mwana ʋana
nseja ʋaseja
umwana
elder
mwana ʋana
ŋgogolo ʋagogolo
mwana
50
mwana ʋana
nseja ʋaseja
ŋgogolo
sehe
mkokolo /
mkoyo
vaɲaluhala;
gɔgɔlɔ
nsêja - asêja
51
grand
parent
hukulu
ŋgeɲɟi
mkenji
mgɛni; mhɛɲɟa;
ńgeni - ágeni
m'yehi
53
chief
mbaha witu mfalme
mutwa
n'kosi
54
god
tʃapáŋga
Muŋgu
muŋgu
muluŋgu
muŋgu
linguluvi
cimluŋgu
55
sickness
ilwala
ntamu
tamu
wound
utamu ʋatamu
kilonda filonda
tamu
56
utámu mátamu
kilónda filónda
sʊkʊlʊ ʋasukulu
wakuhitʃa ʋakuhitʃa
mutwa ʋatwa
muluŋgu /
kjuta
utamu matamu
kilonda filonda
khukhu
guest
sʊkʊlʊ ʋasʊkʊlʊ
ŋgene ʋajene
ɟumbi ʋaɟumbi
kʊːkʊ
52
húkuɾu ʋahukulu
ŋgéne ʋajene
mútwa ʋátwa
kilonda filonda
kilonda
kilonda
khivamba /
khikhoong'a
cilɔnda
57
medicine
mitiʃamba
ŋoda
ŋgoda
mkoda
mtɛla / migɔda ńtěla - mítěla
58
body
mbɪlɪ
mbili
mβɪlɪ
mvili
mhwili; mvili
hyêga - hyêga yega
59
name
lihina mahina
lihina mahina
litaːβu
litaawa
lihina
lihína - máhĭna lihina
60
lihowo mahowo
lihowo mahowo
lififi
lififi
lihɔgɔ
lipúndwa
lipundwa
lipusi mapusi
lisuʋi masuʋi
lihimba mahimba
lipusi mapusi
suʋi masuʋi
lihimba mahimba
liɲalu
limisi
kamlamu /
kapusi
pûsi mbunyâli
pusi mbunyau
sʊβi
liduuma
cihuvi
kího̠bi
kihobi
lisɪmba
libonji
lihimba
líhimba
lihimba
nténde miténde
mbɪlɪ miʋɪlɪ
lihína mahina
ŋoda migoda
mbɪlɪ miʋɪlɪ
lihína mahína
*hyena
lihówo mahowo
lihowo mahowo
61
cat
lipúsi mapúsi
62
leopard
63
lion
lipusi mapusi
kihʊʋi kihʊ́ʋi - fihʊ́ʋi
fihʊʋi
lihímba lihimba mahímba
mahimba
ntende mitende
mbɪlɪ miʋili
lihina mahina
lihowo mahowo /
libondela mabondela
limbuɲali mambuɲali
suʋi masuʋi
lihimba mahimba
kʊku
ŋgene ʋajene
sukulu /
kʊkʊ
ŋgene ʋajene
Ngoni
Matengo
Mpoto
ánijăja ákaníjăja,
ákiníjăja
nseja
ahokolo
bambu
nkolongu
ń̩do̠ngu
tʃapaŋga
utâmu
-lwala
kílonda ílonda
mitela
19
English
Gloss
Lituhi
Litumba
Kuhamba
64
*fruit-bat
liniminími maniminími
liniminimi maniminimi
65
termite
likele
66
termite
hill
67
Nsungu
Luilo
Iwela
Kisi
Pangwa
kiduli fiduli
ligeke mageke
kidúli fidúli
lindʊlindʊliliniminimi - liniminimi mandʊlindʊ lindʊlindʊli khindulinduli
maniminimi maniminimi
li
ligeke ligeke likele
muhwa
mmehe
mageke
mageke
kisʊɣulu kiduli kisugulu kihwa
khisukulu
fisuɣulu
fiduli
ʋisugulu
goat
mene
mene
mene
mene
mene
mhene
mene
68
cow
ŋómbi
ŋombi
ŋombi
ŋombi
ŋombi
ŋgolombi
69
bull
likambaku makambaku
likambaku makambaku
70
calf
litóli - matóli
likambako makambako
litoli matoli
likambaku likambaku
makambaku
litoli ŋombi
matoli
ndala
ŋguku
ŋuku
ŋguku
liɟogolo maɟogolo
kidege madege
lisatu masatu
liɟogolo maɟogolo
lidege madege
lisatu masatu
liɟogolo maɟogolo
lidege madege
lisatu masatu
somba
somba
ɲɟutʃi
lisosólo masosólo
liʋémbe house fly
maʋémbe
tree
71
72
73
74
75
76
77
78
79
louse (of
body)
ŋgʊ́ku
lisosolo masosolo
liʋembe maʋembe
ŋkoŋgo mikoŋgo /
ukoŋgo /
libihɪ - mabihi libɪhɪ mabɪhɪ
Matengo
cinimánima
kínǐma
likɛlɛ
ń̩keki
ciduli; cidulu
mɛnɛ
Mpoto
limea
kikula
imbuhi
imbui
isenga indala /
ŋɔmbɛ
mbukuma
ng’ômbi
ng'ombi
likida
likambaku
ng’ombi
lipôngu
ŋombi
khikwada
litoli
litoli
litoli
ŋhʊkʊ
ng'uukhu
ŋguku
íngo̠ku
ingoko
liɟogolo
lichokolo
liɟogolo
likóngobi
likoŋgobi
kiʝuni
khideke
cidɛgɛ
kíjuni
kijuni ijuni
lisatu
lihatu
lisatu
lǐhátu
lihatu
somba
somba
somba
sɔmba
hô̠mba
homba
ɲɟusi
ɲɟuki
luʝʊki
luyukhi
luɲɟuci
lijusi
injuchi
lisosolo masosolo
liʋembe maʋe