Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

Introduction

The following paper is a phonology and morphology sketch of Luganda, a (‘Great Lakes’)
Bantu language spoken in Uganda, which was written for the third-year course Linguistic
Fieldwork A hosted by professor Jenneke van der Wal at Leiden University as part of the BA
Linguistics. Over the course of thirteen elicitation sessions, couples of students working
together with a native speaker were able to gather raw data in the form of audio recordings
and elicited word lists. We were tasked to each make our own sketch based on this material.

The language
Luganda is a Bantu language spoken by an increasing number of speakers in and outside of
Uganda, a 2014 consensus of which counted 5,560,000 L1 speakers and 5,390,000 L2
speakers (Eberhard et al., 2022). As we will see through many analyses, this language is
typically Bantu, with relatively exotic articulation types and the usage of noun classes among
other features. It is written in Latin script, and is Uganda’s most widely spoken language,
aside from English which (along with Swahili) is the republic’s official language.

The speaker
Our speaker is an adult male native speaker who was born and raised in the capital city of
Uganda, Kampala. After having visited quite a few places in Europe and South America, our
speaker had settled in the Netherlands about five years ago. He still speaks Luganda with
friends over the phone. Other than Luganda, he noted that English is his second native
language. This is reflective of a very interesting diglossic situation. Not only is English the
official and most widely spoken language of Uganda, it is also a highly prestigious language
up to the point that people do not only use it in formal domains like at the workplace, inside
school buildings and in the media, but even youngsters among themselves like to converse
in English when out in public. It was only at home that the speaker was required to speak
Luganda, which was the native language of his father. His mother's native language was
Lugisu, also known as Masaba. This language variety is also a Northeast ('Great Lakes')
Bantu language and thus closely related to Luganda, but as it is commonly known
throughout the Buganda region, people sometimes refer to it as a dialect. The speaker also
speaks French, Spanish, and is picking up on Dutch, and has been a teacher of both French
and English throughout his career. Despite not having any curricular linguistic training, his
knowledge of his own mother tongue as well as his capability of comparing grammatical
structures from different languages is superb and highly beneficial for our research.

Methods
During each of the thirteen elicitation sessions a pair of students was selected to
audio-record our speaker (using a recorder provided by professor van der Wal) offering us
words based on vocabulary lists they prepared. They and the rest of the class
simultaneously gathered data by writing, and the hosts uploaded the audio files to a
SURF-Drive map and their data to a project page on Dative (app.dative.ca), both exclusively
shared by our class. All of our findings were discussed during a separate weekly lecture and
on a discussion board on Brightspace. Our efforts resulted in every student writing a
phonology and morphology sketch of the language, which will (hopefully) be expanded to a
full-fledged grammar sketch by the end of the academic year (2023-2024).

1
1. Phonology

1.1. Syllables

Although it has become tradition to write a phonology sketch in such a way that consonants
and vowels are discussed before moving on to syllable structure, we will start off our
analysis rather counter-intuitively by looking at syllable structure first. The reasoning behind
this is that understanding syllabification processes in Luganda has proven to be extremely
beneficial to the identification of speech sounds and phonological processes (for an overview
of speech sounds and divergent orthography, see A1. Speech sound inventory).

1.1.1. Syllable types


Displaying a hand-picked set of nouns ranging from smaller to larger strings of syllables, we
see that Luganda makes use of monosyllabic to pentasyllabic structures, with CV functioning
as the common syllable type, strings typically consisting of CV.CV, no consonants in coda
position, and independent vowel nucleus syllables are also possible:
1. Monosyllabic forms: ñá [ɲá] 'four; 4
2. Disyllabic forms: bali [bà.lí] ‘those’
3. Trisyllabic forms: eliñó [ɛ̀ɛ̀.lì.ɲɔ́] ‘tooth’
4. Quadrisyllabic forms: olugálo [ɔ̀ɔ̀.lù.gá.lɔ̀] ‘finger’
5. Pentasyllabic forms: okukáñuga [ɔ̀ɔ̀.mù.ká.ɲù.gà] ‘to throw’

1.1.2. Word-initial vowel lengthening


We see that vowels in word-initial position are always long. We cannot say that this is
lexically determined, since these vowels are part of noun class prefixes which necessarily
latch onto nominal roots in any Bantu language (see A2. Noun classes). We would be much
better off by saying this is a phonological process, in which V / #_ > VV. Long vowels are,
however, not only present word-initially, but also occur word-medially either by being-lexically
determined or as the result or assimilation (see 3.2. Vowel lengthening).

1.1.3. Prenasalization
The first roadblock we run into are forms like we following:

ɛ̀ɛ̀mbútɔ̀ ‘bellies’ ɛ̀ɛ̀ñcímà ‘monkey’


musáàmvù ‘seven; 7’ ɛ̀ɛ̀ŋkúùmbì ‘hoe’

The Indo-European ear might want to perceive such nasal elements as closing a syllable
functioning as its coda, and preceding a following consonant which would be the onset of the
following syllable. We are not dealing with an Indo-European language though. Bantu
languages are known for their exocity, and it would be very close-minded to end our
phonological analysis of nasals here. In fact, it would be detrimental to our theory that
consonants in Luganda can only appear in onset position, and simply making an exception
for nasals would be easy, but lazy, atypically Bantu, and non-satisfying.
In a language like Setswana (which I studied last year) it turned out that nasals were
syllabic in the language. They did not appear in a syllable's coda position underlyingly, as
they were separated and carried a distinct low or high tone when our speaker syllabified
words verbally or by means of whistling. Only sometimes during swift speech it would sound
like we were dealing with *CVN, but only as a result of contraction. A Luganda word like

2
eembúto 'bellies' would thus be syllabified as ɛ̀ɛ̀ / m̀ / bú / tò according to Setswana
phonology (assuming the syllabic m carries low tone). An analysis like this would scrap the
idea of nasals appearing in coda position, but it automatically generates the problem of
nasals being able to function as an independent nucleus, meaning that words can end with a
nasal functioning as the final syllable and that a nasal can display a divergent tone as
compared to its direct environment. Luckily, the way our speaker syllabified it was rather
different: ɛ̀ɛ̀ / mbú / tò 'bellies'
A syllabification method like this allows us to analyze such sequences of a nasal and an
obstruent homogenous in place of articulation as prenasalized consonants, as is typical of
many Bantu languages. Thus we can generate the following transcriptions:

embúto [ɛ̀ɛ̀.ᵐbú.tɔ̀] ‘bellies’ eñcima [ɛ̀ɛ̀.ᶮcí.mà] ‘monkey’


musamvu [mù.sáà.ᵐvù] ‘seven; 7’ eŋkumbi [ɛ̀ɛ̀.ᵑkúù.ᵐbì] ‘hoe’

You will notice that vowels preceding such prenasalized consonants are displayed as short
in the personal orthography whereas they are displayed as long (and falling) in IPA. As we
will see, long vowels outside of word-initial position most often carry a falling tone in the
language, regardless of whether the long vowel is lexically-determined or the compensatory
product of a phonological process (like prenasalization) (see …). We can explain the
occurrence of a long vowel before a prenasalized consonant in the following manner, using
the word musamvu ‘seven; 7’:

*[mù.sà_.mvù] The nasal in coda position loses its mora before it shifts to the
following syllable.
*[mù.sà_.ᵐvù] It then becomes prenasalized onto the obstruent it precedes to form
its new syllable’s complex onset.
*[mù.sàà.ᵐvù] This nasal (element) could not take its mora along with it on the
journey, as the C-onset of its new syllable blocks this nasal from
passing on its mora to the nucleus of that syllable, so the vowel of its
original syllable receives the mora and combines it with its own,
resulting in a long vowel.
[mù.sáà.ᵐvù] This long vowel then automatically receives falling tone.

Adding more weight to the argument, we also see forms starting with a sequence of a
nasal and a consonant, like nsɔ̀mà ‘I read’ and mpúlílà ‘I hear’. Using the syllabic method,
we would analyze the first form as ǹ / sɔ̀ / mà and the second forms as m / pú / lí / là, but
luckily the humming of the word by our speaker showed us that we are dealing with two
syllables instead of three and three syllables instead of four respectively, meaning that the
initial nasal-obstruent sequence combine into one onset, indicating prenasalization:

nsoma [ⁿsɔ̀.mà] ‘I read’ ([σ̀-σ̀] when hummed)


mpúlíla [ᵐpú.lí.ɾà] ‘I hear’ ([σ́-σ́-σ̀] when hummed)

With this analysis, there is yet no coda in sight, nor do we have to account for
double-consonant sequences as we have no reason to analyze prenasalized consonants as
anything other than relatively complex single consonants.

3
1.1.4. Palatalization
The next roadblock on our way to defend CV as the prototypical syllable type are forms like
ɛ̀ɛ̀βjɛ́ɛ̀ɲáàɲɟà '(multiple) fish' and ɛ̀ɛ̀βjɔ́ɔ̀jà 'feathers'. Allowing a [j]-glide as part of a complex
onset automatically leads us to think of palatalized consonants, and palatalization as a viable
secondary articulation type in the language. We can explain this finding away by
incorporating noun classes into our analysis (see A2. Noun classes for an overview of the
noun class system). Luganda, being a Bantu language, makes use of noun class prefixes,
which latch on to nouns, verbs and adjectives to mark agreement to a head noun. As for the
above examples, the story becomes even more interesting when comparing these forms
with their singular counterparts and trying to identify which noun classes they employ:
ɛ̀ɛ̀cɛ́ɛ̀ɲáàɲɟà ‘fish’ and ɛ̀ɛ̀cɔ́ɔ̀jà ‘feather’ respectively. We can identify their noun class prefixes
by comparing them to pairs which share somewhat similar characteristics and the most
similar prefixes:

Noun class 7: PB. *kɪ- > eci- & noun class 8: PB. *ʋi-, di- > ebi- 'various nouns, like
diminutives, manners, ways, languages, inanimate objects (SG)'
eci-bála [ɛ̀ɛ̀.cì.βá.là] ebi-bála [ɛ̀ɛ̀.βì.βá.là] 'fruit(s)’
eci-tétéii [ɛ̀ɛ̀.cì.tɛ́.tɛ́.íì] ebi-tétéii [ɛ̀ɛ̀.βì.tɛ́.tɛ́.íì] 'dress(es)’

Reminding ourselves of the prenasalization process, we can assume that something


similar happened in these environments. After all, it is quite suspicious that we see a long
vowel pop up everytime an underlying /i/ of a noun class prefix seemingly turns into a glide /
palatalized element. Using the word ebi-oya ‘feathers’:

*[ɛ̀ɛ̀.βj.ɔ̀.jà] The underlying /i/ of the noun class prefix becomes devocalized (/i/ > [j]) when
preceding another vowel (the o-nucleus of the following syllable).
*[ɛ̀ɛ̀.βʲ_.ɔ̀.jà] It then becomes palatalized onto the preceding consonant /b/ (~[β]) to form a
complex consonant, leaving behind its mora.
*[ɛ̀ɛ̀.βʲ_ɔ̀.jà] This newly-palatalized consonant allows the following vowel into its syllable
claiming it as its nucleus.
*[ɛ̀ɛ̀.βʲɔ̀ɔ̀.jà] The mora left behind due to the palatalization process gets absorbed by the
vowel, making it long.
[ɛ̀ɛ̀.βʲɔ́ɔ̀.jà] This long vowel then automatically receives falling tone.

As for such forms' singular counterparts (eci-oya [ɛ̀ɛ̀.cɔ́ɔ̀.jà] 'feather', eci-eñañja


[ɛ̀ɛ̀.cɛ́ɛ̀.ɲáà.ᶮɟà] 'fish'), all of the above steps apply but producing a clearly distinguishable
palatalized element /ʲ/ onto a palatal plosive /c/ is quite hard, so we cannot hear it at all even
though phonologically it has been through the same process.
All by all, this does not debunk our hypothesis of CV being the main syllable type, as
palatalization does not seem to be underlyingly present, although data to follow might reveal
something else (see 1.7. Labialization).

1.1.5. Apocope
For much the same reason why the palatalized output (/ʲ/) of underlying /i/ belonging to the
NC7-prefix eci- gets elided when preceding another vowel (due to its host being a palatal
plosive), the very same /i/ can get elided completely when between /c/ and another
obstruent. This means that the whole devocalization into palatalization process does not
apply, but /i/ just gets elided as a form of apocope during swift speech: eci-kopo [ɛ̀ɛ̀.cì.kɔ̀.pɔ̀

4
~ ɛ̀ɛ̀c.kɔ̀.pɔ̀] ‘cup’, eci-tétéii [ɛ̀ɛ̀.cì.tɛ́.tɛ́.íì ~ ɛ̀ɛ̀c.tɛ́.tɛ́.íì] ‘dress’. This does not mean that we
would want to adopt *[ɛ̀ɛ̀c] as being the exception to the 'no-consonants-in-coda' rule. It is a
mere process of apocope, and we are lucky that such forms still allow /i/ to pop up when
pronounced slowly.

1.1.6. Epenthesis
There are also entries which might hint to [j] being used epenthetically, meaning to break up
vowel sequences, like the pair (using the NC5 e(li)- and NC6 ama- prefixes) e-jjembe
[ɛ̀ɛ̀.ɟ:ɛ̂ɛ̂.ᵐbɛ̀] 'horn' vs. ama-yeembe [àà.mà.jɛ̂ɛ̂.ᵐbɛ̀ ~ àà.mà.ɛ̂ɛ̂.ᵐbɛ̀] ‘horns’. We are left
wondering what the root of this pair might be, but the variation in the plural form (the latter
lacking [j] between vowels) proves that the speaker is able to produce vowel sequences and
that [j] might be an epenthetic glide.
Also consider [ɔ̀ɔ̀.kù.lì.jà ~ ɔ̀ɔ̀.kù.ɾʲà] 'to eat' (See 2.7.1. /l/ Alveolar lateral approximant for
/l/ > [ɾ] allophony), which stems from Proto-Bantu *lɪ́.a. We could say that the underlying root
in Luganda is /lì.à/ as well, to which an epenthetic glide [j] is added when pronounced slowly
([ɔ̀ɔ̀.kù.lì.jà]). In swift speech the underlying /i/ palatalizes onto the preceding /l/, which turns
out as [ɾ]) while allowing /à/ into its syllable.

1.1.7. Labialization
Aside from palatalized consonants, we also see forms like ɔ̀ɔ̀lwââlà 'fingernail' and
ɔ̀ɔ̀mwâânà 'child'. These two nouns belong to the noun classes 11 and 1 respectively:

Noun class 1: PB. *mʊ- > omu- & noun class 2: PB. *va- > aba- 'humans, animate entities'
omu-ana [ɔ̀ɔ̀.mʷáà.nà] aba-ana [àà.βáà.nà] 'child(ren)’ (V-contraction)
omu-sájja [ɔ̀ɔ̀.mù.sá.ɟ:à] aba-sájja [àà.βà.sá.ɟ:à] 'man / men’

Noun class 11: PB. *lʊ- > olu- 'abstract nouns (SG)'
olu-ála [ɔ̀ɔ̀.lʷáà.là] eN-jála [ɛ̀ɛ̀.ᶮɟá.la] 'fingernail(s)’
olú-búgo [ɔ̀ɔ̀.lú.βú.gɔ̀] eN-búgo [ɛ̀ɛ̀.ᵐbú.gɔ̀] 'bark cloth(s)’

The familiar story applies, taking the word olu-ála ‘finger’:


*[ɔ̀ɔ̀.lw.á.là] The underlying /u/ of the noun class prefix becomes devocalized
when preceding another vowel (the a-nucleus of the following syllable).
*[ɔ̀ɔ̀.lʷ_.á.là] It then becomes labialized onto the preceding consonant /l/ to form a
complex consonant, leaving behind its mora.
*[ɔ̀ɔ̀.lʷ_á.là] This newly-palatalized consonant allows the following vowel into its syllable
claiming it as its nucleus.
*[ɔ̀ɔ̀.lʷáá.là] The mora left behind due to the palatalization process gets absorbed by the
vowel, making it long.
[ɔ̀ɔ̀.lʷáà.là] The attraction of falling tone onto long vowels seems to be stronger than the
presence of (phonologically-generated) long vowels carrying high tone.

The same can be said for verbal roots starting in a vowel which take on the noun class 15
prefix (PB. *kʊ- > oku-) for denoting infinitives: oku-lúma [ɔ̀ɔ̀.kù.lú.mà] 'to bite', oku-sábá
[ɔ̀ɔ̀.kù.sá.βá] 'to beg' vs. oku-ata [ɔ̀ɔ̀.kʷââ.tà] 'to hold', oku-ogela [ɔ̀ɔ̀.kʷɔ̂ɔ̂.gɛ̀.ɾà] 'to talk'
We also see labialized consonants pop up outside of the NC-prefix slot, as in eci-oku-ñúa
[ɛ̀ɛ̀.cɔ̀.kù.ɲú.wà ~ ɛ̀ɛ̀.cɔ̀.kù.ɲʷá] ‘drink’ and omu-luwádde [ɔ̀ɔ̀.mù.ɾù.wá.d:ɛ̀ ~ ɔ̀ɔ̀.mù.ɾʷá.d:ɛ̀]
'sick'. Here, /u/ gets devocalized and palatalized (and even absorbs a following [w], which

5
might as well be only epenthetic) in swift speech, but luckily more careful pronunciations
reveal the underlying /u/ as such.
Based on all this, we would not have to adapt our draft to fit labialization: *C(ʷ)V. However,
when we moved into NC-inflections in relation to possessive pronouns in later sessions, the
tables were turned forcefully. Labialized consonants started popping up by the dozens. To
take the paradigm belonging to noun class 14 (the plural counterpart of noun class 12):

obu-tale [ɔ̀ɔ̀.bù.tà.lɛ̀] bw-aŋge [bʷáà.ᵑgɛ̀] 'my markets'


bw-ó [bʷɔ́] 'your (sg) markets'
bw-é [bʷɛ́] 'his / her markets'
bw-af:e [bʷà.f:ɛ̀] 'our markets'
bw-amwe [bʷa.ᵐwɛ̀] 'your (pl) markets'
bw-abwe [bʷáà.bʷɛ̀] 'their markets'

Such a paradigm brings along complications. Although (1SG.POSS) [bʷáà.ᵑgɛ̀] and the
first syllable of (3PL.POSS) [bʷáà.bʷɛ̀] neatly display vowel lengthening due the
devocalization of /u/, and such lengthening is blocked or re-shortened in (1PL.POSS)
[bʷà.f:ɛ̀] (see 3.4. Vowel shortening), this lengthening rule does not apply in (2SG.POSS)
[bʷɔ́], (3SG.POSS) [bʷɛ́] or the last syllable of (3PL.POSS) [bʷáà.bʷɛ̀].
Even more, there is the personal pronoun (2SG) gwé, which when the session hosts
requested to be hummed came off as hilarious, but ended up being incredibly insightful as
the speaker does not recognize an underlying structure */gù.ɛ́/ in his humming ([σ́]), and the
typical lengthening rule does not apply either (*[gʷɛ̂ɛ̂]). There are also cases in which
labialized consonants appear outside the NC-prefix slot, which don't seem to reveal an
underlying */u/ instead of a labialization element in slow speech, neither does the
lengthening rule apply, nor does humming reveal something underlyingly present, like:

omu-twé [ɔ̀ɔ̀.mù.tʷɛ́] [σ̀.σ̀.σ́] 'head'


oku-mwémwéntuka [ɔ̀ɔ̀.kù.mʷɛ́.mʷɛ̂ɛ̂.ⁿtù.kà] ‘to laugh exceedingly'

We even have to rethink our hypothesis of non-existence of palatalization in the language,


as the following paradigm (NC5) shocks us in the same way as the last one:

eli-ñó [ɛ̀ɛ̀.ɾì.ɲɔ́] ly-aŋge [lʲââ.ᵑgɛ̀] 'my tooth'


ly-ó [lʲɔ́] 'your (sg) tooth'
lyé [lʲɛ́] 'his / her tooth'
ly-af:e [lʲà.f:ɛ̀] 'our tooth'
ly-amwe [lʲa.ᵐwɛ̀] 'your (pl) tooth'
ly-abwe [lʲââ.bʷɛ̀] 'their tooth'

As a side note, this paradigm of possessive pronouns becomes even more interesting
when looking at other types of noun class inflection. We might want to redraw our whole
sketch of the paradigm stems to suit /áà/ as underlying present, as a forms like [kà.βá.kà
wáà.bʷɛ̀] 'their king' or [àà.mà.ɲɔ́ gáà.bʷɛ̀] ‘their teeth’ cannot be explained through
compensatory lengthening. With this, we will have to account for both labialization and
palatalization in our draft of the prototypical syllable type.

1.1.8. Gemination

6
Knowing that long vowels are highly active in the language, we might not be surprised to find
out that consonant gemination is a feature in the language too: e-ttáka [ɛ̀ɛ̀.t:á.kà] 'soil, land',
ffé [f:ɛ́] 'we', omú-bbi [ɔ̀ɔ̀.mú.b:ì] 'thief'. There is no sufficient data to prove that any of such
geminated consonants are the output of some underlying process (like apocope), and we do
have a near-minimal pair the speaker pointed out to us: o-bútto [ɔ̀ɔ̀.βú.t:ɔ̀] 'cooking oil' vs.
obutó [ɔ̀ɔ̀.βù.tɔ́] 'little'. One notable feature of gemination is that it blocks the compensatory
lengthening process of a preceding vowel (see 3.4. Vowel shortening).

1.1.9. Main syllable type(s)


It seems to be quite hard to come up with a prototypical syllable type from which all others
can be derived. Constructing such an extreme prototype would result in *C(ʷʲ)(:)V(:), but this
is highly unnatural as a full realization of all these features combined does not exist in the
language. There is the example of omu-sambya [ɔ̀ɔ̀.mù.sáà.ᵐbʲà] 'hard type of tree', where
we see prenasalization and palatalization on one consonant, but this is the only entry in the
entire database which displays such a capability. Also, we can never be sure that a long
vowel following a palatalized or labialized consonant is lexically-determined instead of
phonologically-produced, although the latter is likely the case most of the time. I personally
have not heard any geminated palatalized or labialized consonants, and gemination blocks
lengthening of a preceding vowel. A better solution would be to propose commonalities: CV
being the most common and single V being possible too; C of which might be either
palatalized, labialized or geminated; V can be long, but gets shortened when following a
geminated C.

1.2. Consonants

Moving on to consonants then, Luganda makes use of of a small consonant inventory (14) if
only counting underlying phonemes of ‘basic’ speech sounds (disregarding secondary
articulation), moderately small if taking into account their allophones and potential loan
phonemes (17)), and average (24) if prenasalized counterparts of the 8 plosive phonemes
and 4 fricative phonemes count as well (https://wals.info/chapter/1).

1.2.1. Consonant inventory

Bilabial Alveolar Palatal Velar

Plosive (p) /t/ /c/ /k/


/b/ ~ [β] (d) /ɟ/ < j > /g/

Nasal /m/ /n/ /ɲ/ < ñ > /ŋ/

Fricative /f/ /s/


/v/ /z/

Approximant /w/ /j/ < y >

Lateral /l/ ~ [d:], [ɾ] < r >


approximant

7
1.2.2. Approximants
We will start off with approximants, as in the grander scheme of things one particular
approximant shares close ties to another speech sound.

1.2.2.1. /l/: Alveolar lateral approximant


The alveolar lateral approximant (w.e.l.a.) [l] is an interesting segment in that it seems to be
in complete free variation with the alveolar tap [ɾ]. We caught on to this variation immediately,
and words which have been cited as having [l] are often repeated but pronounced with [ɾ]
instead, or the other way around:

e-beele [ɛ̀ɛ̀.bɛ̀ɛ̀.lɛ̀ ~ ɛ̀ɛ̀.bɛ̀ɛ̀.ɾɛ̀] 'breast'


mp-úlíla [ᵐpú.lí.là ~ ᵐpú.lí.ɾà] 'I hear'
wa-gulu [wà.gù.lù ~ wà.gù.rù] 'up'

This feature also carries over to the speaker's use of English, which he speaks natively,
but this part of Luganda phonology seems to be integral to such a degree that it hinders a
more standard pronunciation of English: 'belly' ['bɛ.ɾi], 'colour' ['kɔ.ɾɔr], 'cloth' [kɾɔθ].
Remarkably, there were instances during which the speaker persisted that a word needed
to be pronounced with an 'L' (using an English orthographical term), but his pronunciation of
the word sometimes indicated that the phonetic product contained a tap instead: oku-liyá
[ɔ̀ɔ̀.kù.lì.já ~ ɔ̀ɔ̀.kù.ɾʲá] 'to eat' ("It is L."). Such a statement could be based on Luganda
orthography, meaning that the word in written form does contain an 'L'. However, the
language's phonology makes it harder to distinguish between the two allophones in speech.
One constraint (a degree of predictability) we might be able to place on this type of
allophony could be that when multiple occurrences of /l/ are present in a word, either the
latter gets pronounced as [ɾ] or both are pronounced as [ɾ]; [ɾ] is never able to precede [l]
later in the word. The reasoning behind this could be an urge for dissimilation. As multiple
instances of [l] are quite hard to pronounce in swift speech without creating a stutter,
following (or all) approximants turn into something more tap-like ([ɾ]).

Non-assimilation
olu-gálo [ɔ̀ɔ̀.lù.gá.lɔ̀] 'fingers'
olu-ala [ɔ̀ɔ̀.lʷáà.là] 'fingernail'

[l] preceding [ɾ] (dissimilation)


oku-úlíla [ɔ̀ɔ̀.kù.ú.lí.ɾà] 'to hear'
omu-landíla [ɔ̀ɔ̀.mù.láà.ⁿdí.ɾà] 'root'
olu-alelo [ɔ̀ɔ̀.lʷáà.lɛ̀.ɾɔ̀] 'today' (*[ɔ̀ɔ̀.lʷââ.lɛ̀.lò] is incorrect)

Double assimilation
oku-úlíla [ɔ̀ɔ̀.kù.ú.ɾí.ɾà] 'to hear'
omu-landíla [ɔ̀ɔ̀.mù.ɾáà.ⁿdí.ɾà] 'root'

Even so, there are entries present in the data set in which only one [ɾ] is present without it
being pronounced as [l] when repeated, like eci-ré [ɛ̀ɛ̀.cì.ɾɛ́] 'cloud', eci-gere [ɛ̀ɛ̀.cì.gɛ̀.ɾɛ̀] 'foot'
and oku-ógera [ɔ̀ɔ̀.kʷɔ́.gɛ̀.ɾà] 'to talk'. We might rather want to say the allophones are, in
principle, in free variation, with the process of dissimilation being quite common.

8
This phoneme is also one to assimilate completely, meaning both in place and manner of
articulation. One such case is when it turns into [n] when following an alveolar nasal.
Compare the adjectival root lunji ‘good, handsome’ in the following forms:

omu-sájja omu-lunj [ɔ̀ɔ̀.mù.sá.ɟ:à ɔ̀ɔ̀.mù.ɾûû.ᶮɟ] ‘handsome man’


aba-sájja aba-lunj [àà.bà.sá.ɟ:à àà.bà.ɾûû.ᶮɟ] ‘handsome men’

eN-uumba eN-lunj [ɛ̀ɛ̀.ɲûû.ᵐbà ɛ̀.n:ûû.ᶮɟ] ‘good house’


ama-uumba ama-lunj [àà.mà.jûû.ᵐbà àà.mà.ɾûû.ᶮɟ] ‘good houses’

The most fitting theory would be to propose that the underlying nasal of eN- assimilates to
the place of articulation of the following consonant (see XX. Noun classes 9-10 and
prenasalization), resulting in en-. Then the /l/ of the root assimilates its manner of articulation
to the preceding /n/, resulting in gemination. This finding is stunning, as we can now say that
only sequences of a nasal plus an obstruent (plosive or fricative) results in a prenasalized
consonant, and its combination with an approximant like /l/ does not. Hence why we don’t
see usage of *ⁿl in the language. The same process applies when looking at the pair olu-lími
[ɔ̀ɔ̀.lù.lí.mì] 'tongue' and eN-lími [ɛ̀.n:í.mì] 'tongues', the singular using NC11 prefix olu- en
the plural using NC10 prefix eN-, along with the root -lími 'tongue' (see 3.4. Vowel shortening
for shortening of word-initial vowels).

2.7.2. /w/: Labio-velar approximant


For many languages we could pose the question whether [w] is a real underlying phoneme
or whether it is the product of devocalization attaining semi-vowel status. As for the voiced
labio-velar approximant (w.e.l.a.) /w/, we have seen that devocalization is indeed a
phonological process in Luganda (see 1.1.7. Labialization). However, to propose that each
instance of [w] is actually underlyingly /u/ and that the clash of two nuclei (*u.V) caused this
devocalization would be more of a historical analysis rather than a descriptive one, even
though data on [w] as a simple onset consonant is scarce: (noun class 16: PB. *pa- > wa-
'locative') wa-gúlu [wà.gú.ɾù] 'up', wa-ansí [wáà.ⁿsí] 'on the ground' (contraction); NC1
possessive pronouns w-aaŋge [wáà.ŋgɛ̀] 'my', w-aaffe [wà.f:ɛ̀] ‘our’, etc.

2.7.3. /y/: Palatal approximant


The very same is true for the voiced palatal approximant (w.e.l.a.) /j/. We have seen [j] as a
potential epenthetic glide ([àà.mà.jɛ̂ɛ̂.ᵐbɛ̀ ~ àà.mà.ɛ̂ɛ̂.ᵐbɛ̀] 'horns', see 1.1.6. Epenthesis) and
as a devocalized into palatalized element (ebi-eñaɲja [ɛ̀ɛ̀.βʲɛ̀.ɲââ.ᶮɟà] '(multiple) fish'), but we
cannot deny, although our data is scarce, that [j] does pop up as an onset consonant and the
speaker recognizes it as such when humming or syllabifying (instead of a separate nucleus
syllable */i.V): NC9 possessive pronouns y-é [jɛ́] ‘his / her’, y-aabwe [jââ.bwɛ̀] ‘their’. Also,
geminated [j:] is possible, which proves underlying */i/ to be quite a stretch: y:é [j:ɛ́] ‘he / she
(in discourse, not currently present)’.

9
We also see the reverse situation happening, and there are truly some interesting quirks
to this /l/-phoneme. When it, as part of a prefix, precedes a nasal belonging to the root, the
following nasal assimilates its manner of articulation to the lateral approximant, resulting in a
geminate lateral, which automatically fortifies into [d:] (see 2.2.2. Alveolar plosive):

e-ttáka *ddéne [ɛ̀ɛ̀.t:á.kà d:ɛ́.nɛ̀] 'piece of land'


ama-ttáka ma-néne [àà.mà.t:á.kà mà.nɛ́.nɛ̀] 'pieces of land'

Extracting the root of the modifier as *néne from the plural form inflected for noun class 6,
ama- for nouns and ma- for modifiers, its singular counterpart must be inflected for noun
class 5, meaning e(li)- for nouns and a similar prefix for modifiers, most likely not carrying a
form-initial vowel but maintaining /l/ (*l(i-)), which clashes with the following /n/ of the root
*néne.

10
A1. Speech sound inventory

Bilabial Alveolar Palatal Velar

Plosive (p) /t/ /c/ /k/


/b/ ~ [β] (d) /ɟ/ < j > /g/

Nasal /m/ /n/ /ɲ/ < ñ > /ŋ/

Fricative /f/ /s/


/v/ /z/

Approximant /w/ /j/ < y >

Lateral /l/ ~ [ɾ] < r >


approximant

Front Back

High /i/ /u/

High-mid /ɛ/ < e > /ɔ/ < o >

Low /a/

11
A2. Noun classes

NC Number Proto-Bantu Luganda Function

1 SG *mʊ- omu- Humans, animates.


2 PL *va- aba-

3 SG *mʊ- omu- Plants, inanimates.


4 PL *mɪ- emi-

5 SG *lɪ- eli- Various nouns, including liquids and mass


6 PL *ma- ama- nouns.

7 SG *kɪ- eci- Various nouns, including diminutives,


8 PL *vi-, *di- ebi- manners, ways, and languages.

9 SG *n- eN- Animals and inanimates.


10 PL *di-n- eN-

11 SG *lʊ- olu- Abstract nouns.


(10) PL *di-n- eN-

12 *ka- aka- Diminutives.

15 *kʊ- oku- Infinitives.

16 *pa- wa- Locatives.

12

You might also like