(revised April 2010)
© 2008 SIL International
®
Library of Congress Catalog Number: 2008935395
ISBN: 972-1
555671-255-5
ISSN: 1934-2470
Books published in the SIL e-Books (SILEB) series are intended for scholarly research and
educational use. You may make copies of these publications for research or instructional
purposes free of charge (within fair use guidelines) and without further permission.
Republication or commercial use of SILEB or the documents contained therein is expressly
prohibited without the written consent of the copyright holder(s).
Mary Ruth Wise
Bonnie Brown
Lois Gourley
iv
The Bantu Orthography Manual is a resource for developing writing
systems among the Bantu subgroup of Niger-Congo languages. It oers a
strategy for orthography development, combined with a list of resources
for Bantu linguistic information and the condensed advice of a coterie of
respected Bantu linguistic experts. It oers readability and write-ability
considerations whenever applicable. The Manual has a target audience:
linguists gathering information for orthography development.
Procedures for a “participatory approach” to phonological analysis are
described in one chapter. A series of charts help the linguist document
and organize the phonological and morphological information gathered.
The Linguistic Features chapter lists common linguistic characteristics of
Bantu languages, describes the attendant orthographic challenges, and of-
fers suggested solutions, along with the pedagogical rationale for each.
v
Denition of orthography ......................................................................................................2
The purpose ...........................................................................................................................2
The goal .................................................................................................................................2
Foundational assumptions for orthography development .....................................................2
Elements of the development process ....................................................................................3
Purpose of this manual ..........................................................................................................3
Introduction ...........................................................................................................................4
List of graphemes for the segmental aspects of a Bantu writing system ................................7
Vowels ..............................................................................................................................7
Consonants .......................................................................................................................7
Vowels ..................................................................................................................................10
Vowel systems ................................................................................................................10
Vowel harmony ...............................................................................................................12
Vowel length ..................................................................................................................14
Predictable shortening of an underlyingly long vowel ..................................................21
Semivowels ....................................................................................................................22
Vowel elision (also a word break issue) .........................................................................25
Consonants ..........................................................................................................................26
Liquids ............................................................................................................................26
Fricatives ........................................................................................................................27
Nasals .............................................................................................................................27
Aricates ........................................................................................................................29
Implosives ......................................................................................................................31
Double articulated stops ................................................................................................31
Geminate consonants .....................................................................................................31
Prenasalization ...............................................................................................................32
Labialization and palatalization .....................................................................................33
Velarization ....................................................................................................................36
vi
Morphophonological Processes ...........................................................................................36
Phonological conditioning at morpheme boundaries .....................................................37
Vowel coalescence/assimilation .....................................................................................39
Principles of Tone Marking in an Orthography ...................................................................40
Verbs ..............................................................................................................................42
Nouns .............................................................................................................................43
Functors .........................................................................................................................43
Word Boundary Principles ...................................................................................................43
Clitics vs. axes: two dierent animals .........................................................................45
Augment vowels .............................................................................................................47
Associative marker (wa, ba, gwa, gya, etc.) ...................................................................49
The conjunction na .........................................................................................................51
Vowel elision / coalescence in clitics .............................................................................51
In verbs ...........................................................................................................................52
Generalizations ...............................................................................................................53
Troubleshooting Problem Areas ..........................................................................................54
Probable Test Needs ............................................................................................................54
Who Should Be Tested .........................................................................................................55
Types of Test ........................................................................................................................55
Informal vs. Formal ........................................................................................................55
Giving formal tests .........................................................................................................55
Appendix A: Determining Functional Load for Tone ...........................................................60
Appendix B: Glossary of Linguistic Terminology .................................................................63
Appendix C: Morphological vs. Phonological Rules .............................................................65
vii
The goal ...............................................................................................................................73
Foundational Assumptions for Orthography Development ..................................................73
Goals ....................................................................................................................................73
Linguistic rationale ..............................................................................................................73
Overview ..............................................................................................................................74
Initial Orthography Development: studying phonology and morphophonology... ..........74
Discovery of contrastive sounds ..................................................................................74
Phonotactics: the combinations and combinatory restrictions of
Consonants and Vowels, particularly in root-initial position. .....................................74
Morphology and Morphophonology ...............................................................................74
Approach: Participatory Research ...................................................................................75
Word Collection ...................................................................................................................76
Goals of word collection phase .......................................................................................76
Word collection: wordlists vs. semantic domains ...........................................................76
Roles of Participants .......................................................................................................77
Instructor ........................................................................................................................78
Computer Scribe .............................................................................................................78
Contact person with hosts ...............................................................................................78
Group scribe ....................................................................................................................78
Discussion facilitator .......................................................................................................79
The groups ......................................................................................................................79
Recommended Schedule .................................................................................................79
Elicitation process ...........................................................................................................80
Data Entry ............................................................................................................................81
Computer tools to assist in word collection and card printout .......................................81
Printed word collection forms .........................................................................................81
Keyboarding words .........................................................................................................81
Printing cards ..................................................................................................................82
Other printouts ...............................................................................................................82
Challenges: ..................................................................................................................82
viii
Preparation for Phonological Analysis ............................................................................83
Step 1. Weed out all compound nouns as well as loanwords and phrases ..................83
Step 2. Print out the word cards .................................................................................84
Step 3. Generate a phonology report ..........................................................................84
Workshop use of cards and laptops: an overview ................................................................84
Vowels, Week 1 ....................................................................................................................87
Week 1, Vowels ....................................................................................................................87
Vowels Step 1. Discover lexical roots of nouns with C-initial roots ................................87
Vowels Step 2. Chart distribution of each vowel within nouns,
in which V1 = V2...........................................................................................................88
Vowels Step 3. Study vowel combinations, -CVCV noun roots in which
V1≠V2. .........................................................................................................................88
Vowels Step 4. Discover lexical roots of verbs. Study vowel distribution
in verbs ...........................................................................................................................90
Step 5. Study vowel length .............................................................................................90
Consonants, Week 2 .............................................................................................................91
Consonants Step 1. Look at the CVCV noun roots, sorting by C1 (and V1) ....................91
Consonants Step 2. Sort nouns by C₂ ..............................................................................93
Consonants Step 3. Look for complementary distribution of consonants ........................93
Consonants Step 4. Sort consonants in the CVC verb roots, px–C
1
VC-V .........................94
Consonants Step 5. Sort consonants in the CVC verb roots, px–C
2
VC-V .........................94
Consonants Step 6. Finalize list of consonant phonemes ................................................95
Consonants Step 7. Study and document V-initial roots of nouns and verbs
and vowel harmony (may be done by TA and a few speakers of the
language while others work on Step 9) ...........................................................................95
Consonants Step 8. Identify all noun class prexes (and variants) .................................96
Consonants Step 9. Make alphabet charts .......................................................................97
The Alphabet ........................................................................................................................97
Alphabet Chart Considerations .......................................................................................98
Unresolved alphabet issues ...........................................................................................100
List Revision and Text Collection: Ongoing activities this week and next .........................100
Tone, Week 3 .....................................................................................................................101
Tone Week Step 1. Study tone on verbs ........................................................................101
Tone Week Step 2. Study tone on nouns .......................................................................102
ix
Tone Week Step 3. Record function of tone, nalize wordlists and
alphabet charts..............................................................................................................105
Tone Week Step 4. Edit and review ..............................................................................105
Tone Week Step 5. Close workshop ..............................................................................106
Things to do each time you show someone the wordlist: informal testing of preliminary
orthography (Translate) .....................................................................................................107
Orthography Sketch and Writers’ Guides ...........................................................................109
Nouns, Week 1 ...................................................................................................................110
Step 1. Begin charting adjectives ..................................................................................110
Step 2. Chart noun phrase numerals .............................................................................110
Step 3. Chart demonstrative phrases .............................................................................111
Step 4. Chart demonstrative forms: other .....................................................................111
Step 5. Chart interrogative forms ..................................................................................111
Step 6. Chart pronominal forms-1 .................................................................................111
Step 7. Chart pronominal forms-2 .................................................................................111
Step 8. Chart pronominal forms-3 .................................................................................111
Step 9. Chart diminutives and augmentatives ...............................................................111
Step 10. Chart possessive pronouns ..............................................................................111
Step 11. Locatives .........................................................................................................112
Step 12. Chart associative constructions .......................................................................112
Verbs, Week 2 ....................................................................................................................112
Step 13. Chart Copular Forms .......................................................................................112
Step 14. Chart Copular Forms Part 2 ............................................................................112
Step 15. Chart Morphophonological Processes, Perfective ...........................................112
Step 16. Chart Verbal mood: Imperative Plural ............................................................112
Step 17. Chart Verbal mood: Imperative Singular ........................................................112
Step 18. Chart Verbal mood: Subjunctive .....................................................................113
Step 19. Chart relative phrases, Object Relative and Subject Relative..........................113
Step 20. Chart Use of the Augment on Nouns, continue charting verb forms ...............113
Step 21. Check charts with a linguist ............................................................................113
Verbs and checking of charts, Week 3 ...............................................................................114
Step 21. Print charts .....................................................................................................116
Step 22. Edit stories in groups, referring to the writing rules the group chose during
group charting ..............................................................................................................116
x
Step 23. Closing ceremony and distribution of printed materials .................................116
Appendices
Appendix A: Alphabet Chart Template .........................................................................117
Appendix B: Lesson 1: Goals and products of Workshop 1 ...........................................118
Appendix C: Lesson 2: Helping Readers with our Spelling Decisions............................121
Appendix D: Lesson 3: Consonant Symbol Choices .......................................................125
Appendix E: Lesson 4: Morphemes and Words .............................................................129
Appendix F: Lesson 5: For TAs: Spelling Principles.......................................................131
Appendix G: Lesson 6: Goals and Products of Workshop 2 ...........................................132
Appendix H: Lesson 7: Word Break ...............................................................................137
Appendix I: Lesson 8: Story Editing ..............................................................................141
Appendix J: Noun-Class Prexes for Proto-Bantu .........................................................142
Appendix K: Duruma Noun-Class Morphology..............................................................145
Appendix L: Orthography Workbook Forms .................................................................148
Contents ........................................................................................................................150
1
The Bantu Orthography Manual is not a computer tool but a resource to aid in
orthography decisions. It oers a suggested strategy for orthography development with
a workshop approach, combined with a list of resources for Bantu linguistic information
and the condensed advice of a coterie of respected Bantu linguistic experts: Rod Casali,
Myles Leitch, Oliver Stegen, Constance Kutsch-Lojenga, Ken Olsen, Mike Cahill, Karen Van
Otterloo, and Bill Gardner.* It oers readability and write-ability considerations whenever
applicable, and is written with an informal style to facilitate easy access to information.
Part 1, Bantu Orthography Manual, has a target audience: the Ordinary Working
Linguist (OWL).
1
As such, it does not oer the degree of linguistic detail one can nd in
such extensive, deep and informative collections of Bantu linguistic information as Nurse
and Philippson (2003), though it refers the reader to many such resources.
Part 2, Participatory Approach to Orthography Development provides step-by-step
instructions for carrying out a participatory approach to orthography development, a
concept originated by Constance Kutsch-Lojenga. Its style is less formal and very practical.
The process involves workshops with groups of participants possibly representing several
related Bantu languages.
While the Manual has an academic focus, the Participatory Procedural Guide is less
formal and very practical. It attempts to guide an OWL through a participatory approach
to orthography development. It hyperlinks to the Bantu Orthography Manual, which
should be used often as a reference, as well as a series of eight Literacy Lessons, plus
Proto-Bantu morphology charts and a large Orthography Workbook Form. All are designed
for use during a workshop.
The Linguistic Features chapter of the Manual provides suggestions to help the linguist
and mother-tongue speakers choose symbols to represent the linguistic features of their
languages. It gives an overview of many signicant phonological and morphological
processes as they interact in Bantu languages. Examples of the orthographic challenges are
given and advice is oered as to how these phenomena might best be reected in a writing
system.
2
This chapter can also be used as a complement to the Bantu Phonology Tool, a
computer tool for phonological analysis of Bantu languages.
The principles and suggestions given try to incorporate pedagogical, perceptual, and
sociolinguistic considerations into the decision-making process for graphemic choices. In
this way, it is hoped that the writing system for a given Bantu language will become the
possession, the useful tool, of the mother-tongue speaker.
*
Special thanks go to these contributors: Rod Casali, Constance Kutsch-Lojenga, Bill Gardner, Myles
Leitch, Karen Van Otterloo, Oliver Stegen, Keith Snider, and Helen Eaton.
1
The phenomena described focus on East and Central Africa Bantu languages, with the hope of publishing
a future edition that reects a wider area. Swahili is often referred to as a typical language of wider com-
munication. However, the procedure can be adapted to other parts of the Bantu-speaking world, where other
languages function as the LWC.
2
Native speaker perception and preferences are embedded in our participatory approach to phonological
analysis, which precedes the graphemic decisions, and with orthography testing in local communities
(addressed in other chapters).
2
Consists of two parts:
1. Symbols (in our case, alphabets)
2. Writing rules
• Alphabets require linguistic input from the Bantu dictionary and phonology tools
• Writing rules need these, plus input from Bantu grammar and discourse tools
• Both alphabets and writing rules need input from psycho-linguistic, sociolinguistic,
educational and political realities in the context
• Empowerment of mother-tongue speakers to read and write their languages as a result
of their own informed orthography decisions
• Advancement of academic knowledge and professional skills, both our own and those
of our partners
Develop an orthography for a Bantu language group, in partnership with speakers of that
language (laying a linguistic and educational foundation for the application of that orthog-
raphy to reading and writing)
• Community ownership/involvement in every stage of the orthography development
process, not only to develop orthographies which are accepted by local communities,
but to encourage local understanding of the rationale underlying orthography decisions
• The centrality of linguistic analysis to orthography development, inclusive of
phonology, morphophonemics, grammar and discourse (since these aspects of a spoken
language overlap and inuence one another on the surface, they will seem to compete
with one another for prominence in a writing system)
• Mother-tongue speakers’ perception should play a signicant role in orthography
decisions. That perception can be developed and enriched for those who take part in
the orthography development process
• Orthography-in-use as goal, and also as means for constant feedback and evaluation
• Questions of readability and write-ability will be considered throughout the
development process
3
Chapter 2 - Representing Linguistic Features
• Revisions will be ongoing, and will necessarily reect the political, educational and
social context of the writing system
• The Orthography Development tool is more a concept or a process than a nely tuned
mechanism; i.e., it will not require a computer program
• Templates for the products to be produced while applying the tool/concepts could be
made (i.e., for a Word book, an Alphabet Book, and/or a Writers’ Guide or transition
primer).
• Workshops to produce these and plans for testing followed by application in
communities are the actual tool (or the format/vehicle for implementation)
• Interaction with Ordinary Working Linguists, or OWLs (people who are now developing
and testing their linguistic analyses and applications), to see what challenges they have
encountered in their work and what info (and wisdom) they’ve gained as a result.
A suggested sequence for such an approach is given in Chapter 5, with activities listed
under each objective. The attendance and involvement of local language speakers in each
activity described is assumed; ordinary working linguists and data entry personnel will
also be involved in most of the activities listed.
This orthography manual is not a complete guide for phonological analysis, but is de-
signed for applied linguistics, specically orthography development. It is intended as an
aid toward development of trial orthographies for Bantu languages, using a workshop ap-
proach. It assumes the reader has some familiarity with the grammatical and phonological
characteristics of Bantu languages, and is therefore brief in its descriptions of some Bantu
linguistic phenomena. It contains references, as well as footnotes, and a list of suggested
background reading materials. The last appendix in Part 2 (Appendix L: Orthography
Workbook Form) is a worksheet to be lled in during a series of two workshops.
4
This chapter provides suggestions to help the linguist and mother-tongue speakers
choose symbols to represent the linguistic features of their languages. It gives an over-
view of many signicant processes that occur in Bantu languages and how these phenom-
ena might best be reected in a writing system. We begin with an acknowledgement that
linguistic features are not the only factors in orthography development: “On the surface,
orthography selection and development are linguistic issues; but in practice they are loaded
with imperatives arising from a number of sources” (Eira 1998:171).
Orthographies are complex visual representations of language and thought which are
designed to facilitate literate communication. Because of the intertwining of phonology,
morphology, and several cognitive processes in the decoding of print, sociolinguistic, and
pedagogical considerations for orthography decisions will be presented in this chapter,
though it is organized around linguistic features of Bantu languages. Just representing
sounds with graphemes is not enough. Language is more than sound, and reading and
writing are more than recognition and transcription of sounds alone. A good orthography
enables readers to quickly recognize meaning, and its spelling rules are as simple as pos-
sible, to aid the writer.
Solid linguistic analysis, while not the only factor in orthography decisions,
1
is essential
to the development of a good orthography. It under-girds any writing system which ac-
curately represents the speech and perception of the mother-tongue speaker. Toward that
same end, the following principles and suggestions try to incorporate pedagogical, per-
ceptual, and sociolinguistic considerations into the decision-making process for graphemic
choices. In this way, it is hoped that the writing system for a given Bantu language will
truly become the possession, the useful tool, of the mother-tongue speaker.
William Smalley’s ve criteria for orthography decisions (1964:34) are still widely used
today, though he had transfer from mother-tongue to the language of wider communica-
tion (LWC) in mind. In Africa today, our situation is often the reverse, with mother-tongue
literacy coming later. His criteria with my annotations are:
1. Maximum motivation for the learner (Smalley refers to issues of ethnic and
national identity)
2. Maximum representation of speech (phonemic writing, with some exceptions
where morphology aects phonology, and vice versa)
3. Maximum ease of learning (simplicity of rules)
4. Maximum transfer [to/from] literacy in a LWC, and
5. Maximum ease of reproduction (for publication purposes)
1 “
A better use of linguistics is as a source of insights about orthography options, and as a tool to probe the
orthographic insights of native speakers” (Bird 2000:29).
5
Chapter 2 - Representing Linguistic Features
One further point could be added:
6. Maximum recognition/transmission of meaning.
Smalley alludes to preservation of morphemes in his second point. Bantu languag-
es, though, are particularly capable of packing multiple morphemes within one word.
Phonological processes alter the surface structure of those adjoining morphemes signi-
cantly. Since consistency of morpheme shape facilitates spelling, reading uency, and
comprehension of written text, a Bantu writing system’s ability to help the reader recognize
morphemes merits focused attention.
This orthography manual is not a guide for phonological analysis, but for applied linguis-
tics, specically orthography development. It is necessarily specic in its focus; it is not a
substitute for general knowledge of orthography principles. I advocate familiarity with Bird
(1998), Gardner (2001, 2005), Kutsch-Lojenga (1993), Smalley (1964), Snider (1998, 2005),
and Ssemakula (2005). These sources will give you some generalizations about the phono-
logical processes and morphological structures of Bantu languages or explain basic orthog-
raphy principles not detailed in this manual. A course on orthography principles is also
oered in the summers at the University of North Dakota.
A word about shallow versus deep orthographies is necessary here. There is a continuum
between deep, meaning-based writing systems in which the primary units of meaning
represent morphemes (such as Mandarin), and more shallow, alphabetic writing systems
which are based on phonemes, such as Italian, Serbo-Croatian, and Luganda. Deep orthog-
raphies represent the morphology of the language more than they do the phonology, while
shallow orthographies aim to closely represent the phonology of the language. Even within
alphabetic writing systems a variety of depth exists, between the more morpheme-preserv-
ing orthographies such as English and French and the more phoneme-based orthographies
of Swahili and Lugungu.
An ideal orthography matches the language it represents
. “Shallow orthographies are
characteristic of languages in which morphemic relatives have consistent pronunciations”
(Mattingly 1992:150). At the opposite end of the spectrum are languages which require a
deep orthography. They have morphemes which are subject to a lot of phonological varia-
tion and/or many homophones which must be distinguished in writing.
There is a tension between the strengths of each kind of orthography. The shallowest
orthography gives the new reader/writer lots of control. With recognition of a handful
of symbols he/she can recognize and spell anything they can speak or say, based upon a
one-to-one correspondence between sound and symbol. But no writing system is actually
capable of completely representing speech. Both readers and writers are assisted by writ-
ing according to word-level phonology, rather than phrase-level phonology. Moving towards
a deep orthography allows homophones to be distinguished while words whose pronun-
ciation varies in context (the dog vs. the apple) can be given a xed representation. This
allows the experienced reader to access meaning rapidly. The issue of morphology and its
eect upon the phonology of a language (morphophonological alternation) will arise very
frequently in the process of orthography development for a Bantu language.
Orthographies should usually be developed with the mature reader in mind. Therefore,
preserve morphemes when possible, especially if a surface-level representation seems to be
6
Bantu Orthography Manual
in equal competition with a morpheme for graphemic representation. Think beyond the in-
dividual segment. Think of a whole word. Lexical changes which take place within a word
are usually written. (For more explanation, see Part 1, Appendix D.)
2
Lugungu, for example,
has a rule that changes the phoneme /r/ into the phoneme /d/ when a morpheme ending
in /n/ is prexed. That morphophonemic change, /kuruga/>/nduga/ when the prex /n/
is added, takes place within the boundaries of a word and should be written.
When it’s possible (sociolinguistically) to avoid digraphs, do so—but you must consider
the transferability of your symbols to the LWC (language of wider communication) in your
area. If the LWC is Swahili, your spelling options are rather narrow.
Spell the way people speak in slow speech, rather than rapid conversation. Elision of seg-
ments dropped in rapid speech is not recommended for several reasons:
1. If a phoneme or a morpheme is written fully, people can recognize meaning
quickly. For example, the Bantu associative marker is /Ca/ or /CCa/. The initial
consonant of this word changes depending upon the class of the noun it modies,
leaving only the word-nal /a/ to give it consistency of appearance. In rapid
speech, this one vowel is often elided and replaced, if the noun it precedes
is vowel-initial. For example, in Ikizu (E402J), the associative marker has an
underlying <> as its vowel, but because it is pronounced together with the
following word, in rapid speech the <> coalesces with a vowel-initial augment
and elides. The associative, a word in its own right, is written with the vowel
that would ordinarily be the augment of the following noun. Visually, then,
recognition of the associative marker is completely lost to the reader whenever it
precedes a vowel-initial noun.
2. If a phoneme is written fully, ambiguity is kept at a minimum. For example, if
a certain vowel is elided in rapid speech, if it is not represented in writing, the
reader may be unsure which vowel is missing.
3. The apostrophe, generally the symbol of choice for indicating elision, is used in
the standard Swahili alphabet as a part of the trigraph <>. When the LWC
includes this trigraph, mother-tongue reading uency is slowed and accuracy of
spelling may be hindered if the apostrophe is further used to indicate elision or
phonological attachment of clitics. Apostrophes have several uses in many Bantu
orthographies, sometimes indicating a glottal stop (in Cameroon), the attachment
of clitics to larger words, the deletion of a vowel across word boundaries, or as
part of the consonant digraph <>. This multi-purpose usage of <> slows
uency of reading and hinders accuracy of spelling. Avoid such multiple uses of a
symbol.
At times (rarely) for sociolinguistic reasons, an orthography may have to represent a pho-
netic sound [ ], rather than just representing the contrastive phonemes / /, of a language.
For example, a language group may insist upon writing a distinction between [l] and [ɾ] for
purposes of reading transfer to a LWC, even though they may not be contrastive phonemes
for their language group. The symbols for sounds are enclosed in [ ] and the graphemes
in < >.
2
This principle takes on great importance when spelling and word break decisions are made, for associa-
tive markers in particular (also see section on word breaks).
7
Chapter 2 - Representing Linguistic Features
The best grapheme options are listed rst; less desirable ones follow them. The format for
each grapheme suggestion will be like this one:
ɑ:] Options: <>
1. Vowel Systems
2. Five-vowel systems: <>
3. Seven-vowel systems: <> or <>
4. Vowel Length: <>
5. Semivowels: <> and <>
6. Vowel elision: see Vowel elision
1. Liquids
[l] Options: <> or<>
[ɾ] Options: <> or <>
2. Fricatives
[ð] Options: <> or <>
[Ɵ] Options: <>
[ɤ] Options: <> or <>
[ß] Options: <> or <> or <>
This list is not exhaustive
3. Nasals
Velar Nasal
[ŋ] Options: <> (or <>or <> or <>
Palatal nasal [ɲ] (see “Labialization and palatalization” below)
Homorganic nasalization
4. Aricates
[tʃ] Options: <> (or <> or <> or <>)
[ts] Options: <> or <>(less desirable Francophone)
8
Bantu Orthography Manual
[dz] Options:<>
[dƷ] Options: <> (Swahili inuence) or <> (French inuence)
[ʒ] Options: <> or <>
5. Implosives
[ɓ] Options: <> or <> or <> (or <> or <>)
[ɗ] Options: <> or <> or <> (or <> or <>)
6. Double articulated stops (Labialized velars)
[k͡p] Options: <> or rarely <>
[g͡b] Options: <> or rarely <>
7. Geminate consonants
8. Prenasalization - Prenasalized labiodental fricatives
[
ͫ
f] Options: <> or, less optimally, <>
[
ͫ
v] Options: <> or, less optimally, <>
9. Prenasalization - Prenasalized palatals
[͡tʃ] Options: <>
[͡dʒ] Options: <>
10. Prenasalization: Prenasalized velars
[
n
k] Options: <> or <>
[
n
g] Options: <> or <>
11. Prenasalization: Prenasalized double articulated stops
[
n m
͡kp] Options: <>
[
n m
͡gb] Options: <>
12. Labialization and palatalization - Labialized consonants
[tʷ] Options: <> or <>
13. Labialization and palatalization – Palatalized consonants
[tʲ] Options:<> or <>
14. Labialization and palatalization - Palatal nasal
[ɲ] Options: <> or <> or <>
15. Palatal nasal contrasting with palatalized alveolar nasal [nʲ] and palatalized palatal
nasal [ɲʲ]
[ɲ] Options: <> or <> or <>
9
Chapter 2 - Representing Linguistic Features
[nʲ] Options: <> or <>
[ɲʲ] Options: <> or <>
Labialized and palatalized consonants contrasting with CSV sequences
[kʷe] Options: <> or <>
[kuwe]Options: <> or <> or <>
[pʲa] Options: <> or <>
[pija] Options: <> or <>, <> or <>
16. Velarization: see Velarization
10
Bantu Orthography Manual
Since there are only ve vowel letters in the Roman alphabet, we are left with the option
of altering these currently existing letters to multiply the grapheme options.
Bantu orthographies most commonly have added ~ or ^ over a vowel to distinguish it
from a similar grapheme, or put a line directly through it, as in <>. Underlining and,
rarely, the placing of a dot under a vowel have also been used. These latter two options
make letter/sound recognition slower, but with nine-vowel systems the need for contrastive
representation requires ingenuity. The ability to type the characters is important for litera-
cy, but several Unicode-compatible options are now available.
Bantu phonology is highly sensitive to morphological considerations. Underlying vowel
distribution within specic morphological slots is highly restricted, both in seven-vowel
and ve-vowel languages. Not only are vowels limited as to their distribution within a
word, but there are rules which restrict their occurrence in sequence, especially within
stems (Hyman 2003:46).
3
Typically, there are more contrastive vowels stem-initially than in
the rest of a Bantu word.
The standard ve Roman alphabet letters are used for these systems, representing the
vowels closest to those of the LWC in the perception of mother-tongue speakers.
Swahili’s ve-vowel
system is shown
below (Maddieson:16)
i ninety
ɛ be late
a rst
ɔ hippo
u porridge
i/І u
e/ɛ o/ɔ
a
Two kinds of seven-vowel systems are typical, with contrast for advanced tongue root
(for simplicity’s sake, we are considering +/-ATR distinction as equivalent to tense/lax).
3
For an example, see Figure 1, Tharaka Vowel Coalescence.
11
Chapter 2 - Representing Linguistic Features
ɩ ʋ
e o
ɛ ɔ
ɑ
ɩ u
I ʊ
ɛ ɔ
ɑ
Rangi (F33), Nyakyusa (M31), Nande (DJ42), and Vwanji (G66) are examples of the
upper type seven vowel system. Oliver Stegen (pc) provided the following examples from
Rangi.
ɩ mushroom
ɪ spider
ɛ abdomen
ɑ drizzle
ɔ fallow eld
ʊ mosquito
u strength
A word about the distribution of vowels in seven-vowel systems may be helpful.
Axes often contain a smaller vowel inventory, most typically ve vowels. They may ex-
hibit phonetic variants with the ve phonemes, though, if they harmonize with the vowels
in the roots. The vowels in the augment and prex actually can help prepare the reader to
recognize the root.
Word-nally, even roots of words may have a smaller vowel inventory. There is often
even devoicing of these word-nal vowels.
Bila (D32), like Budu (D332) and some others in northeastern DRC, has a nine-vowel sys-
tem with ATR-based vowel harmony (Kutsch-Lojenga 2003:450-474). A nine-vowel system
will have to distinctively represent the vowels /ɛ/ and /ɔ/.
12
Bantu Orthography Manual
i u
ɪ ʊ
e o
ɛ ɑ ɔ
Vowel harmony is a common phenomenon in which a vowel or vowels harmonize in
some quality with other vowels in a word. It can be seen as the spreading of a phonetic
feature (back, high, round, advanced tongue root) within a word. The spread is usually
leftward from the root, or leftward from a derivational extension. The vowels of one group
tend to occur with each other in words, to the exclusion of vowels of the other group. The
phoneme /a/, however, is usually a member of both groups in Bantu languages, so it is not
aected.
Vowel height harmony is very common, and usually aects only the vowels of derivation-
al suxes such as the causative, applicative, reversive/separative, neuter, and combinations
of these.
ATR harmony, on the other hand,.occurs only in Bantu languages with seven or more
vowels. The vowels are divided into two mutually exclusive groups, one known as +ATR
vowels, and the other as -ATR. As stated earlier, the vowel /a/occurs with both groups. It
often blocks the spreading of the harmony in a given language.
In a language with ATR harmony, all roots (noun roots, verb roots and others) will con-
tain only vowels of one group or the other, but not both. The “dominant” value of ATR
(usually +ATR in narrow Bantu) is the one that causes the other value to change to be like
it. In a system where +ATR is dominant, a -ATR vowel can change to become +ATR, but a
+ATR vowel will never change to become -ATR.
In most Bantu languages, the +ATR feature spreads leftward from a triggering vowel.
This may be a root vowel aecting a prex as in (2), or a sux vowel causing a change in
the vowels to its left. This means that even the vowels in the root can be aected.
Vowel height harmony is shown in (1) with altered vowels in bold, and ATR (advanced
tongue root) harmony as reported for many Bantu languages such as Rangi (F30), in (2)
from Stegen (pc), and Malila (3) from Helen Eaton (pc) and Constance Kutsch-Lojenga. The
Rangi example shows that the –ATR nominal prex is changed to +ATR when it occurs
with a +ATR root.
(1) Swahili:
-imba ‘sing’ + applicative >
-kata ‘cut’ + applicative >
-vuta ‘pull’ + applicative >
-leta ‘bring’ + applicative >
-soma ‘read’ + applicative >
13
Chapter 2 - Representing Linguistic Features
(2) Rangi:
mʊ- work mʊrɪmɔ
snake sp. mʊlalɔ
mouth mʊlɔmɔ
God mʊlʊgʊ
mu- slope mugiritʰɔ
whistle muluri
Kutsch-Lojenga says of Malila (M20), “The [+high] initial vowel of the -ile sux is [+ATR]
and changes all other vowels of the verb word to its left into [+ATR], however many /a/-
vowels are found in between. This means changes in the verb-root vowel from /ɪ/ to /i/
and from /ʊ/ to /u/ as well as the the allophonic realisations [e] and [o] of underlying /ɛ/
and /ɔ/.” The allophonic dierences between e /ɛ and o/ɔ are not written (note that the /a/
is not aected by the harmony process).
(3) Malila
kʉmʉpɨmba ‘to carry him’
amupimbile ‘he has carried him’
kʉmʉbhʉʉzya ‘to tell him’
amubhuziizye ‘he has told him’
Mother-tongue speakers are aware of the changes from one vowel phoneme to another, in
these examples. This means the changes should be represented in the orthography. Some
changes are only allophonic, and these are not written. In both ATR languages exemplied
above, when a +ATR vowel follows underlying –ATR mid vowels /ɛ/ and /ɔ/, these mid
vowels are pronounced as [e] and [o]. These distinctions, only discerned by the phoneti-
cian, are of course not written.
Write word-internal changes which the native speaker is aware of, even if this means a
given morpheme, such as a noun class prex or even a verbal root, may have two written
forms .
4
That would mean that word-internal phonemic (not phonetic) changes due to vowel
harmony rules should be written. This rule will make spelling intuitive for the writer.
The reader is already accustomed to the morphophonologically conditioned alternations in
speech, so will learn to access meaning despite the resulting altered appearance of some
morphemes. The reader is often actually assisted by this surface-level representation of
vowel harmony, because it usually moves leftward from the root. The reader, beginning
at the left end of a word and moving to the right, nds a lot of predictability of the form
the root will take, based upon the rst vowel he/she sees. Indicating the vowel harmony
4
A rare result of vowel harmony which alters a vowel in the root is ambiguity between two roots, but the
context usually clears it up for the reader and the writer is helped immeasureably by writing what he/she
actually hears.
14
Bantu Orthography Manual
changes also means that the reader does not have to read the entire word before he can
pronounce its beginning correctly.
The following section oers suggestions for orthographic representation of the vowels of
a given language.
1. In choosing graphemes for each of the vowels, decision-makers must rst consider
which of them are most closely associated with those of the LWC. These will
probably be given commonly used vowel symbols. The remainder then will
probably require diacritics to distinguish them. Possible diacritic modications
to a vowel symbol are: <←><>, and <>. In writing systems which
represent tone above any syllable, the rst option, <←>, is preferable to
facilitate recognition of both the vowel symbol and its tone. The Tharaka use of a
circumex (<>, <>) above two of its vowels would be inappropriate if tone
were written over any of its vowels (it is not).
2. If the orthographies of surrounding languages use certain diacritics for vowels,
this may aect your choice of symbols, especially if many people have learned to
read using that system.
5
3. For consistency, always use diacritics with the same + or – ATR attributes,
because the contrast will be a natural one for the mother-tongue speaker. For
example, do not use a diacritic on a [-ATR] front vowel, while also using it on a
[+ATR] back vowel.
4. One suggestion, all other things being equal (i.e., never violate rule 3), is to add
the diacritics to the least frequently occurring vowels. This facilitates speed for
typists and writers, since the symbol won’t have to be written as often. It may also
facilitate speed of symbol recognition for readers.
The suggested vowel graphemes are shown below, and they avoid diacritics above the
letter.
V Options
5V orthography: < >
7V orthography: either < > or < >
9V orthography: < >
Some Bantu languages have phonemic vowel length, while others do not.
Hyman (2003:42) lists ve [potential] sources of vowel length in Bantu:
5
This assumes that the number of vowels in these nearby languages is equal to the number in the
language in question. Underdierentiation of vowels for a language should be avoided at all costs!
15
Chapter 2 - Representing Linguistic Features
1. Phonemic length
2. Concatenation (either across morpheme boundaries or through consonant elision)
3. Gliding plus compensatory lengthening
4. Compensatory lengthening preceding a moraic nasal plus consonant
5. Penultimate vowel lengthening
This chapter addresses point 1 under “Phonemically long vowels.” It addresses points 2,
3, 4, and 5 briey, under “Conditioned long vowels.”
Rod Casali identies two more sources of vowel length.
1. elision of one of two adjacent nonidentical vowels, with compensatory
lengthening of the remaining vowel (for an example, see the Tharaka vowel
coalescence chart, gure1
2. phonetic lengthening of underlying short vowels that bear surface contour tone
In regard to this last point, it should also be noted that a long vowel (either phonemi-
cally long or one in which length phonologically conditioned or attributable to morphologi-
cal concatenation) may be longer, or at least perceived as longer, when it is realized with a
contour tone (falling or rising) rather than when it has a level tone.
Some Bantu languages have phonemic vowel length, while others do not.
Proto-Bantu has been reconstructed as having seven vowels /į,
6
i, e, a, o, u, ų/ plus
phonemic length (Hyman 2003:42). Many, but by no means all, Bantu languages have re-
tained phonemic length in their vowel systems. It is easy to determine whether a language
has phonemically long vowels by sorting bisyllabic (CVCV) noun and verb stems using the
participatory approach to language analysis. If there is phonemic vowel length, these stems
will fall into two sets: CVCV and CVVCV. Noun stems containing consonant-glide sequences
(Cy or Cw) or nasal-consonant sequences (NC), e.g., mb, nd, ng, are not used in this initial
sorting for phonemic length. They will be used in the sorting for conditioned vowel length.
If the sorting of the stems with glides and prenasalized consonants also results in two sets,
there is no orthographically signicant conditioning for vowel lengthening, and all long
vowels must be marked as such.
If a language has phonemically long vowels, they must be marked in the orthography
as long. The orthographic representation of phonemic vowel length is doubly important in
languages where the marking of phonemic tone is done sparingly, because the fewer phone-
mic cues the reader has, the harder it will be for the reader to determine the meaning of a
given word.
Most Bantu orthographies denote long vowels by doubling the vowel symbol. The use of
two identical vowel symbols seems to match Bantu speakers’ perception of vowel length:
<>, <>, <>, <>, <>.
6
The vowels į and ų constitute extra-high vowels. The subscript diacritic is a convention peculiar to Proto-
Bantu vowels.
16
Bantu Orthography Manual
People do not generally object to the writing of double vowels once they realize that
vowel length is an important contrast in their language. The use of double vowels will be
unfamiliar even to most literate people, however, and at rst may make it harder for people
to spell their language correctly.
7
Thus, spelling rules must be taught and practiced once
the orthography is introduced. With time, readers and writers will realize the usefulness
of long vowels in distinguishing words whose contrast would be otherwise ignored by the
writing system.
In many of the Bantu languages with contrastive (phonemic) vowel length, one also nds
noncontrastive (phonologically conditioned) vowel length.
Conditioned vowel length is usually found in these environments:
• before a prenasalized consonant
• following a labialized or palatalized consonant (Cy or Cw) or a semivowel (y or w)8
If a language has phonologically conditioned length, as a great number of Bantu languag-
es do,
9
speakers will signicantly lengthen any vowel in the two environments mentioned
above. Speakers will not be able to distinguish a short vowel from a long vowel in these
environments. They will all sound the same length to them. Thus, there will be no vowel
length contrast possible in this environment.
10
Whether the vowels in these conditioning
environments should be written as short or long in the orthography depends on a number
of factors and is best determined by testing the perceptions and preferences of both readers
and writers.
Depending on the language, the actual phonetic duration of such conditioned long vowels
may be nearly the same as a phonemically long vowel, or it may be something in between.
Ganda (JE15) and Sukuma (F21) both have phonemic length and conditioned length.
However, the conditioned long vowels in Ganda are much closer in duration to underly-
ing long vowels in Ganda than they are in Sukuma. Sukuma lengthened vowels are almost
exactly intermediate between underlying long and short vowels (Maddieson 2003:37).
In some languages, perhaps like Sukuma, speakers may feel that the vowels in the
length-conditioning environments group with short vowels, rather than long. If this is the
case, the obvious choice is to use a single vowel to write the phonologically conditioned
long vowels, since it matches the speakers’ perception, and they will automatically ap-
ply the correct degree of phonetic lengthening to vowels which occur in the conditioning
environments.
An example of a language where conditioned vowels are not quite as long as phonemi-
cally long vowels is Tharaka. However, whenever a vowel precedes a prenasalized stop, the
vowel still sounds long to a mother-tongue speaker. Thus, actual phonetic length of these
7
During literacy lectures at the rst orthography workshop, it is useful to give examples from Dutch, Eng-
lish, or other prestigious languages which do manifest double vowels in their writing systems (e.g., English
<>, <>).
8
Sometimes a Cy or Cw will condition following length, but a simple glide (w or y) will not. All environ-
ments should be checked before the linguist decides where vowel length is contrastive and where it is not.
9
According to Hyman (2003:52), “In most Bantu languages there is no vowel length opposition before an
NC complex.”
10
A possible exception would be the result of morphemes rather than phonological processes.
17
Chapter 2 - Representing Linguistic Features
vowels is not necessarily an indicator of how they are perceived by speakers. However,
despite their perception that the conditioned length vowels are long, because this length is
not contrastive, the Tharaka orthography does not write the conditioned length with dou-
ble vowels. Thus, readers must be taught (requiring a lesson on prenasalized stops) to write
such vowels as short.
In languages where speakers feel that the vowels in these environments are clearly long,
writing this length, even though it is not contrastive, may make spelling easier for them.
Mother-tongue speakers may want to write all the vowels which sound long to them as
long. This is one of the options for representing conditioned long (CL) vowels.
But because speakers will automatically produce the conditioned length in the condition-
ing environment, they can often best be written with single vowels, even when they are
perceived as long. If the orthography choice is made to represent CL vowels with a single
letter, writers must be taught to write such vowels as short, despite their perception (requir-
ing a literacy lesson on prenasalized stops).
This is the choice taken by the ocial Ganda (EJ13) orthography. A note in the orthog-
raphy statement or worksheet, with examples, will be sucient to explain the simple envi-
ronments that condition the automatic lengthening.
In Ganda, following semivowels and preceding prenasalized stops, vowels are
always written short (http://www.buganda.com/language.htm).
[tuunda] ‘sell’ is written <>
[kitaanda] ‘bed’ is written <>
[kiimpi] ‘it is short’ is written <>.
There are pros and cons to each option, and these vary from language to language. They
should be discussed by representatives of the language community and choices should be
tested by a literacy specialist and/or a linguist before a nal decision is made.
The disadvantages to writing all long-sounding vowels as long:
• Using double vowels to represent noncontrastive length makes words longer.
• If both contrastive and noncontrastive length are indicated orthographically, truly
contrastive length will not stand out to the reader like it does when double vowels
are used only to indicate contrastive length.
• In some cases, truly contrastive length which results from morpheme concatenation
will be masked by the use of double vowels for the representation of conditioned
vowel length. The following examples can be found under “Other vowel length
issues”.
The disadvantages to writing all vowels with conditioned length as short:
• In languages where tone markings are used frequently, the use of a single vowel
to indicate a vowel with conditioned length can complicate the representation
of falling or rising tone. Such contour tones can be indicated by a sequence of a
high tone mark and a low tone mark for falling tone (or the opposite order for
rising tone) when they occur on a long vowel indicated by two letters. This allows
all tones to be marked using only two symbols. But if a contour tone occurs on a
conditioned long vowel which is represented by a single vowel, it is necessary to
18
Bantu Orthography Manual
introduce a circumex and/or a breve or caron to indicate the contour tone(s).
Discussion of these issues and a good degree of testing the tendencies and perceptions of
both readers and writers will be necessary to determine whether or not conditioned length
should be written with two vowels or one.
In some languages in which there is phonemic vowel length, this length is always
contrastive,
11
even in the environments where length is conditioned in many other lan-
guages. In such a language, either a long vowel or a short(er) vowel may occur in the
conditioned environments, and the long vowels still contrast with short vowels no matter
what the phonological environment is. In such a case, long vowels must always be written
as long, and short ones as short. This is apparently the case in Matuumbi (Odden 2003:532)
as well as in Rangi (F33), where there is contrast preceding a prenasalized consonant in the
following two noun stems:
Rangi: ‘child’ vs. ‘beehive’
We will now look in some detail at various spelling challenges for languages with con-
ditioned lengthening (all highly typical of Bantu languages) and discuss the orthographic
considerations relevant to each.
Sometimes, vowel length (or a series of two consecutive vowels in separate syllables) re-
sults at a morpheme boundary, from the juncture of a vowel-nal morpheme and a vowel-
initial one.
Bungu (F25), which seems to have no phonemic length in stems (Stegen:4), ex-
hibits a contrast in verb tenses that needs to be represented by the use of a double vowel.
This conjoining of two vowels often only happens where the verb stem is vowel-initial, as
in the example here, where the stem is -anda. (See a similar example from Ndali (M301)
later in this section.)
Bungu <> ‘I am beginning’ vs. <‘I have begun’
As Stegen states, long vowels can result from cross-morphemic vowel sequences, at times
involving vowel assimilation in case of nonidentical vowels. This can occur through direct
concatenation as in the following.
Rangi ‘he has shut [implied: the door]’ (from á+ék+ire)
Or it may involve consonant elision as in the following example.
Rangi ‘he has sung for him’ (from á+mú+ímb+ir+ire)
In most cases, such long vowels will be written with double letter, as they can be seman-
tically distinct from the same segmental sequence with short vowel, as demonstrated in the
example below in contrast to the example above.
11
This should not be confused with contrastive length that occurs in these environments only in a few
selected words, occasioned by morpheme concatenation or found in one or two specic grammatical mor-
phemes.
19
Chapter 2 - Representing Linguistic Features
Rangi: ámwiimbíré ‘he should sing for him’ (from á+mú+ímb+ir+é)
However, since Rangi may have contrastive long vowels in all phonological environ-
ments, all their long vowels are written with a double vowel.
Helen Eaton (pc) writes concerning languages in the Mbeya area of Tanzania, “We are
nding that for those languages where lengthened vowels in conditioned environment are
perceived by MT speakers as just as long as underlying long vowels, the morphologically
lengthened vowels also sound just as long, and no longer.” Take Malila (M24), for example.
1 <> ‘to say’
2 <> ‘I have begun’ (from na+anda)
3 <ée> ‘I have taken’ (from na+ega)
4 <i> ‘cl.2-good’ (from abha+inza)
5 <ii> ‘cl.2-black’ (from abha+ilu)
Eaton says, “The underlined vowels are all considered the same length by MT speakers.
If we taught that the morphologically lengthened vowels before NC (2 and 4) needed to be
written long, that would match up with forms like 3 and 5, but dier from forms like 1, in
which the vowel is perceived as just as long.”
As long as there is not any meaning contrast lost by writing all conditioned length vow-
els with a single letter, it is simplest if the rule can be maintained without exception.
Real languages are complex, and even in the extremely regular Bantu languages, there
will be cases which necessitate an exception to your spelling rules. Do not question your
rules just because you nd an exception or two. If you nd extensive and systematic sets of
exceptions, consider revising the rules.
Individual exceptions to the rule of writing conditioned length with a short vowel may
need to be made in cases of verbs with vowel-initial stems. This does not mean that the
“write them short” rule is incorrect or invalid. The exceptions will involve a meaningful
contrast, and this fact may make teaching them fairly concrete. Testing will be needed to
determine the best option for dierentiating otherwise ambiguous words or morphemes.
In Tharaka (E54), vowels are always lengthened about 1½ beats before prenasal-
ized stops. Their orthography follows the practice of writing a conditioned long vowel
with a single letter. This means that they have a spelling rule which says not to write
a double vowel preceding a nasal plus consonant. However, the following words are
contrastive in meaning (quite possibly the second word is actually a clitic plus another
morpheme):<> ‘but’ and <> ‘now’. Since they have dierent meanings, they
need to be written dierently, as vowel length (and tone) in the second word indicates a
20
Bantu Orthography Manual
very distinct lexical item. Thus, they need an exception to their rule, saying that in the
word <> ‘now’ a double vowel must be written, even though it precedes a prenasal-
ized consonant.
Note that if instead the Tharaka writing system simply required that any vowel that
sounds long must be written with a double vowel, both of these words would be written
<>. The contrast in meaning between the two words would be lost.
In Ndali (M301), strict adherence to the rule of writing vowels short in a conditioned
length environment results in neutralized contrasts, such as that between <>‘he
has stamped’ from /a-kanda/ and <> from /a-ka-anda/ ‘and he started’. In such a
case it will probably be best to teach readers and writers to reect the morphemes which
are otherwise involved in the loss of contrast, writing /a-ka-anda/ as <> in order
to maintain the semantic dierence between the two forms. This can be taught in a les-
son focusing on vowel-initial verb stems, which often cause exceptions to length rules, due
to the concatenation of a vowel-nal tense or person morpheme and the vowel-initial verb
stem.
In a case where concatenation results in three vowels in a row, writing them as three
consecutive vowel letters may not be accepted. The following Tharaka (E54) verb in the
distant past exemplies such a situation. The person marker is ba-, the past marker is a-
and the root is -ak-, resulting in three <>’s in a row.
Tharaka: /baaakire/ ‘they built (long ago)’ (from ba+a+ak+ire)
The Tharaka example /baaakire/ is written orthographically as<>, because they
decided to disallow triple sequences of identical vowels in the orthography. Kithinji et al.
give the following explanation.
In some cases it is possible that three identical vowels occur in sequence…
In such cases all three vowels will not be written but only two. For ex-
ample, ba + a + akire, ‘they built’ is written baakire (not baaakire).
This means that in the surface form this will not be distinguished from
the recent past, ba + akire, [also] spelt <>. The two forms are
pronounced with dierent tone patterns, and the context will determine
in most cases which form is intended. (Kithinji et al. 1998:3)
Especially if it can be shown that there is a dierence in the syllable pattern of such a
word, writers may be convinced that writing all three vowels is actually a good option.
In a few Bantu languages, phonemic length has not been retained from Proto-Bantu, but
contrastive length has developed from the loss of intervocalic consonants or morpheme
concatenation. In such cases, there will not be any phonologically conditioned length.
Among these is Tonga (M64) (Maddieson 2003:38).
21
Chapter 2 - Representing Linguistic Features
It is also possible that a contrastive long vowel may be retained in certain tense prexes
even in a language which otherwise has no phonemic vowel length (Nurse 2003:100). This
is especially true of length due to morpheme concatenation or the loss of an intervocalic
consonant, but can also be due to certain retained or regained vocabulary items. Botne
(2003:425) notes that in Lega (D25) long vowels occur most commonly in ideophones
(which generally tend toward onomatopoeia and intonational length) and rarely in other
words—he lists only three other words that have long vowels. Thus, even in a language
with no systematic phonemic vowel length, there can be the need for an occasional double
vowel. This would call for the singling out of such morphemes as teaching points in an or-
thography guide and any literacy pedagogy.
In some languages, there are certain specic vocabulary items which contain sequences
of nasal plus consonant that do not condition length in the preceding vowel. Often these
are not reexes of Proto-Bantu NC sequences, but rather reect the historical loss of a
vowel, usually <u>, from a <mu-> prex. In other cases, they involve a prenasalized con-
tinuant, such as nj, nz, rather than a prenasalized obstruent. The presence of such non-
length-conditioning sequences does not rule out the use of a rule that a vowel be written as
short preceding a prenasalized consonant. Testing should be done, especially with readers,
in such cases.
There are also many cases in Bantu languages in which underlying long vowels undergo
predictable shortening in some contexts:
A number of languages allow a long vowel only in the penultimate or antepenultimate
syllables, and shorten any long vowel which would occur before that. Some languages, like
Safwa (M25) and Kifuliiru (DJ63), count morae rather than syllables and allow no more
than two morae maximum to follow a long vowel. This sort of shortening usually applies
both to conditioned vowel length and to phonemically long vowels.
This may present the need for a spelling rule by which any phonemic length in a verb
stem is preserved in stems which would otherwise lose their contrastive length when sever-
al suxes are added. It is not an issue in nouns and other words which do not have signi-
cant suxation.
The following example (Van Otterloo, pc) shows that from the literacy point of view, it is
sometimes preferable to write underlying length rather than reecting phonological short-
ening in verbs with contrastive length in the stem.
Fuliiru (DJ63) has two verb stems with lexical contrast for vowel
length,<> ‘to put’ and < > ‘to crow’. <> will lose the vowel
length in its radical, phonetically, whenever there are three or more morae following in its
suxes, e.g., with the long causative /–iis/ plus the nal vowel. With the three following
morae, the stem vowel shortens. [-biik-] is shortened to [-bik-] in [kubikiisa] ‘to cause to
put’. If the shortened stem vowel is written with a single letter, the contrast will be neutral-
ized orthographically between the two verbs.
22
Bantu Orthography Manual
The underlying length in the stem should be preserved if testing veries that readers can
learn to identify the verb root, and hence meaning of the whole word, in this way. The verb
radical may be concrete enough that writers would be able to deal with maintaining non-
phonetic length there (Van Otterloo). In this case, then, the verb with causative [kubikiisa]
would be written <>.
The other implication of the presence of a shortening rule is that in an orthography
which represents conditioned vowel length using double vowels, a large percentage of
stems would be aected by any length-preservation rule such as that mentioned in the pre-
vious paragraph. Instead of involving only stems having phonemically long vowels (usually
a relatively small number) it would also theoretically involve any stem in which there is a
phonologically conditioned long vowel.
If a spelling rule preserves the morphemic shape of a verb stem such as [-biik-] above,
this will either necessitate also preserving length in a verb stem such as [-geend-] (in which
the vowel length is merely conditioned, and not contrastive) or else readers and writers will
need to be taught to dierentiate between phonemic and conditioned length when decid-
ing which length to retain. If the decision is made to retain all length, testing will need to
be done to determine whether it is possible for readers and writers to identify the verb root
and to determine that in its unsuxed form it exhibited phonological vowel lengthening.
This extrapolation might be dicult enough that it would outweigh the ease factor of a
decision to write all vowels that sound long with a double letter, and be a factor in deciding
to write conditioned long vowels with a single letter in all instances.
• Languages with phonemic length contrasts write single versus double vowel letters,
e.g., <> vs. <>.
• Long vowels resulting from morpheme concatenation should be written with double
letter, at least in cases where meaningful contrast would otherwise be neutralized.
• Conditioned length vowels which are perceived as short should be written with
single letter.
• Vowels which occur in an environment which conditions length in the language,
but which are perceived as extra long (i.e., there is a tone change across the vowel,
or the length results from morpheme concatenation and is perceived as longer
than that of monomorphemic conditioned length vowels in the same environment)
should generally be written with double letter, subject to testing of readers and
writers. Such words will need to be taught as exceptions to the rule of writing a
single vowel in a phonological environment which conditions vowel length.
Mother-tongue speakers’ perception of a vowel as a consonant in certain contexts will
probably determine your choice of symbols for semivowels. Again, however, certain ambi-
guities can be created when a phonemic (and morphemic) distinction is lost when replacing
a certain vowel with the semivowel. For a ve-vowel language this is rarely a problem.
/i/ is always written as <> before any other vowel besides itself, and /u/ is always
written as <> before any vowel besides itself, as shown below.
12
12
A morpheme is lost, though, when two identical ones adjoin, as can be seen in /mi+i/ <miri> and /
u+u/, <unga> in the Duruma example above.
23
Chapter 2 - Representing Linguistic Features
from Duruma (E72d) nouns, showing what happens at the juncture of noun
class prexes and vowel-initial roots:
class 4 pl: mi-
mi + iri → <> bodies
mi + ezi → <> months
mi + adine → <> trees, sp.
mi + oho → <> res
mi + uho → <> rivers
class 14 sing: u-
u + ira → <> song
u + embe → <> razor blade
u + ari → <> food from maizemeal
u + ongo → <> brain
u + ungs → <> our
In seven-vowel languages, on the other hand, things can be more complicated. See Figure
1 Tharaka Vowel Coalescence, for an example. Both /ʊ/, spelt < >, and /u/, spelt <>
in certain contexts can be pronounced or perceived as consonantal. If a <were writ-
ten in both instances, an important lexical contrast would be lost! The decision was made
to write <> before any vowel besides itself as <>, but to write /u/ before any vowel
besides itself as <>.
Surprisingly (since the higher of the two vowels would seem to function more like a
palatal), the same loss of distinction takes place with glide formation after either /i/ or /î/
within a Tharaka word. In the case of /i/ and /î/, people chose to write /î/, the lax front
unrounded vowel, as <>before any vowel besides itself. The /i/ remains the same in the
writing system, before any vowel besides itself. This is common in Bantu languages, includ-
ing Swahili.
13
As with the prenasalized stops, Tharaka semivowels usually bring about (conditioned)
length—in the vowels immediately following them. The following is an example (Kithinji
1999:5).
14
Watho, ‘law’, will sound as if a long aa sound is present. But we know that
the root of the word is -atha, to which is added the (abstract) noun class
prex û, which changes to <w> before a. Since the underlying form only
has one vowel, only one vowel will be written. This is still pronounced
long.
13
Rod Casali (pc) says, “I have a strong suspicion that the [ATR] distinction…is, in many seven-vowel
languages, not neutralized phonetically either, but that what has commonly been transcribed as [y] is really
a high front vocoid…and that what is transcribed as [w] is really a high round vocoid.” Or, there may be a
“greater desyllabication with the [-ATR] than the [+ATR] vowels,” with the contrast not really neutralized
phonetically. If this is true, the Tharaka orthographic choices made perfect sense!
14
I have altered the formatting of the original article so that the symbols for phonemic, phonetic, and or-
thographic representations are consistent in this manual.
24
Bantu Orthography Manual
The same applies to the other semi-vowel<>. When this occurs before a
single vowel, that vowel will sound as if it is long. For example, <>,
‘his’ is made up of /î + -ake/. It sounds long but there is only one /a/
vowel present in the underlying form.
Here are some other examples.
û+ onokio = <> salvation
û + ega <> good
î+akwa <> mine
î + enu <> yours (pl.)
When two identical vowels follow a semi-vowel, then both vowels will be written. This
is a long syllable and will sound long with or without the presence of the semi-vowel. For
example, ‘you built’ (distant past) is <>. Here the dierent parts making up the
word are /û + a + aka + ire/. Here we see that one <> vowel is the past tense marker,
and the other <> is the rst letter of the word ‘build’. Other examples are the following.
û + a+ anjîîria = <> he started
û + a + athaga = <> he reigns
î + a + amba = <> it spread
î + a + akire = <> it applied itself
Also note that two vowels that join together to form one long vowel will also both be
written after a semivowel, like the following.
wa + ûgu = <> they said
ya + ûgu = <> it said
So basically, the Tharaka spelling rule is: where there is a SV (semivowel) syllable which
only has one vowel in the underlying structure, then only one vowel will be written, but
where there are two vowels in the underlying form, then two vowels will be written. In
pronunciation (the surface form) it is possible that both will sound the same. A trial period,
and eventually testing, may be necessary to determine whether people can grasp series of
rules such as this. The Tharaka adults found their vowel length rules to be very dicult to
master at rst. The children, though, who received instruction in Tharaka grammar as part
of their spelling exercises, may be mastering it.
This is another example of phonological rules impacting the surface representation of
morphemes. In this instance, in order to preserve an important semantic contrast, the mor-
pheme, or underlying form, was represented in the orthography.
If a noun root starts with a vowel, for example Kwaya, E251(J) /eki-aanga/ ‘skull’, pro-
nounced [ekyaanga], there may be an underlyingly long vowel root initially (regardless of
the prenasalized stop which follows it). Mother-tongue speakers are acutely aware of its
presence.
25
Chapter 2 - Representing Linguistic Features
What shows us the underlying length?
The common occurrence of long vowels in this position in the language, even where
there is no possibility of semivowel insertion.
The existence of two morae, shown by a tone change across a long, or double, vowel.
: Write both of the double vowels which follow the vowel in the prex,
if there is an underlyingly long (contrastive with short V-initial roots) vowel in this posi-
tion. The written word will exhibit three vowels in a row, so people may nd it helpful to
attach it to the noun it precedes, with a hyphen. The prex is usually a clitic. Insertion of
a hyphen breaks up the 3-vowel sequence visually, and sets apart what amounts to two
words. This choice may benet the writer, giving more simplicity and consistency in spell-
ing rules for prexes. It preserves the integrity of true vowel length, which is underlying,
not conditioned. The writer doesn’t have to remember the complicated rule, “Never write a
long vowel after y or w, or before prenasalized consonants, except after a prex.”
<> or <>
Never insert a semivowel between two vowels unless speaker perception calls for it.
Otherwise you make recognition of morphemes which are indicated via these symbols,
such as the causative or the passive, harder for the reader, and you make words longer.
If at all possible, avoid using an apostrophe and deleting a vowel in a writing system,
even if the vowel is rarely spoken. Usually, the issue of insertion of an apostrophe arises
when a short word comes before a V-initial word. Since in rapid speech the vowel at the
end of the short word is dropped out, mother-tongue speakers want this vowel dropped in
the writing system. Doing this creates several problems for the reader.
There may be another /t-/ particle in the language which can occupy the same position
in a sentence. In Tharaka, removing the /i/ from the representation of the negative form
of the verb ‘to be’, /ti/, may leave its interpretation ambiguous. Maybe the letter elided was
/a/ instead, giving <> a completely dierent meaning.
15
Elision makes word recognition harder for the reader in another way: it visually unites
two words with very disparate meaning and function, making it harder to recognize the
noun or adjective to which the smaller word is now attached (while normally such words
stand on their own).
Adding an apostrophe after an <> also creates unnecessary challenges for rapid recog-
nition of the commonest symbol for the velar nasal, <>. Since apostrophes are already
in use with this digraph, and a <> is the only intervening contrastive symbol, it is advis-
able to avoid the apostrophe in marking elision.
15
The Tharaka particles /ta/ and /ti/ are examples. The rst of these particles means like, as. The second
is the negated form of /i/, is not. In the orthography it was decided that the rst would always be written as
<ta>, despite its elided pronunciation before a V-initial word. On the other hand, /ti/ is written in full be-
fore consonants, and regularly as < t’ > before vowels. So, though the appearance of the morpheme under-
goes change, at least there is no ambiguity between ti and ta because ta is always fully written out. Writing
the full form of /ta/ was a deliberate measure to maintain its distinction from /ti/ (Kithinji 1999:5).
26
Bantu Orthography Manual
Last, if there are already tone symbols or vowel diacritics written above the writing
line, adding apostrophes will add to the complexity of eye movement and word/grapheme
recognition.
Where elision is a fast-speech phenomenon, there will be considerable pressure
upon the linguist to write “the way people speak.” It may help local people who want to
write “the way we speak,” to see the need to at least preserve the morpheme for one of the
possible elisions. People may agree, for example, to always write out <>, but to elide
the <> in /ti/, as in the Tharaka example above. Don’t expect new literates to have the
nesse we experienced literates have, in distinguishing the correct word from the context!
When there’s a question (and an option), spell out morphemes as they’re pronounced in
slow speech.
On the other hand, if the underlying vowel is lost completely, even in slow speech, do
not force the issue or you’ll make the language very hard for people to write. Choices can
still be made based upon what people know about their language. If a form has been fully
elided and become semantically one, such as <> ‘my child’, it can be writ-
ten as such, with no apostrophe needed. In this case, two words have merged semantically,
forming a compound, a new morpheme which is an amalgamation of the two.
Preserve word image, especially the smaller grammatical words. As Scott Satre (pc) says,
“Americans don’t always pronounce the <> in ‘his/her’ in the middle of a sentence in
English, but we always or almost always write it, depending on the ‘register’ of the com-
munication.” Mother-tongue speakers should not have to rely upon sounding out everything
they read orally in order to get meaning from a text. Even though Americans also say “jago
(e.g., “tʌ thʌ stor)?” in rapid speech, the separate words <> are written as
such.
Hyphenation: When you have a series of words which are semantically closely connected
but not merged, and it’s lengthy for the reader, consider hyphenation, as in sister-in-law.
Because of borrowed words from other languages, some Bantu languages have a contrast
between /l/ and /r/. If this is the case for the language in question, both should be written.
Many Bantu languages have a single phoneme with allophonic variation between [l] and
[ɾ]. It is optimal to write only one, especially if most of the mother-tongue speakers are not
already reading a language which distinguishes the two in writing. If there is one liquid,
<> is usually chosen (especially if it is already in use with a LWC).
[l] Options: <> or <>
[R] Options: <> or <>
: In Zinza E23(J), /ɾ/ does not occur as a phoneme contrasting with /l/, but
is written, in accord with Roman Catholic tradition, for proper names.
27
Chapter 2 - Representing Linguistic Features
In many cases, voiced Bantu fricatives become stops following a nasal so one grapheme
represents the allophonic variants, unless, in rare cases, people are acutely aware of the
distinction. This may happen in places where a LWC such as Kiswahili is very inuential.
Gungu, E101(J), is one language in which plosive b and fricative b are contrastive. The
fricative is written <>, while the plosive is written by doubling it<>.This decision
was made because the fricative b is more common and corresponds to the plosive /b/ in
the orthographies of surrounding languages (Moe 2002:2).
[b] Options: <> or <>
[ß] Options: <> or <> or <>
Test people for their preferences. Some factors to consider are 1) frequency of occurrence;
2)transferability issues (plosive b is written with a single <>in the European languages
and Swahili); and 3) mother-tongue speaker perception, e.g., in many Bantu languages, /b/
is a weak consonant. Pre-nasalization strengthens the voiced bilabial: the fricative [ß] and
the stop [] are allophones in this case.
16
Spelling options become narrower if the language also has an implosive bilabial stop.
Mother-tongue speaker perception of the three sounds (plus their use in the language of
wider communication) will probably determine the choice between the symbols as listed.
Many Bantu languages have a phonemic velar nasal [N] which in Swahili and many oth-
er languages is written with the trigraph <>. In order to avoid the long, visually com-
plex trigraph, and because the symbol exists on typewriters, some have chosen the symbol
<>Another problem with the trigraph <>is created by the apostrophe. Its use in
a digraph may inhibit uent reading because apostrophes will also be used to represent
elision, commonly after an <>! The simple IPA symbol <> is an excellent representa-
tion for this simple phoneme. However, an important consideration in selecting a graph-
eme for the velar nasal is transferability to the nearest LWC (Alas for languages spoken in
Tanzania!).
[ŋ] Options: <> (or <>or <> or <>
17
16
People’s perception of sounds which are allophones for their language may be so acute that the distinc-
tion must be written, in rare cases. Overdierentiation in the writing system may be necessary, if spellers
are consistently confused when they write their mother tongue. This may occur when the inuence of the
LWC is very strong among a language group. An example of this is the Kwaya language, in which velar nasal
fricative hardens to a stop after a nasal. For this language, speakers insisted that its fricative variant be
written distinctly, with a <gh>.” All the related language groups were able to use one symbol for what is
actually one phoneme, /g/.
17
Shona languages of Mozambique (S10) use <n’>, which has the advantage of being a digraph instead of
a trigraph, but the disadvantage of looking like a syllabic n (which Shona does not have (Bill Gardner, pc).
28
Bantu Orthography Manual
:
In Ngbaka (an Ubangian language) /ŋ/ is written as <> and /ŋg/ as <>.
In West African orthographies, a simple<> represents this phoneme. The problem with
reproducibility of this symbol may be diminishing with increased computer use.
In Swahili <> represents the velar nasal, as in <> ‘cow’.
: The biggest challenges for writing the velar nasal arise when it is contrastive with
a prenasalized velar stop. If this is your challenge, see Prenasalization.
Some Swahili noun class prexes (i.e., for classes 9 and 10) consist of homorganic pre-
nasalization (nonsyllabic) preceding voiced consonants. /N-boga/ ‘vegetable’ is written
<>, /N-dege/ ‘bird’ is written<>, /N-jaa/ ‘hunger’ is written <> and
/N-goma/ ‘drum’ is written <>, for exampleSince the homorganic change is word-
internal, the change is written. Voiceless obstruents may undergo the same process.
18
: In Swahili, the velar nasal /ŋ/, a phoneme in the language, is written by a
trigraph <>. However, when a velar consonant is prenasalized, the simple letter <>
is used, to keep words visually shorter and simpler. For example, /Ngoma/ ‘drum’ is written
<> rather than <>.
Usually, prenasalization for Bantu languages is written with for bilabial se-
quences <>,< > and <> precedes the stop for alveolar <> <>, palatal
<>,<> and velar <>, <> sequences. For more on pre-nasalization issues, see
Prenasalization.
:
mbókà village
ndákò house
ŋgómbà hill
Bantu languages with voiceless nasals represent these in their orthography, e.g., in Gogo
<, , , >. An alternative is using a strike-through of the voiceless nasals, as
< ̶> if the orthography is already full of digraphs, or if the consonant sequence /Nh/
also occurs in the language.
[N̥] Options (voiceless nasal): <> or <>
Syllabic nasals are often word-initial in Bantu, usually predicative morphemes or realiza-
tions of the prex . They usually arise from elision of the vowel in a prex or clitic.
18
Another assimilative process is described by Meinhof’s Law, or the Ganda Law. The post-nasal consonant
is nasalized and assimilates to the place of articulation of the nasal, thus creating a geminate nasal. See the
glossary for an example.
19
In nasal-consonant sequences which are heterorganic (having two dierent points of articulation), the
29
Chapter 2 - Representing Linguistic Features
The simplest way to represent syllabic nasals, and to keep them visually distinct from
prenasalized sounds, e.g., /
ͫ
b/ or /ⁿd/, and from other nonsyllabic nasals, is actually to
write the elided vowel in the case of the prex, but this is not often done.
20
: noun classes < > (syllabicity unwritten)
Swahili only has syllabic bilabial nasal: /ṃ/, as in /ṃtu/ <> ‘person’ and /ṃti/
<> ‘tree’. The /ṃ/ is a noun class prex attached to a monosyllabic stem. Monosyllabic
noun stems make the prex syllabic.
For Swahili adjectives, classes 1 and 3 are distinguished from classes 9 and 10 by syl-
labicity vs. homorganicity, e.g., <> /ṃdogo/ ‘little [implied: person]’ vs. <>
/<dogo/ ‘little [implied: object]’. The syllabicity distinction is not written.
21
: Sena (N.40) has minimal pairs for syllabic vs. homorganic nasals, according to
Barbara Heins (pc).
<> ‘clay plate, pot lid’ /
ͫ
bale/ <> ‘brother’ /ṃbale/
<> ‘war’ /
n
kʰo<do/ ‘path’ /ṇkʰo
n
do/
Heins says, “The pronunciation of ph, th, kh, d and b is aected by prenasalization (the
<> indicates aspiration). <>and <> are normally implosive and lose their implosiv-
ity, as in ‘clay plate’ above. In the aspirated stops the consonants are in the process of drop-
ping out as in Shona. <> ‘no’ is pronounced [N] with a prominent [ʰ], as in ‘war’
above.”
[Ṇ] Options: <Ṇ> or <> or <+ elided vowel> or <>
: the apostrophe is probably the most commonly used marker to distinguish
syllabic nasals from others, but apostrophes have their problems, being used for several
other things, such as elision, possessives and as components in trigraphs. Zero marking
may work if minimal pairs, such as those in Swahili, are few.
Due to the inuence of European languages, the voiceless palatal aricate [tʃ] is often
spelled . Orthographers have sometimes opted for the simpler <>.
nasals are always syllabic. Additionally, syllabic nasals with homorganic sequences occur if the nasal is a
class 1 or class 3 prex, e.g., m-bepari ‘capitalist’, m-bembe ‘seducer’, m-beregezaji ‘libuster’, m-bono ‘castor-
oil plant’, while in noun class 9, all homorganic NC-sequences are truly prenasalized, e.g., mbinu ‘method’,
mbolea ‘manure’, ndoo ‘bucket’, and ngamia ‘camel’. Despite minimal pairs such as cl.3 m-buni ‘coee plant’
versus cl.9 mbuni ‘ostrich’ the contrast is not written (Oliver Stegen, pc).
20
For clitic grapheme suggestions, see section on Word Boundary Principles.
21
Karen Van Otterloo (pc): The dierence is that classes 1 and 3 have an underlying m- prex, while 9/10
has an underlying N- (unspecied nasal) prex. The dierence is not audible when the adjective begins with
a bilabial consonant: all four classes would be m-bichi (‘unripe/wet/raw’). They probably wouldn’t go with
class 1, but could go with either 3 or 9/10. The dierence is that the class 9/10 prex is a nasal that always
assimilates to the point of articulation of the C, while the class 1 or 3 prex is m-. Example: m is a prex in
mbwa ‘dog’, and is simply word-medial in kwamba ‘that’.
It might be better to say the contrast is due to word position. An initial NC sequence is usually pronounced
with a syllabic nasal, while a word medial NC is pronounced as a nonsyllabic homorganic nasal.
30
Bantu Orthography Manual
[tʃ] Options: <> (or <> or <> or <>)
When a language has prenasalization which must be written (see prenasalized palatals),
thus adding to the alphabetic (and digraph) inventory, it’s especially helpful to keep the
number of graphemes down to a minimum with the <> choice.
Another reason for using <> rather than <> occurs when there is phonemic as-
piration, usually written with an <>. <> can then represent the aspirated voiceless
palatal aricate, while <> represents the nonaspirated voiceless palatal aricate, e.g., in
Ndau (S15a). Hence, a simple <> is recommended when possible, sociolinguistically.
:
dry beans sʰu sʰururu
island with reeds tʃiwi
These segments are not uncommon in Bantu C, for example see Babole (Leitch 2003), or
in Eastern Bantu. Normally [ts] and [dz] can be written as <>and <> consistently
within a language. In Francophone countries, with languages where [ts] and [tš] are con-
trastive, they are commonly written as <> and <>, as in French ‘Tchad’ for English
‘Chad’. In addition, [ts] and [tš] can be dialectal variants within a language, and in this
case the simpler [ts] <>can be used for both.
[ts] Options: <> or <> (less desirable Francophone)
[dz] Options: <>
: Dibole (C101) has forms like the following (Leitch 2000)
<> ‘open space in river’ /dzlé
́
/ vs. <d> ‘liver’ /di-l/
<> ‘evil tormenting spirit’ /e-dz/
<> ‘red, brown’ /ts/
<> ‘verbosity’ /mo-tsk/
“Usually the aricate is more common than the fricative…so <>is used for the ari-
cate. However, symbolizing the fricative then becomes a problem (Gardner 2005:2).”
[dƷ] Options: <> (Swahili inuence) or <> (French inuence)
[ʒ] Options: <> or <>
<> makes a nice parallel with the voiceless fricative <>. “Other languages choose
to reserve the <> for the fricative and use <> for the aricate (e.g., Sena). However,
this can end up introducing lots of extra s into the written language (Gardner 2005:3).”
31
Chapter 2 - Representing Linguistic Features
Lubwisi, Bantu J, has an implosive b and d. The implosives are written by doubling the
letters: <>and <>Since Shona languages have both implosives and, rarely, ex-
plosives, they use <> and <> for marking the commoner implosives and <> and
<> for the explosives, which are spoken with a bit of a murmur. This choice gives con-
sistency of appearance with their murmured nasals, written <> and <>.
[ɓ] Options: <> or <> or <> (or <> or <>)
[ɗ] Options: <> or <> or <> (or <> or <>)
Using an apostrophe as part of a digraph is a bad idea. Don’t do this unless you abso-
lutely have no choice, i.e., you have a plosive voiced bilabial contrasting with both a frica-
tive voiced bilabial and an implosive voiced bilabial. Even then, I recommend looking for a
fricative symbol for /β/, such as <>.
[k͡p] Options: <> or rarely <>
[g͡b] Options: <> or rarely <>
Since the velar stop has usually been demonstrated to slightly precede the articulation
of the bilabial, the digraph with the velar stop preceding the labial will usually be chosen
over <> or <>, but there are exceptions such as Nyarwanda D61(J) and Shona (S10)
(Maddieson 2003).
Ganda (E15) has a series of geminate consonants that are realized phonetically either by
length or some sort of fortis articulation, according to “A Primer on Speaking and Writing
Luganda.” The Baganda refer to this lengthening of consonants as “stress.”
The Luganda consonants may be either stressed or hit softly. The letter < is a typical
example of this. It is either strongly hit which requires the sound [ as in ‘a thief’,
or it may be hit softly as in < ‘a bad person’. The double consonant in a word is used
to indicate a single strong emphasis on the consonant. This should not be confused with a
repetitive consonant sound. Following are some more examples:
to kill vs. to release
smoke vs. wife of
let them come back vs. bells
The exception is the special consonant <. When this is “stressed” in a word, only
32
Bantu Orthography Manual
the <> part is written twice as in <> rather than <>. One example is
<>.
22
The formation of geminate consonants (a rare phenomenon) may indicate deletion of a
high vowel from between them. Testing of an intelligent but naïve illiterate would be the
best way to ascertain people’s perception (see Gudschinsky 1973:132–133 for details).
We could choose one nasal symbol, for instance <>, to symbolize prenasalization
everywhere. However, existing African orthographies write <>before the prenasalized
bilabials. Even in the case of palatal and velar nasals, a special letter for these consonants
has never been used; rather, <> and <>are written for the palatals, <> and
<> for the velars. Following are listed more problematic phones/phonemes.
There is actually a distinction between the prenasalized <> and the heterorganic
<> in some languages.
:
fold [it] for me
fold [it] for him
One may nd prenasalized labiodental fricatives where a choice is possible between writ-
ing <> and <> versus <> and <>.
[
ͫ
f] Options: <> or, less optimally, <>
[
ͫ
v] Options: <> or, less optimally, <>
Most orthographies represent these as <> and <>, but some have opted for
<>and>If the language also allows a sequence of distinct phonemes represented
by <> and <>, for example, using this digraph will probably complicate recognition of
phonemes in the language. That would be unusual, though.
Virtually all Bantu writing systems represent the voiced and voiceless prenasalized
palatals as <> and <>, rather than something more graphemically complex, such as
<>or <>.
[͡tʃ] Options: <>
[͡dʒ] Options: <>
from Tharaka:
͡tʃaana child
22
Available on the internet: A Primer on Speaking and Writing Luganda (http://www.buganda.com/Lu-
ganda.htm#elem)
33
Chapter 2 - Representing Linguistic Features
͡dʒoka snake
According to Ron Moe (2002:8), “All existing Bantu orthographies write the prenasalized
velars as <> and <>, rather than <> and <> or <>and <>. The
‘correct’ pronunciation is automatic on the part of the reader, of course, conditioned by the
stop’s point of articulation. As always, creating a trigraph with insertion of an apostrophe is
not optimal.”
[
n
k] Options: <> or <>
[
n
g] Options: <> or <>
Of course the only question is which nasal symbol to use, or , at the front of a trigraph.
Usually is chosen to represent the nasal.
[
n m
͡kp] Options: <>
[
n m
͡gb] Options: <>
[tʷ] Options: <> or <>
The orthographic challenge is choosing the spelling of a morpheme, [Cu] when another
vowel follows the [u]. Syllable rules often determine the writing rule, converting the labial
feature to <.
from Tharaka:
mû + ambais writtenthief
ba + kû + andais writtenplanters
[tʲ] Options: <ti> or <ty>
As with labialization, the choice of graphemes depends upon the structural (e.g., syl-
lable structure) rules of the language and possible morpheme preservation issues. The same
principle is in operation as the one for labialization, but the necessity of an underlying /u/
converting to a consonant /w/ when the underlying rst vowel is high and rounded doesn’t
seem to be as strong for /i/ converting to a consonant when the rst underlying vowel is
high unrounded.
: The following chart of vowel sequences for Tharaka
(Kithinji 1999:3-4), depict-
ing the vowels in orthographic (not phonetic) form, illustrates this.
34
Bantu Orthography Manual
The vowels listed down the left side are those which come rst and those along the top
are those which come second.
It is important to note that the vowel <> only changes to <> before <>
when <> stands alone as a V syllable. Because of the language’s phonological rules (syl-
lable structure), when /I/ follows a consonant, as in <> or <>, it continues to be
written as <>and does not change to <>.
/rû + a/ is commonly pronounced as one syllable (CSV), and is therefore spelled
<>.
/rî + a/ is commonly pronounced as two syllables (CV-V), and is therefore spelled
<>.
Most Bantu languages have a palatal nasal [ɲ] which is normally written with the di-
graph <>. The <> option corresponds well with Swahili and most, if not all, other
Bantu orthographies. This does, however, conict with the use of <>to represent a
consonant followed by a palatal. This is not a serious problem unless the palatal nasal [ɲ]
is contrastive with palatalized alveolar nasal [nʲ](See examples below).
[ɲ] Options: <> or <> or <>
Many Bantu languages do make a distinction between /ɲ/ and /nʲ/, especially on the
western side of Lake Victoria. Most eastern African languages write the palatal nasal
<> as above, reserving <> for the palatalized nasal.
23
:
Zinza E23(J) in Tanzania with <> for palatal and <> for palatalized alveolar
nasal.
Gungu E101(J) and Gwere E17(J) in Uganda use <> for palatal and <> for palatal-
ized alveolar nasal (Oliver Stegen, pc).
23
The palatalized nasal is a phonetically complex phoneme. For more on this subject, see “Palatals” and
“Palatalized nasals,” The palatal nasal, on the other hand, is a single, simply articulated consonant.
35
Chapter 2 - Representing Linguistic Features
In some Bantu languages there is an emic contrast between the palatal nasal [ɲ] and the
palatalized n [nʲ]. As mentioned above, Gungu has a three-way contrast between the pala-
tal nasal [ɲ](<>‘urinate’), the palatalized nasal [nʲ] (<>‘heal’),and
the palatalized palatal nasal [ɲʲ] (<> ‘cause to increase’ [Moe 2003:5]).
[ɲ] Options: <> or <> or <>
[nʲ] Options: <> or <>
[ɲʲ] Options: <> or <>
24
There are problems with all of these options. Most orthographies use <> for the pala-
tal nasal [ɲ], but this choice conicts with representation of the palatalized alveolar nasal
[nʲ]. If palatalization is normally transcribed as <, then<>will provide a consis-
tent representation of that phoneme. But if both the palatal nasal and the palatalized alveo-
lar nasal are contrastive emically and written identically, orthographic ambiguity will be
the result for users of the writing system. The palatal nasal [ɲ] is by far the most common
of the two. The palatalized alveolar nasal [nʲ] is most often written <>.
Sociolinguistically, mother-tongue speakers of Bantu languages have proven unwilling
to write the distinction. The need for distinguishing the two may be compared to the need
for distinguishing [ð] and [θ] in English—minimal pairs for them are rare indeed, with
the voiced interdental fricative occurring in functors and the voiceless in content words, so
they are both spelled <>.
Test with connected text for comprehensibility and uency of reading, as well as people’s
perception of the options.
Some languages permit a contrast between <> (i.e., a labialized consonant such as /
kw/) and <>. The Gungu words <>‘yield’ and <>‘tell’ are illustra-
tive. The same contrast can occur between <> and <> (i.e., a palatalized conso-
nant such as /py/ and a sequence such as /piy/, and between <Cj> and <Cij>. The Gungu
words <> ‘rhinoceros’ and <>a loanword for‘money’, demonstrate the same
contrast. Other similar sequences are possible, of course.
[kʷe] Options: <> or <>
[kuwe] Options: <> or <> or <>
[pʲa] Options: <> or <>
[pija] Options: <> or <>, <> or <>
Regarding the last option listed, it may be inappropriate to use<>to represent the
palatalized voiceless bilabial stop, since it may already be used to represent a palatal stop,
not palatalization. Choice of symbols depends upon whether the <> symbol is already
24 “
The solution of writing a vowel here, at least for Lugungu, is a good one because the palatalization of
the nasal is probably the result of a high [+ATR] vowel following the nasal” (Rod Casali, pc). If a high vowel
is the cause of such a phenemonon, the vowel should be written.
36
Bantu Orthography Manual
in use, or whether, alternatively, <> can have a dual role, representing both a voiced
palatal stop/j] and palatalization of a voiceless stop <>.
In addition to murmured consonants, all Shona (S10) languages have a set of ‘labiovelar-
ized’ consonants (Gowlett 2003:614). According to Gowlett, “[the] w element carries both
labialization in the form of lip-rounding, and velarization in the raising of the velum.”
With such an extensive and complex consonant inventory, digraphs and trigraphs are
unavoidable.
from Shona:
“In addition to the basic phonemes, there are three major modications: Prenasalization,
Velarization, and Palatalization. Voiced oral stops and fricatives can be prenasalized
(
m
b,
n
d,
N
g,
M
v,
n
z,
n
ʐ,
ɲ
j), represented by a preceding <> or <> in the orthography.
Velarization, rather than labialization, is represented by a following <> (e.g., <>,
<>), and has several interesting phonetic realizations, as does Palatalization, represent-
ed by a following <> in <>, <> and <>” (Gardner 2001:59).
Shona, like many Mozambican languages, has an orthography that avoids the use of
special characters. Besides the modications mentioned above, the consonantal symbols
in Table 1 are used (Gardner 2001:59). The large consonantal inventory requires many
digraphs. Use of IPA symbols might be a better option for future orthographies, but LWC
transfer issues are the challenge.
:
: ɓ b ɗ d ʋ v ʂ ʐ ʃ ʒ tʃ ɲ ŋ ɦ
It’s hard to keep this section distinct from the segmental items above it since most Bantu
orthography issues have a morphophonological aspect! Donald Burquest (1993:172) says
that there may be no dierences between the representations of a structural phonologist, a
generative phonologist, and a lexical phonologist.
This will not be the case for the Bantu languages. Morphology and phonology will impact
one another repeatedly.
25
In the words of Keith Snider (2005:1),
The ‘trick’ in developing a good orthography is to write words the way
the mother-tongue speaker perceives a word to sound, not necessarily the
way the mother-tongue speaker actually pronounces it. This is great for
the immature reader. There is an added bonus, however, and that is that
when we write this way, we are also able to maintain a constant word-
image. Writing the way mother-tongue speakers perceive the language
25
This is a very concise description of the issues. For more information, see Edmondson, et al. 1998 or
Snider 2005. Snider’s entire article can be found in Appendix D (Part 1).
37
Chapter 2 - Representing Linguistic Features
to sound meets the needs of both mature and immature readers with the
same orthography.
At other times, however, the mother-tongue speaker is well aware of a dierence in pro-
nunciation of a morpheme. For examples of this in English, see You can have your cake and
eat it, too: An orthography that meets the needs of both mature and immature readers, by Keith
Snider (2005:3). in Appendix D (Part 1). The following are Bantu examples of morphopho-
nological challenges and possible solutions.
Ron Moe, referring to hardening of voiced fricatives or liquids after a nasal, writes, “It
would be possible to keep the underlying form of the root even when a nasal prex is at-
tached. For instance the /in <(root:-rug-) ‘to come from’ could be maintained
by writing <which phonemically is [nduga] ‘I come from’. This would maintain
a consistent morpheme image. However, most orthographies do not attempt to maintain a
consistent morpheme image (certainly not in East Africa!). Some people have confused the
principle of maintaining a consistent word image with maintaining a consistent morpheme
image. Lexical phonology helps us to understand why the two are dierent… The results
of lexical rules (such as the Gungu rule that changes the phoneme /rinto the phoneme
/dwhen a morpheme ending in is prexed) should be reected in their spelling (Moe
2002:5). The result of such a lexical rule for Lugungu, E101(J), is <>.
If allomorphs are morphologically conditioned they probably should be written reecting
their pronunciation, as above; if the variation is simply phonologically conditioned within
the shape of the word, then the morpheme should be preserved in the spelling. In other
words, show a change when morphemes come together. Don’t show it when it’s just the
result of a phonological process such as compensatory lengthening.
Function words, especially short ones, are best written consistently, despite variation
in their pronunciation. There is great advantage to having their form constant. Though
they are often monosyllabic and therefore phonologically attached to a noun or verb, they
should generally be written disjunctively
26
(see section on word breaks below).
The following example from the Lugwere writers’ guide (Moe 2003:4)
27
illustrates phono-
logical conditioning of a (word-internal) morpheme:
:
The prex <-> ‘Class 7’ becomes <-> when it is followed by a vowel. However, the
<> is pronounced the same as <>. Both are pronounced [ʧ]. If a [ʧ] sound is from an
underlying [ki-], it shall be spelled <>. Otherwise it shall be spelled <>.
This option benets uent readers and uency in general. The morphophonological alter-
nation has eects beyond the word level, as can be readily seen by the matching prexes
in Option 1. The redundancy built into the spoken language is reected in its written form,
making words recognizable more quickly (with the letter <> in the prexes of both
words, as in <>).
28
26
Literacy Note: they will need to be taught as sight words (memorized and recognized as whole words),
but with a larger context whenever possible.
27
The formatting of the original has been altered so that the symbols for phonemic, phonetic, and ortho-
graphic representations (< and >) are consistent in this manual.
28
The <k> in the prex of the rst word tells the reader that the possessive pronoun following the noun
38
Bantu Orthography Manual
<> ‘my thing’
<> ‘their secret’
<> ‘that trap of his’
<> ‘our thing’
<> ‘at the village’
<> ‘that secret’
<> ‘to giggle’
<> ‘I stray’
<> (Lusoga) ‘he squats’
:
The prex <-> ‘Class 7’ becomes <-> when it is followed by a vowel. However the
<> is pronounced the same as <>. Both are pronounced [ʧ]. Whenever a word has
the [ʧ] sound, it shall be spelled<>, no matter whether it is from an underlying <->,
or from an underlying <>.
<> ‘my thing’
<> ‘their secret
<> ‘that trap is his’
<> ‘our thing’
Option 2 has the advantage of one symbol, one sound, if after testing with careful, slow
speech you nd that people perceive the sounds to be identical ([ʧ] in both contexts).
<>, as in Swahili <> ‘room’. Swahili may long ago have lost the underlying
morpheme (noun prex ) with the substitution of [tʃ] before some vowels in class 7, but
is true to its perception by the mother-tongue speaker in its orthography, with a phonemic
distinction between the variants.
29
The prex is aricated at times, but not consistently:
‘shoe’potato’mirrorfood’
paper’toilet’
will of course begin with the same letter.
29
This is described as a velar palatalization process in Nurse 2003:55.
39
Chapter 2 - Representing Linguistic Features
Follow these steps to choose spellings for morphemes when phonological processes alter
their pronunciation:
1. If the morphophonological process in question takes place across word
boundaries, retain a constant visual image of the word. Associative markers, for
example, are usually subject to phrase-level phonological changes rather than
word-internal change, on the underlying vowel /a/, preceding vowel-initial
nouns. Such grammatical words should have their integrity maintained: write
morphophonemically across word boundaries (remember the “slow speech”
principle).
2. Determine whether morphophonological processes take place between a root
morpheme and its axes.
3. If so, determine the following:
4. Do the changes aect the root morpheme?
5. Do they aect the ax?
6. Do they aect both root and ax?
7. If they do, determine whether the sounds which result from morphophonological
processes are phonemes in their own right (phonemes which already exist in the
language, and for which there is a symbol available). Try writing content words
(nouns, verb roots, words with clear meanings) phonemically and grammatical
particles morphophonemically.
Some vowels, originally distinct, change when they adjoin one another. Since these
modications are word-internal and involve changes into sounds which do exist elsewhere
as phonemes in the language, reecting the phonological changes in writing is usually best.
Some morphemic contrasts are lost in this process, but beginning readers really benet,
especially for languages which have many vowels in the verbs. The vowel which is assimi-
lated doesn’t completely lose its inuence; its inuence upon syllable length often gives
evidence of its presence in speech.
Figure 1 (see the chart of Tharaka vowel coalescence, or coalescent assimilation) illus-
trates the phonological processes which take place when dierent vowels adjoin one anoth-
er and the orthography choices reect them quite closely.
1. In general, spell words the way they sound after word rules have applied, but
before phrase rules have applied (Snider 2001).
2. Determine whether the resulting sounds of the morphophonological process are
sounds which may occur elsewhere in the language. If the sounds occur nowhere
else, then they are conditioned allophones which should not have independent
representation in the alphabet (Van Dyken, et al. 1993).
30
30
For more on this subject, see “How to make spelling choices” LinguaLinks version 5. [CD-ROM]. Dal-
las: SIL International, 2002. n. pag. Available: LinguaLinks Library bookshelf (see www.ethnologue.com/
LL_docs/lit_bkshlf.asp).
40
Bantu Orthography Manual
Tone will be usually be underspecied in the orthographies of Bantu languages. You
want to represent tone as simply as possible, while striving to maintain a constant word
image to aid readers in morpheme recognition. Your main task here is to determine what
must be disambiguated, i.e., to discover any contrasts which are conveyed only by tone.
Here are some generalizations.
1. Expect not to mark surface tone. Expect to mark something more abstract, on the
word level, if you mark tone. If you mark tone on a surface level, the reader will
be forced to sound out every melody (i.e., sequence of tones on each syllable)
every time he encounters a new word, in order to gain meaning.
2. It is the functional load (relative importance of a linguistic feature in conveying
grammatical or lexical information) of lexical tone which determines whether
or not it must be marked.
31
Is tone the only indicator of a grammatical or lexical
distinction, or is there another indicator as well, within a word or phrase? If
you consistently nd dierent tone melodies across words of the same syllable
pattern and same part of speech, it’s a strong indication that lexical tone should
be written. However, even in such a case, as few as one or two tone melodies may
actually require marking (See the following Rangi example in “Nouns,”).
3. “If there are more than 100 minimal pairs for tone in a given language, then
writing of lexical tone should be considered. This is very unlikely for Narrow
Bantu Languages. If one decides to write tone in a language, it is absolutely
necessary to come up with a good way of teaching it”, according to Kutsch-
Lojenga (pc). If lexical tone must be marked
32
, the most frequently occurring
lexical tone is considered the default; hence it is not marked. For example, the
root of a verb in a two-tone system could be marked by nothing if it is high and by
a symbol over the appropriate stem syllable/s if it is low.
4. If both grammatical and lexical tone are found to carry a signicant functional
load, consider a double system of orthographic representation, representing
both grammatical and lexical tone. In this way, the needs of both beginning and
mature readers can be met. That is, you ensure consistency of writing, plus more
immediate recognition of meaning.
Steven Bird (1998:23) provides the following example from Kako (A93), a Bantu
language from Cameroon. The injunctive form of the verb is indicated by a single
diacritic at the end of the verb, while the tonal pattern of the word (conveying a
lexical distinction) is indicated by a single diacritic above the rst vowel of the
verb stem.
31
See “Functional Load” in the glossary, for a Lendu (Central Sudanic, nonBantu) example, in which tone
has a very HIGH functional load. See also Appendix A, Determining Functional Load for Tone, for one proce-
dure for calculating functional load.
32 “
In languages that have a tendency to be monosyllabic, there are numerous instances of tone distin-
guishing between lexical items or grammatical constructions. Bantu languages are [usually] on the other
end of the scale. They tend to have long words, and tone is less frequently the only distinguishing feature
between nouns especially. Tone plays a role which is much greater in the verbal system…there are not many
instances where tone is the only feature distinguishing between two otherwise identical words” (Kutsch-
Lojenga. 1986).
41
Chapter 2 - Representing Linguistic Features
he does-INJ circumex=falling tone
he listens-INJ circumex=high-low
he works-INJ circumex=high-low-low
he sings-INJ circumex=high-low-low-low
Such a meaning-based representation of tone seems quite appropriate in a
language where an entire melody can be compactly represented by one symbol
per tone melody rather than one symbol per tone, or where tonally distinct words
appear the same in isolation, or where there are morphemes manifested solely
by perturbation of the tones of nearby words. In both situations, we may want to
adopt diacritic symbols which are iconic representations of an underlying tone
pattern (Bird 1998:23).
5. Along the same lines, for languages in which lexical tone need not be marked, but
grammatical tone does, there are two possibilities. The rst marks the tense at
the beginning or end of a word. The second marks it directly over the contrastive
morpheme.
Mark a verbal morphemic distinction with a symbol at the front of the verb, such
as this. The value of this marking system is that the reader/writer simply knows
that this tense, i.e., distant past, is indicated by that symbol. No awareness of the
locus of an ax or a syllable to which tone is attached is necessary.
In this example, Kwaya speakers chose to distinguish between present perfect
completive tense and the distant past tense (2 days ago and before). Therefore,
the distant past tense will be marked with a word-initial colon.
<naateekere> ‘I have already cooked’
:naateekere> ‘I cooked (two days ago)’
Snider suggests positioning the diacritic in some consistent position directly over
the morpheme which would otherwise be ambiguous.
Diacritic marking of grammatical tone can be successfully employed when
tone does not play a great role in distinguishing lexical items, but there
are nevertheless a signicant number of grammatical distinctions indi-
cated by tonal dierences. In this case, one does not mark tone per se.
Instead, one uses diacritics in a consistent manner to indicate the dierent
grammatical distinctions. An acute accent on the verb, for instance, might
be used to signal perfective aspect, regardless of how tone is realized in
this construction. Similarly, a circumex might indicate imperfective as-
pect (Snider 1992:7).
42
Bantu Orthography Manual
6. The best way to discover what additional grammatical tonal phenomena must be
marked is for the local speakers to start using the edgling orthographyand keep
track of any cases of sentence-level ambiguity (i.e., a sentence may be written
according to orthographic rules but still have two dierent interpretations because
of unspecied grammatical tone phenomena).
7. Floating tones: If oating tones are contrastive, represent words that have oating
tones dierently from words that do not have them. In other words, mark the
dierence on the word that causes the dierence, as opposed to the adjacent word
that undergoes the change.
33
If we need to represent each tone melody in a unique
fashion anyway, and since the oating tone of a particular word is part of its
inherent melody, it only stands to reason that the word that initiates the oating
tone should be the one that gets the special treatment. (For more examples and
explanation, see Snider 1998. SIL Africa Area has a copy, on CD.)
8. Determine the functional load of the verb tone contrasts (this is after contrastive
melodies have been identied during phonological analysis). Many Bantu
languages will have two or occasionally three tone classes (underlying pitch
patterns) for verb roots.
34
In others, there is no longer a lexical tonal contrast.
If there is more than one tonal melody for verb roots, test readers for the need
to mark tone on the verb root. Minimal pairs are just a clue to possible heavy
functional load, not the only indicator. Give the verbs a context when testing.
9. Verbal extensions: Verbal extensions not only have a reduced vowel system—
they usually also lack tonal distinctiveness.
35
Don’t expect to write any lexical
tone on verbal extensions because they don’t normally have underlying tone.
A grammatical tonal pattern, however, will be realized across the whole verb,
including any extensions.
10. There will probably be some Bantu languages, particularly in the west, which
require marking of verbal tone both lexically and grammatically. See point 4
above for details on using a dual marking system which indicates both lexical and
grammatical tone. Determination of the necessity for such dual markings may
require testing.
I recommend using the two distinct kinds of marking whenever both lexical and
grammatical tone must be written: marking lexical classes of verb or noun roots
tonally by the use of a diacritic over a vowel (if tone class marking carries a high
functional load for that language) and marking the grammatical tonal melody of
a specic grammatical verb form with an abstract symbol, such as an exclamation
33
A oating tone (generally a L tone) causes a following H tone to be realized at a lower pitch than it
would be if the oating L tone were not there.
34
Proto-Bantu had two lexical tone classes for verbs, and they still exist in the great majority of Bantu
languages (Maddieson 2003:40). Narrow Bantu languages often have a third exceptional class, referred to
as toneless verbs. This class of toneless verbs adopts its surface tones from its surroundings or from adjacent
tonal morphemes. On the surface a toneless verb may appear to function at times like a high tone verb and
at other times like a low tone verb. This is most often the case with very common, and hence often irregular,
verbs.
35
Schadeberg (2003:148) says, “In the verbal word, they generally adopt the (rst) tone of the nal sux
(‘tonal extension assimilation’ by anticipation).”
43
Chapter 2 - Representing Linguistic Features
mark, colon, hash, or other mark. The use of such a meaning-based symbol at the
front of the verb alerts the reader to the correct tonal melody of the word. Such a
symbol will not be taught as a mark of pitch/tone, but as a grammatical marker
(i.e., # equals recent past tense), and should be recognized like a sight word. This
amounts to morphemic writing of grammatical tone (Bird 1998:16).
In languages with shorter words, distinctive markings directly over the vowel
may still be the easiest. Many Grasselds Bantu languages use diacritics for
grammatical tone. According to Anderson and Alomotor (2005), Awing requires
the marking of habitual aspect. Diacritics are used for this purpose, with two dots
over he vowel <ӓ>. The rule is that this diacritic be placed on the rst vowel
preceding the verb in a sentence. Note the contrast in the following example
sentences.
11. Try to nd a system to distinguish tone for nouns if lexical tone proves to make
a distinction for several minimal tone pairs (e.g., 5 percent or more of words in
the lexicon). Use the pitch pattern of the noun in isolation, unless for some reason
underlying pitch patterns are neutralized in isolation. Expect to nd at least two
main pitch patterns for nouns; many languages will have more, such as Bila in
Eastern DRC and Tharaka in Kenya.
36
12. Marking tone in functors (small grammatical words such as to and too ) can
be very important, if sentence order does not already tell the reader which
grammatical word they are reading. Most important is to distinguish functor
minimal tone pairs that can be confused in relatively short sentences.
Word breaks have always been a challenge in writing Bantu languages, partly because of
the extensive axation of monosyllabic morphemes.
We can assist uent word recognition by matching the intuitive knowledge of mother-
tongue speakers. The following guidelines should help you do this (Van Dyken, et al.
1993:7). Examples added are mine.
1. The principle of : A morpheme qualies as a word if it
clearly communicates meaning, even when heard or seen in isolation. Conversely,
each language has certain morphemes which cannot communicate meaning in
isolation, such as Bantu noun class markers.
2. The principle of : This requires that each written word convey
only one concept. When compound words are formed, it is because the two words
join to form a new and unique concept.
36
Proto-Bantu had four distinctive tone patterns for disyllabic nouns: LL, LH, HL, and HH, which might be
expected in most Bantu languages; even reduced tone systems with only two or three lexical tone distinc-
tions on nouns might still require tone-marking. Tharaka has an extra-high tone.
44
Bantu Orthography Manual
3. The principle of : Word space can show that two words
are pronounced with dierent stress patterns (e.g., blackbird and black bird are
distinguished orally by dierent stress patterns). Ignoring these dierences can
lead to ambiguities in the written text which are absent in speech. If both blackbird
and black bird were written with a space between the two component words, the
dierence in the two meanings would be lost.
: In Tharaka, the proclitic /wa/ ‘each’ has an entirely dierent function
(and stress pattern) from the /wa/ written separately, meaning possession, so the
writing system reects the distinction.
4. When a morpheme demonstrates , it is usually written as a separate
word, even if it doesn’t make sense in isolation. In Swahili,
Or are both correct (Translation:
When will your guests arrive?). Since the adverb clearly is not obliged to follow
the verb and its axes in this example (), but is movable (it is isolatable
as well), it should be written as a separate word from the verb, for readability
considerations.
37
5. The principle of A functor is written separately from the word with
which it usually occurs when other words can intervene between them. In Bantu
languages the subject prex, negative marker, and tense and aspect markers tend
to be bound axes; because these markers are so closely linked to the verb stem
there is no possibility of inserting any other words between them and the verb
stem.
: The associative marker, usually consisting of a prex plus –a, e.g., the
/ya/ in Swahili , ‘the house of my younger relative’,
should be written either as a clitic or as a completely separate word from the
following noun. Other words may intervene between the associative and the noun,
as in this example: ‘the house of this younger
relative of mine’. The associative should not be attached to a noun as a prex,
because it is a separate word. Since it is monosyllabic, however, people may
prefer to treat it as a clitic. (See following section on clitics.)
However, in some verb forms, the prexes may be bound to an auxiliary, while
the main verb stem constitutes a separate unit. If it is possible to insert other
words between the associative marker and the verb stem, it is quite possible that
there is a verbal auxiliary, and thus there is a word break in the verb. In this case,
people will respond favorably to writing the verb stem as a separate unit. The
subject marker will probably be perceived as a prex on the word which precedes
the main verb stem within the verb phrase. Phonological clues such as vowel
harmony, vowel length, stress, and tone may help guide the word break decision.
If not, test speakers with a variety of treatments of word breaks in such verbs.
37
Based upon its mobility, a poor word break choice for Swahili’s interrogative particle /je/ is shown here.
It can occur in two locations in a sentence: <Je, wamefanya nini?> ‘What? What have you been doing?’ and
<wamefanyaje?>’You have been doing what?’ It should be written disjunctively in both instances, though
the stress patterns of speech have supported its conjunctive spelling.
45
Chapter 2 - Representing Linguistic Features
Tone can also make a dierence in grammatical meaning, and of course that
written distinction must be attached to the part of speech it modies.
6. The principle of is often a criterion for writing certain
grammatical elements as separate words. In Bantu languages, one can substitute
‘the boys’ for the independent pronoun ‘ they’ in <> ‘they played’
or <> ‘the boys played’ but not for the same prex in the
verb. One cannot say, for example, */wavulanalicheza/. One must say /wavulana
walicheza/. Hence the /wa-/ morpheme is written as part of the word.
As stated earlier, there may be much elision in rapid speech. When at all possible, choose
the full form with the dierent part of speech as a separate word, striving for the constant
word image. Good readers always recognize units which are larger than the individual
grapheme; they quickly recognize and retain syllables and morphemic units. Remember (at
the risk of overstating it), spell words the way they sound after word rules have applied,
but before phrase rules have applied (Snider 2001:324).
A linguistic consultant will take advantage of phonological indicators such as tone pat-
terns which can only occur at certain places within a word, vowel length indicators, and
vowel coalescence indicators, in nding word breaks. Coalescence of vowels within a word
results in a dierent vowel quality than coalescence of those same vowels across a word
boundary, following phrase-level rules. These patterns, phrase-level and word level rules,
can really help one distinguish word breaks, and mother-tongue speakers will probably
quickly identify them. All of these indicia should be used to inform orthography develop-
ment. The following issues are thornier and very common.
One major cause of confusion plaguing the issue of word division in Bantu stems from
the failure to distinguish between axation and cliticization. The failure to distinguish be-
tween monosyllabic axes, which are an integral part of a word, and monosyllabic words
which are separate words, phonologically cliticized to another whole word, is a major cause
of the confusion regarding word boundaries in Bantu.
Clitics are phonologically joined to a host word and yet are not axes. Whether they
need to be represented as separate words or as words attached, such as by hyphen or apos-
trophe, to another word depends on whether or not this joining of the clitic to the host
word causes phonological changes such as tone or stress changes, or changes in vowel
length or quality. In order to preserve the word shape both of the clitic and the host word,
it is not recommended to represent a clitic in the same way that one represents an ax.
Clitics are a common occurrence in Bantu nominal and verbal phrases. Mother-tongue
speakers, especially those with less exposure to print, will often want to see them written
conjunctively with the word on which they depend because they are phonologically con-
nected. A writing system must do more than represent sounds. It represents the function
and meaning of sounds and morphemes to the reader. The big challenge with clitics is rep-
resenting their separateness, yet still acknowledging their phonological attachment to the
word. The big question to ask is: “Isit an ax, or is it a clitic?”
This appears to be a simple question, yet depending on the morpheme involved, it may or
may not be that simple! The six principles found above under “Word Boundary Principles”
46
Bantu Orthography Manual
can help you determine if a morpheme is a clitic or an ax. Once the question is answered,
these steps are suggested by Karen Van Otterloo (pc):
If it is an ax, it should be written as part of the word.
If it is a clitic, the OWL should decide whether there are any phonologi-
cal consequences for the host word. Is there a change in stress? Does
vowel lengthening result? Are there tonal changes in the host word? If the
answer to any of these three questions is “yes,” it will probably be best to
indicate orthographically the clitic’s dependency on the host word (e.g.,
with hyphenation).
If the answer is “no”—there are no such consequences of cliticization
upon host words in the language—then a literacy consultant should deter-
mine the mother-tongue speaker reaction to writing the clitic separately
or together. If mother-tongue speakers choose writing the clitic conjunc-
tively, the next decision whether to use a hyphen to mark the joining.
A useful generalization is that locatives attached to nouns will usually be clitics.
Corresponding locative concords, used with verbs, pronouns, adjectives, and demonstra-
tives, will be axes.
As long as this dierence between the nominal and nonnominal elements matches the
perception of mother-tongue speakers, there will be no problem in getting them to write
the locative plus noun as a clitic (or separate word) plus a noun, while writing the concord
prex plus verb as a single word.
As we’ve said, Bantu languages have strong prohibitions against monosyllabic words,
thus pressuring such clitics as the Tharaka /-ni/ ‘in, on’ to be written as part of the nouns
they follow. Try a hyphen to show the separateness, yet phonological and semantic de-
pendency of such a locative.
38
Tharaka /tûragita≠ni/ ‘on the tractor’ would be written
<> rather than <>. The relative separateness of the locative
even helps to visually preserve the appearance of the word <>since the slight
lengthening of /a/ is probably due to the fact that it is now penultimate.
39
If there are other
enclitics in the language, they should probably be written in the same way.
Options: [-] or [‘]
Hyphenation is one possible indicator of that close relationship. It shows dependence yet
distinction. Another is use of the apostrophe.
The rationale for using an apostrophe is that with cliticization, there is almost always
vowel elision or coalescence taking place. The apostrophe does double duty in such cases,
signaling to readers that it is tied in with the following word and automatically allowing
for any elided vowels (Karen Van Otterloo, pc). In Francophone countries, the symbol is
usually well-received.
38
As further support for writing locatives as clitics, Bresnan and Mchombo (1995) have shown that in the
various Bantu languages they have studied, locative markers used with nouns are not prexes, while the
nonlocative noun class markers are.
39
Another indication of the distinctiveness of the Tharaka locative is the fact that the tone of the word-
nal vowel on the noun always changes preceding the locative. As Karen Van Otterloo (pc) says, “If it adds a
new element to the melody, it is probably a clitic.”
47
Chapter 2 - Representing Linguistic Features
A minus to use of the apostrophe is that in some orthographies the apostrophe has other
functions, as in Swahili, where it is actually part of a trigraph <>. It’s best not to
use the symbol to represent such a diversity of linguistic phenomena. A typical mother-
tongue English speaker, highly literate, struggles with confusion as to whether to spell the
possessive its with an apostrophe as it’s (incorrect), as is done with the possessive, e.g.,
<>or the contraction for it is, <>. This recurring problem is due to the multiple
functions of the apostrophe and the multiple functions of /-s/ both in the spoken language
and in the English orthography.
If, however, mother-tongue Bantu speakers choose to use <> to represent the velar
nasal in their orthography, overuse of the apostrophe is not a problem!
Augment vowels (e.g., a, e, o) sometimes come before the noun class markers, e.g.,
/ama-/, /omu-/, and /eki-/.
Many Bantu languages have, in addition to their noun class prexes, an augment or
pre-prex, which in some languages functions as a denite marker. Swahili does not use
the augment, so we will exemplify it by using Fuliiru D63(J), a language of eastern Congo.
Fuliiru’s augment
40
appears as a single vowel preceding the noun class prex, e.g.,
u-mu-ndu ‘person’ a-ba-ndu ‘people’ and i-ki-ndu ‘thing’. This pre-prex is generally not sepa-
rated from the noun (Van Dyken, et al. 1993:10).
There are some languages in which the pre-prex functions as a clitic rather than a pre-
x, however. In such cases, it may best be written as such. To write it as a completely se-
parate word is not a good option because the form of the vowel is dependent on the prex
that follows, and in some cases the length of the vowel is also dependent on what follows
it.
The presence of augments is determined by dierent factors in dierent languages, but
they typically carry minimal semantic information. Since they always occur word initially
40
According to Van Otterloo (To appear), in Fuliiru a noun class prex is always an ax. Locative mark-
ers, on the other hand, while often presented in the literature as unquestionably “noun prexes,” function
syntactically, at least in Fuliiru, as monosyllabic words rather than as prexes. If we compare, for example,
the class 19 diminutive nominal prex, with any locative marker, we see that there is an essential dierence
between the two in terms of separability. The diminutive prex usually replaces the normal noun class pre-
x of a noun which is being diminutivized. Especially in class 9/10 nouns, however, the diminutive is some-
times added to a complete noun, preceding its normal noun class prex. Even when added to a complete
word, however, it is still clearly an ax and not a clitic, because it is inseparable from the noun to which
it is attached. Compare ny-úùmbà 9-house and its diminutive, 19-9-small house, with the phrase
ìyó gíìndì nyúùmbà that other house and its diminutive counterpart yìhyó híìndì hínyúùmbà that other small
house. Note that the hí- prex is never separated from the noun to which it is attached. Rather, it remains
attached to the noun, and the modifying words within the phrase have matching concord prexes.
A locative marker of any class, on the other hand, also precedes a whole noun, e.g., ny-úùmbà 9-house/
mú’ny-úùmbà in the house (18-9-house). However, the locative marker, unlike the diminutive prex, is
separable from the word to which it is phonologically attached. Observe how the locative marker appears
only at the left of the phrase as modifying words are added: mú’gíìndì nyúùmbà in another house, mwìì’yó
gíìndì nyúùmbà in that other house. Clearly, the mú is independent from any specic part of speech, and
rather is joined phonologically to whatever word which happens to follow it within the phrase. Thus, the
locative marker can be separated from the noun to which it seemed to be attached in the original phrase
mú’nyúùmbà in the house. It simply remains the head of the phrase as other components are added, and
there is no concord within the noun phrase that agrees with the class 18 locative marker.
48
Bantu Orthography Manual
if they occur at all, they invariably interact with the monosyllabic and therefore cliticized
associative markers, e.g., , ‘of’, coming before them in an associative phrase. For
example, in Fuliiru /i-bi-ndu by-a u-mw-ami/ becomes [ibindu byomwami] ‘the things of the
king’, according to Karen Van Otterloo (pc). The augment itself may be an ax or a clitic,
depending on the particular language. This issue can probably be determined by following
the word break criteria given by Kutsch-Lojenga (1993:7), applying the principles of mobil-
ity, movability, separability, conceptual unity, and referential independence. Consider, for
example, whether it is possible to insert another word between the augment and the noun
it precedes, or whether it is possible in any circumstances to pronounce the noun in a sen-
tence without the augment.
The best option for longterm readability is preserving the underlying morpheme, <bya
umwani> in this example, though speakers may want to write it as it sounds as part of the
phrase.
: <or <> or <or <>
The rst option is axation. In the Malila and Shi examples which follow, the augment
vowel should be written together with the rest of the word. Kutsch-Lojenga (pc) says, “the
quality of the vowel depends on the vowel in the prex which follows (in Malila it must be
the same as the prex vowel; in Mashi, it is e, a, o, according to the front, central, back
vowel of the prex); often (and certainly in Malila), the whole word, including the aug-
ment, forms a tonal unit.” Tonally, augment + prex + noun form a unit.
Malila (M24) examples with augments:
ladle knife
girl congregations
girls rope
bones muscle
For all of them, the following orthography was chosen: <>, and not <
>; <>, and not <>.
In the following examples from Shi (D50), the vowel of the augment agrees in front/cen-
tral/ or backness with the prex vowel. For front and back, it is one degree lower than the
prex vowel (Kutsch-Lojenga pc). Tone is not marked in these examples.
people (indenite)
people (denite) not: a bantu
dogs (indenite)
dogs (denite)
trees (indenite)
trees (denite)
49
Chapter 2 - Representing Linguistic Features
The second possibility is that the augment is a clitic. An indication of semantic separate-
ness (though phonological attachment) of a clitic is vowel lengthening. According to Karen
Van Otterloo (pc), if the vowel of a noun prex is not lengthened when the noun stem is
monosyllabic, but the vowel of the augment is lengthened when the noun is short, this can
be one indication that the augment is a clitic rather than a prex. In Fuliiru an augment
undergoes lengthening in such a situation, and because of its mobility and separability it is
clearly identiable as a phrase-level clitic. The augment, in such a case, can be considered
a monosyllabic word which attaches phonologically to the following noun or adjective. The
following examples are from Fuliiru (Van Otterloo, to appear).
a) í kí-he [ííkíhe] AU’7-time
b) í kí-nògòsho [íkínògòʃò] AU’7-hoof
c) á bá-nà [ááβánà] AU’2-four
d) á bá-àná [áβāānà] AU’2-child
An augment, then, may be a monosyllabic word, and as such may be written separately.
41
Associative markers should usually be written separately from the nouns to which they
refer. A word break challenge arises when the word following the associative begins with
a vowel (often an augment). This vowel often alters the pronunciation of the associative
marker when it is part of a noun phrase, because the vowel of the augment typically
harmonizes with the vowel in the class prex (Gardner 2005:30) in the word which comes
after it.
The following example is from Ikiizu (E402).
1/1a
umuryakari
bride
wa
of
baana
children
2/2a
abaryakari
brides
be
of
kesebe
udder
41
Literacy specialists seem to disagree with linguists on this point. Literacy people hope to facilitate reading
uency and comprehension by indicating the semantic separateness of the distinct words. They hope readers
will eventually be able to read silently, because comprehension is much better with silent reading; the reader is
not forced to simultaneously decode print, vocalize, and listen with comprehension to his own speech.
50
Bantu Orthography Manual
3
umuri
root
gwi
of
nyumba
house
4
imiri
roots
jɨ
of
ngɨrɨ-gwasi
warthog
The associative has an underlying <> as its vowel, but because it is pronounced to-
gether with the following word, the <> coalesces with the augment and elides. For more
detail, see section on axes vs. clitics.
Since word level rules, not phrase-level rules, usually guide spelling, it is best both for
writers and for readers if associatives are always written with <-> word nally. When
associatives precede vowel-initial nouns, mother-tongue speakers may want to write them
reecting the vowel coalescence processes of their language. This makes the associative
marker dicult for the reader to recognize because its form will vary enormously, and
because the one unvarying letter before all consonant initial words, <->, will be missing.
Such surface-level writing, though it helps new readers, eventually slows spelling and writ-
ing rules. Acceptability to mother-tongue speakers must be considered, however, so present
the options carefully (see Part 2, appendix C).
Options: <- or –>
In some languages, e.g., Babole (Leitch 2003), there may be a dierence between associa-
tives used for possession and those used for all other purposes. For possession there is an
“extra” initial vowel that is absent in the other associative forms. This is seen in the con-
trast between (1) and (2) below.
(1) < >
8:food 8:ASS 2:visitor
‘food (set apart) for the visitors’
(2) < í à>
8:food 8:ASS me
‘my food’
The additional morpheme -í- in (2) looks like a remnant of an ‘augment’ and only occurs
in possessive associatives (as opposed to generic associatives). In rapid speech, this initial
vowel commonly elides, and is easy to miss, but the distinction, if it exists, must be repre-
sented in the orthography. For additional details see Leitch 2003:419.
Options: <
> or
<
51
Chapter 2 - Representing Linguistic Features
Options for writing them do not include hyphenation. Associative markers should not
look like locatives, stuck onto the beginning of the noun. Also, since in Swahili the associa-
tive is written disjunctively, people should not object to writing them separately from the
nouns they precede. People do not usually question the suggestion. A problem arises, how-
ever, in that many people will want to follow phrase-level, rather than word level (lexical
rules), as pronounced when the associative is part of a noun phrase in their speech. For
example, an associative marker which precedes a vowel-initial noun will often have its
pronunciation altered by the vowel of the noun it modies, in rapid speech, or by vowel
harmony rules.
In the following example from Ngoreme, note the changes in the vowel of the associative
marker for some noun class 1/1a words:
1/1a
The Ikiizu (E402), on the other hand, chose to just write the associatives and the loca-
tives in their underlying forms. After all, they have seven vowels, and the variant speech
forms are many indeed. They also are convinced that the associative is always “wa,” de-
spite its morphing in speech before every V-initial noun.
This is a nonagreeing associative marker, according to Van Otterloo (pc). As such, it
should be written separately, or as a clitic attached to the word it precedes with <> or
<>. A word break will facilitate recognition of the word which follows it and will also
avoid merging dierent parts of speech into one written form. If a word break is not pos-
sible, try indicating its separateness with an apostrophe or a hyphen.
[na=] Options: <> or <> or <>
As noted earlier, much elision and coalescence of vowels is done at phrase-level in rapid
or normal speed speech, and since these vowel changes result from phrase-level rules
rather than word level rules, they should not be represented in the orthography. However,
with clitics, there may be undoable elision, even with unnaturally slow speech. Especially
in the case of augments which are clitics rather than prexes, there may be word level
coalescence of vowels between the augment and another preceding clitic such as the asso-
ciative marker. Since such elision/coalescence is not truly a phrase-level rule, it may best be
indicated orthographically.
In determining whether the augment must be written on the noun after an associative or
conjunction, try to nd out two things:
1. if there are any instances when the augment cannot be pronounced at the beginning
of the noun, because it changes the meaning when it is pronounced together with the noun.
2. if there are two ways to pronounce the augment in the same phrase (one reecting
fast speech and one preserving the underlying forms).
If either of these conditions is true of the language, write the augment in its full form,
and write the associative in its underlying form, usually <Ca>. No elision! Readers will
52
Bantu Orthography Manual
more quickly recognize the associative and the noun which follows it, and within a few
weeks of practice they’ll read aloud as they actually speak.
If possible, determine a base form for each clitic and maintain that shape (spelling). This
will help develop uent reading, but in the early stages of literacy, it will make the teach-
ing of writing more dicult if it is conicting with word level rules.
If the coalescence involves grammatical words which cannot be isolated from another
word even at the slowest rates of speech, this indicates a word level rule; the spelling
should indicate the functor as part of the word. The solution is to reect linguistic reality. If
the augment is actually a clitic, it should be written as such.
The noun plus augment would appear orthographically as a clitic group:
e.g., u’mundu. In such a case, the only option for the conjunction plus
noun, phonetically [nomundu] would be to write: <
>. This
would accurately reect the linguistic facts that the augment is a clitic
and that word-level coalescence takes place between the conjunction and
the augment. This would maintain the word shape of the noun stem in
both cases. And while such a solution would result in several shapes for
the conjunction no’, ne’, na’, etc., (e.g.,< > the thing, <>
and the thing, <> the people, <> and the people), this dif-
ferent shape would still accurately reect the shape of the conjunction at
word-level. The same sort of allomorphy occurs in noun prexes at word
level, and no one would even think of trying to undo it for the sake of
maintaining word shape. For example, in Swahili, the class 2 prex wa-
may appear as <-> as in watu people, or as <>- as in <> all
or<> good and no one would think of writing <> or <
> just to keep the morpheme shape the same (Karen Van Otterloo, pc).
In summary, aim to reect the linguistic status of the lexical units (the “linguistic re-
ality”). Also, try to reect word level rules, not phrase-level rules. The best result of an
orthography decision for a clitic is one which allows people’s silent reading to be uent
(recognizing meaning quickly) and allows their oral reading to make adjustments, so they
can pronounce a word as it would be spoken in rapid speech. People can usually be given
some linguistically acceptable options, and then are able to choose between them, depend-
ing upon how they want to indicate the coalescence.
Bantu verbs have so much axation that people tend to think that everything preceding
the verb stem must be a prex. As Karen Van Otterloo (pc) points out, however, there are
usually:
…at least one or two verb forms in which there is an auxiliary (or even
multiple auxiliaries in a single verb, cf. English will have been gone, and
Kifuliiru <>‘they are intentionally singing
again’). In verbs with auxiliaries, each auxiliary is a separate word from
the main verb. The main verb is gone in the English example, and
‘sing’ in the Kifuliiru example. The most common separate-word auxiliary
will probably involve a form of the verb ‘to be’, often or ,
which will bear the subject prex, perhaps along with other[s].
53
Chapter 2 - Representing Linguistic Features
The main verb which follows may be innitival, or it may also have a subject or object
prex, or even be a bare stem with no prexes.
Expect a copula which is followed by a main verb to be written separately.
Expect to nd two separate verbs whenever you encounter two subject prexes in a
stream of speech, as in the Fuliiru sentence /bagwetibagashubiyimba/ ‘they are intention-
ally singing again’; a natural word break will be signaled by a subject prex, so <
> will be the rst attempt at word breaks. In Fuliiru there are other indi-
cators which show that there is also a word break between /bagashubi/ and the main verb
/yimba/, thus giving <>, a three-word verb form (Karen Van
Otterloo, pc).
Expect the presence of an innitive marker to also indicate the beginning of a separate
verb. You may nd either the normal innitive marker () or a prex of noun class 5,
which is usually an /i-/ or an /e-/.
Expect to nd a subjunctive nal vowel (usually /-e/) on the main verb of a subjunctive
verb form. If the subjunctive ending has instead been replaced by a default nal vowel, this
may indicate there is a word break in the verb. Often the subjunctive ending will not show
if there is an auxiliary between the prexes and the main verb.
Options: <> or <> or <>
As we said when we began, linguistic analysis will undergird any writing system which
accurately represents the speech and perception of the mother-tongue speaker. Toward
that same end, the principles and suggestions given in this chapter incorporated pedagogi-
cal, perceptual, and sociolinguistic considerations into the decision-making process for
orthographic choices. Your team should now be prepared to make word break and graph-
eme choices, which will be recorded in the Workbook Form. Most tables and charts in the
Workbook Form are set up to hyperlink the reader back to the linguistic features described
above.
54
In the words of Keith Snider, “As with everything else in orthography design, try some-
thing and then test it thoroughly” (Snider 2001:6). If you know what to look for and how
to identify it, you will be able to assess the of the linguistic analysis, the social
of the writing system, its and even .
Usually, two aspects of orthography require testing: comprehension and u-
ency. Comprehension can be tested by having subjects read one or more selections and then
asking content questions. Fluency can be tested by having subjects read one or more selec-
tions orally and then noting the places where they stumble. The people who conduct the
test should record the dierent readings on audio cassette, and then later note the number
and type of reading errors made.
A note on the value of tape-recorded tests: A literacy specialist who is not a mother-
tongue speaker of the language is likely to miss people’s mistakes with oral reading. He/she
will be limited in distinguishing some minimal pairs, for example, especially in the ow of
speech in a text. Since the passage can only be read once by each reader, there are no sec-
ond chances for the tester to note everything! If the test is recorded, though, you can listen
repeatedly for problems, along with a mother-tongue speaker or two.
You may also need to test people’s ability to the language, applying its spelling
rules. There are also two aspects that should be tested: accuracy and speed. A dictation test
usually suces, with the tester noting the number and type of errors made. If people are
very slow in writing the words dictated, the tester should note this as well.
1. Speakers’ perception of a feature of their language which usually has lower
speaker awareness (i.e., breathiness, tone, length)
2. Aspects of the language which are left ambiguous for the reader—level of
diculty in guessing whatever is not specied or contrasted
3. Word break (and comprehensibility) questions with long words or clitics
4. Vowel deletion when it comes across word breaks (i.e., Is it possible to write out
both words fully?)
5. Vowel coalescence and morpheme loss or other loss of contrast
6. Spelling rules—which, if any, confuse people?
7. The extent to which tonal contrasts need to be marked
8. Mother-tongue speakers’ response to altermother-tongue representations of tone
55
Chapter 3 - Orthography Testing
This depends upon the purpose of the test. Usually, though, test:
1. Members of dierent dialect areas
2. People who are literate in a language of wider communication
1. Just ask people if they have diculty reading or writing the same feature in a
variety of contexts.
2. Elicit mother-tongue speaker reaction to a problem area. Asking people “Which
option do you prefer?” predisposes them to choose whatever they’ve been
accustomed to seeing in another language. Instead, say, “Which option is true of
your language?”
3. Observe people trying to read or having problems in literacy classes.
Are certain symbol sequences dicult to teach?
Are students having diculty spelling certain symbol sequences?
1. Dene the problem to be tested.
2. Design a test which will give you information on how to solve the problem.
3. Decide whether the test should be oral or written. You can ask people to read
aloud, to write what is read to them, or to read silently and express their
comprehension of what was read (i.e., a minimal pair).
4. Carry out the test.
1. Ask a mother-tongue speaker to prepare two texts that are written dierently,
each using a dierent version of the orthography feature you are testing. The
passages must be reasonably dicult.
2. Ask individual mother-tongue speakers to read both passages aloud.
3. Listen for hesitations, repeats or errors.
4. Ask each person which text they prefer and why.
5. Write your observations after each oral test.
6. After conducting several tests, look for a pattern. If certain problems recur, do
they present reading problems?
56
Bantu Orthography Manual
: to test the symbolization of a fricative b [β], which is contrastive with
a plosive b, write one test using to represent the fricative, and another text in
which the fricative is written .
: text will show things a simple list cannot. Certain features in the context
may make up for ambiguities in the spelling system, so that the reader can still
read with uency and comprehension, given enough context (even a comparison
of short phrases such as vs. , vs. ).
Rather than have someone read two dierent stories, you can opt for two lists
of sentences, for example, written two dierent ways. See how much people
stumble, if at all, when they come to the redundant morpheme, now unwritten.
Sarah Gudschinsky (1973:126) suggested that reading a text involves two steps:
1. recording the oral reading of a passage
2. assessing the comprehension of it
The test can then be studied as to how many real mistakes were made and why. Was the
passage understood clearly? Were there any ambiguities for the reader? Did he or she have
to read the passage more than once to get the meaning? An orthography which requires
more than one reading is problematic.
1. Ask a mother-tongue speaker to prepare two wordlists that are written dierently.
2. Ask other individual mother-tongue speakers to read both wordlists out loud.
3. While the people read, listen for any hesitations, repeats, or errors.
4. Ask people which list they prefer and why.
5. Write your observations after each test.
6. Look for patterns in your data….
: the language has been previously written, with no dierentiation
of [b] and [β]. If the PTEST has shown complementary distribution of the two
phones,
1. Ask a mother-tongue speaker to prepare one wordlist in which both [b] and [β]
are written the same.
2. Ask a speaker of the language to read the words.
3. Note any problems they have. In the case of minimal pairs or other forms of
ambiguity, ask the reader to tell you the meaning of the word/words in question.
4. Ask the mother-tongue speaker to prepare a second wordlist, in which the two
phones are represented dierently.
5. Ask the speaker if he or she perceives a dierence in the sounds of the two. Lists
of minimal pairs can be very useful but are not always possible.
6. Find out whether the speaker is consistently able to sort the words into two
categories (i.e., stop and fricative).
57
Chapter 3 - Orthography Testing
1. Ask a mother-tongue speaker to prepare a wordlist or short sentences with items
to test.
2. Ask a mother-tongue speaker to read the wordlist or short sentences to a small
group of mother-tongue literates.
3. Have literates write the words as dictated. Note where people have a problem.
: Vowel length is conditioned before prenasalized obstruents. Mother-
tongue speakers, however, hear these vowels as long. You would prefer to write
only true vowel length in the language, because sometimes vowel length is
truly contrastive, either lexically or grammatically. You give the spelling rules
to a group of mother-tongue speakers and then give a dictation test to see how
dicult it will be for them to write vowel length only when it is contrastive. Note
trends.
To devise a multiple choice cloze test with options written out:
1. Ask a mother-tongue speaker to write several sentences containing the
phenomenon in question, such as distant past tense.
2. Remove the word containing the problematic element, replacing it with a
______________.
3. Write the word, spelled in the alternative mother-tongue ways you are
considering, below the sentence.
4. Ask mother-tongue speakers to read the sentences, circling the spelling they prefer
for each missing word.
: This type of test combines silent reading and writing. It’s useful for evaluating
people’s comprehension of some underdierentiated element. It is easy to assume that the
context will give the reader all he needs, to compensate for that underspecication. This
test may clarify his ability to really do this.
It also simply shows you the reader’s preferences. The Tharaka language uses extra-long
vowel length to indicate distant past tense. Tone is the only other indicator of that tense,
but tone is not written in any of the surrounding related languages, so writing of tone is
not an option. A mother-tongue speaker devises a test in which there are other cues which
indicate to the reader that the event described in the sentence took place long ago. For
example, ba + a + akire, ‘they built’ is written two ways (, ), and partici-
pants are asked to write the word in a blank in the sentence, in the form they prefer. Seven
or eight similar sentences are given.
Long ago, Noah and his family_________________() an ark.
Further testing may be required, if no satisfactory solution is found to your spelling
dilemma. Try representing a morpheme such as this one, which is conveyed orally through
tone as well as length, by using a punctuation mark to indicate tense/aspect at the begin-
ning of the word (for example, * might indicate distant past tense, automatically cuing the
reader to a melody and rhythm associated with this tense/aspect feature on the verb). Such
58
Bantu Orthography Manual
a symbol would communicate meaning directly to the reader, rather than pronunciation.
Test again for readability (and acceptability) with several people.
The test above is limited. It doesn’t show the tester or the testees whether the distinction
is necessary; it only shows what people prefer. To provide more insight, either for yourself
or for the readers, provide two texts written with the dierent representations (see above)
and ask comprehension questions. People can read silently or aloud. The tester gains infor-
mation, either way, and the testee’s awareness of comprehension issues is heightened.
Text example version 1: Long ago, Noah and his family an ark…
Text example version 2: Long ago, Noah and his family an ark…
1
Text example version 3: Long ago, Noah and his family an ark…
Comprehension question: When did the story take place? What word or words tell
you this?
Testing for tone, in particular: The real issues at stake with respect to tone and orthogra-
phy are what tonal melodies are in contrast and at what level of the phonology is the mother-
tongue speaker most aware of these contrasts? So discussing how to represent contrasts is
really secondary to discussing which contrasts to represent and at what level of the
phonology or morphology they should be represented.
Not marking tone in any way is probably the easiest strategy for writing,
but it can pose problems for reading comprehension. Whether it does pose
problems for readers or not depends on how great a functional load tone
bears. One way to determine how great the functional load is, is to take
texts of varying degrees of diculty, not mark tone on them in any way,
and then use them to test reading uency and comprehension. A good
strategy when testing a zero marked text is to note in particular where
the diculties lie, if indeed there are diculties. (Snider 2001:5-6)
The three things to look for when testing for any possible tone marking rules are:
1. uency of reading
2. comprehension of a text (not a wordlist)
3. spelling accuracy and ease
It is always simplest to begin with less marking and to add whatever is needed to mini-
mize ambiguity
2
for the reader. These are the recommended steps to take in testing.
1. Look for ambiguity in a list of isolated words:
1
The tonal melody (marking the tense in question) is represented by a symbol preceding the verb
2
For more information on tonal ambiguity analysis for African languages, see Roberts, 2008.
59
Evaluating Test Results
a. Generate an exhaustive list of homographs in the language (disregarding tone at this
point)
b. Look for sets of ambiguous words (homographs), and identify their respective parts
of speech.
c. These sets will be your reference for all further research
2. Analyze the frequency and distribution of these homographs in texts (a variety).
These are also known as tonal minimal pairs.
3. Start with a text unmarked for tone, and a strong reader who has plenty of exposure
to the orthography. Perform a miscue analysis of oral reading.
a. I recommend using a tape-recorder, but testing can be documented on-the-spot as
well.
b. Make sure that both your reader and listeners are completely unbiased. They don’t
need to know the purpose of the test, other than that we are checking to see if their writ-
ing system works well for them.
c. Use a 1st person narrative text (or some other text the reader and listeners have not
seen or heard before), as clean and up-to-date as possible. Everyone will listen as a strong
reader reads the text aloud.
d. Instruct the listeners raise their hands when the melody is wrong. (If the melody is
wrong, also ask if the meaning is unclear).
e. On your own copy of the text, perform a miscue analysis as the reader reads aloud.
Mark all mistakes, self-corrections, and long hesitations. If comprehension is questionable,
the listeners and the reader can also be asked a few comprehension questions.
f. If melody irregularities are identied by the group of listeners, or if there were
signicant hesitations and autocorrections or confusion about meaning, use an equivalent
text, which is marked for tone. Use the same reader, if possible (of course he/she must al-
ready be well-practiced with reading the tone marks).
g. Compare the results with those from reading of the unmarked text.
Document everything as thoroughly as possible.
If there is conict between written results and verbal opinions expressed, give priority to
written results.
If there is conict between comprehension of a text and verbal opinions, give priority to
the results of the comprehension test.
60
Most Bantu languages will not exhibit a high functional load for tone. In Bantu A lan-
guages, and some others, however, the role of tone may be too important to be ignored
orthographically. Stephen Anderson, Cameroon, uses the Bantu Phonology Tool, PTEST, to
generate two lists of tone contrasts in a specic language (one list for verbs and another for
nouns). These lists can be rather long, especially if the language is from Cameroon. Steve
then edits them to provide a unique list of all the multiple tone contrasts. The following
two paragraphs are his words (pc):
To understand the process, imagine a simple tone triplet
1
. The computer
will actually generate three tone pairs instead of a tone triplet. I correct
this by editing. Also, depending on the quality of the input, the computer
may generate lots of pairs that really have to do with only a single root.
That root may have multiple meanings (either real homonyms or just mul-
tiple meanings of one root). I correct that as well by editing. When I am
done, I have a single list of all the tone contrasts, without what I consider
false data that would skew the results. Once I have my unique list of tone
contrasts, then I am ready for the next step, marking on a separate sheet
the number of contrasts that occur, feeling this may be a more accurate
indicator of “the functional load of each contrast”. In practice, this has
been very helpful in deciding which tones to write and which to dare to
leave unwritten (our major hurdle here).
Sometimes a tone contrast is for more than a pair of words, so I also get
rid of any homonyms or apparent homonyms so that the tonal roots oc-
cur only once in each multiple tone contrast. Once the (dual) lists of tone
contrasts have been fully edited, I go through them slowly and marks
each tone contrast by making a line between two tones on a rough chart,
as below:
Then, each time I get an additional pair of words for a contrast already
indicated, I make a little slash mark on that particular line to indicate
quantity of contrasts.
The rst column in the table below is actual data from a 440-wordlist of Rangi nouns.
Oliver Stegen (pc) shows the contrasting tone melodies for which minimal pairs have been
found, and the second column gives the count (“slash marks on the line”). L=low, H=high,
1
Concerning the computer writing a tone triplet as three tone pairs, ‘muumba’: with RM# melody it means
‘Creator’, with FL ‘Maasai’ and LL# ‘barren woman’. Instead of treating the three as a triplet RM# vs FL vs
LL#, the computer will list three pairs: 1. RM# vs FL, 2. RM# vs LL#, 3. FL vs LL#. Steve’s editing triplets
back into the contrastive tone melody list will lengthen the chart. The benet: the linguist could see three-
way distinctions immediately at a glance, which would be lost when displaying tone ** only.
61
Appendix A: Determining Functional Load for Tone
M=mid, F=falling, R=rising tone and # =a boundary tone phenomenon (e.g., ‘x’ in
Anderson’s example above).
FL# vs HM# | = 1 instance
FL# vs LL# ||||| = 5 instances
HM# vs LL# || = 2 instances
HM# vs RM# | = 1 instance
RM# vs LL# | = 1 instance
FL vs HH | = 1 instance
FL vs LH |||| = 4 instances
FL vs RH | = 1 instance
HH vs LL | = 1 instance
HH vs LH | = 1 instance
HH vs RH | = 1 instance
RH vs LH | = 1 instance
It can now quickly be seen that in Rangi, falling versus low tone melodies may need to
be distinguished lexically (high instance of vs ). By contrast, even though both
and occur as distinct tone melodies in Rangi, they never contrast. Consequently,
they can be grouped together and marked with the same pattern (in the case of Rangi, no
tone mark at all versus tone mark on the rst stem vowel for falling melodies). Similarly,
vs never contrast, nor do vs . So basically, this chart shows us that the
functional load of tone is a lot higher on the rst stem syllable (contrasting , , and )
than on the second/nal (where no consistent contrast could be identied, and consequent-
ly, tone is never marked on the nal stem syllable in Rangi). A revised chart, taking into
account only the tone on the rst stem syllable, would then look like this:
F vs H || = 2 instances
F vs L ||||||||| = 9 instances
F vs R | = 1 instance
H vs L |||| = 4 instances
H vs R || = 2 instances
R vs L || = 2 instances
62
Bantu Orthography Manual
At the end of this process, the following can be seen:
1. Which TONE PAIRS actually contrast, which do not
2. Which TONE PAIRS have a high functional load lexically, and which have lower
functional loads
The result of this process: clues as to which tones need to be distinguished from other
tones and which do not. “Depending on the language, one can often lump several tones to-
gether as carrying minimal functional load. Sometimes it is good to leave the largest group
of tones unmarked in order to minimize tone marking” (Stephen Anderson, pc).
The dual lists function completely independently. Verbal and nominal tone usually in-
volve completely dierent issues. One possible outcome, then, of the tonal contrast counts
may be a decision to mark tone for some verbal forms or for verbal roots, while not mark-
ing it for nouns.
The nal step involves going through the minimal pairs of words which one has decided
not to mark dierently (orthographic homonyms), to see if the two meanings are likely to
be confused in context. If there is little chance of confusing the orthographic homonymns
in textual material, then one feels better about minimal marking of tone. If certain words
are potentially confusing, however, it is better to nd some creative way to dierentiate
them orthographically.
This is often the case with grammatical words, and small words which carry less seg-
mental or context-dependent information. They often occur frequently in text and aect
people’s oral and silent reading uency. This is why, for example, it is so important in the
Spanish language for the stress dierence between <esta> ‘this’ and <está> ‘it is’ to be
reected in the orthography.
‘I walked’
‘we walked’
‘I will walk’
‘we will walk’
‘I should walk’
‘we should walk’
‘I am walking’
‘we are walking’
63
a monosyllabic (in Bantu languages) word which is phonologically bound to
another whole word.
a phonological phenomenon involving the deletion of one
segment accompanied by an increase in the length of another, usually adjacent to
it, thus preserving syllable length or “weight”. This happens often with vowels
in Bantu languages, in which vowel length, while not lexically contrastive in a
certain instance, is phonetically created. Automatic vowel lengthening in the
penultimate syllable is one example; another occurs when a consonant is elided
between two vowels. A third source of compensatory lengthening takes place
before a prenasalized stop. Fourth, it can take place after a glide formed from a
vowel (Gardner: 2005:17).
a linking verb which has little independent meaning. It serves to relate dierent
elements of the clause structure, especially subject and complement. A prime
English example is the word “is”, though there are many others which function
similarly.
a rule of consonant dissimilation. A voiceless stop becomes voiced if the
consonant in the next syllable is voiceless. (Example: Gikuyu, from Kikuyu). In
other words, “When two successive syllables in a stem each begin with a voiceless
plosive, then the rst of these becomes voiced” (Schadeberg 1999:391). This
process usually occurs in prexes, as in the Kikuyu example above, in which the
prex /ki-/ becomes /gi-/ before /-kuyu/. In some languages only traces of this
law remain; in others it has never applied.
relative importance of a linguistic feature in conveying grammatical
or lexical information. Example: If tone is the only indicator of passive tense
for Kitharaka verbs, its functional load is very high in that context. If both tone
and vowel length are indicators of passive tense, then tone’s functional load is
lower. Lendu, a Central-Sudanic language, provides an example of grammatical
tone carrying a very high functional load. This is what the Lendu orthography
would communicate, without tone depicted in the writing system. See also BOM
Functional load.doc for a method to determine functional load of tone for Bantu
verbs.
“I walked”
“We walked”
“I will walk”
“We will walk”
“ I should walk”
“We should walk”
64
Bantu Orthography Manual
“ I am walking”
“We are walking”
the expression of a semantic contrast using grammatical forms
rather than content words. Grammaticalization takes place over time in a
language, and words which carry distinct meaning, such as nouns and verbs, can
erode and become functors.
the occurrence of two vowel sounds without pause or intervening consonantal
sound
“Given two consonant-groups in a word, separated by a vowel and
themselves aspirated, and provided that they are in the same root, then one
(and normally the rst) is deprived of its breath feature,” i.e. it is de-aspirated
(Schadeberg 1999:391).
language of wider communication, such as French, Swahili, Portuguese or Luganda
: prescribes the formation of geminate (long)
nasals when a nasal adjoins a consonant, and the assimilation of the nasals to the
point of articulation of the underlying consonant. This happens when post-nasal
voiced consonants become nasalized, e.g. Ganda E15(J) (Hyman 2003:51):
/n≠bomb-a/ m≠momb-a ‘I escape’
/n≠limb-a/ n≠nimb-a ‘I lie’
/n≠jung-a/ ɲ≠ɲung-a ‘I join’
/n≠gend-a/ ŋ≠ŋend-a ‘I go’
There are several variants of this rule, one of which (the Kwanyama Rule)
removes prenasalization from the second consonant in the root (Schadeberg
2003:148).
An alternation between phonemes when two morphemes
are adjoined
a clitic which depends upon a following word (i.e., an article such as the, a, or
an in English)
process in which a language turns obstruents followed by i and u into
strident fricatives or aricates
a common phenomenon in Bantu languages in which vowels are divided
into two mutually exclusive groups. The vowels of one group tend to occur
with each other in words, to the exclusion of vowels of the other group. Vowel
harmony can be seen as the spreading of a phonetic feature (back, high, round,
advanced tongue root) to certain vowels, and not to others, within a word. The
spread is usually leftward from the root.
ↆↆↆↆ
65
(Lecture given to Orthography class, SIL-UND, June 29, 2005)
Mature readers need a constant visual word image
Mature readers read by sight (Venezky 1970), and do not take the time to sound out the
words they read. For this reason, a good orthography maintains a constant word-image.
This helps minimize the eort it takes for a developing reader to memorize the shape of
each word.
Immature readers need to spell words the way they perceive them to sound
Beginning readers often sound out their words. For this reason, a good orthography bases
the spelling of its words as closely as possible on the way the native speaker perceives the
words to sound.
There can be tension between the needs of mature readers and the needs of immature
readers.
If we write words the way they are pronounced in order to help immature readers as
much as possible, this can cause problems for mature readers. This is because the pronun-
ciation of words changes in dierent environments and writing words with these changes
means that the mature reader will not have a constant word image. Here is an example
from Chumburung, a language spoken in Ghana.
Words spoken in isolation
‘Ko (person’s name)’
kòfí
kùŋú
‘head’
1
This paper is an expansion of §2 in Snider (2001). Since I am the author of both works, I have taken the
liberty of including in the present work a number of sections verbatim from the earlier work. I have also
taken the liberty of editing some of these sections liberally to better t the present purposes. This has made
it dicult to acknowledge which material comes from the former work and which material is new and so I
have not done that.
2
This is true only for alphabetic orthographies. A logographic orthography, in which there is a one-to-one
correspondence between a symbol and a word or morpheme, sidesteps this issue, of course.
66
Bantu Orthography Manual
Words spoken in context
kòfú kuŋu ‘Kofi’s head’
Notice that the [í] in ‘Kofi’ changes to [ú] in ‘Kofi’s head.’ One thing that nonlinguists
are not always aware of is that there is often a difference between the way a native
speaker pronounces a word and the way he/she (hereafter he) perceives it to sound. In the
case of this Chumburung example, while the native speaker pronounces ‘Kofi’ with an [í]
in isola-tion (i.e., citation form) and with a [ú] when it comes before the word for ‘head,’
he doesn’t realize he is making this change. In other words, he does not perceive this
change.
The “trick” in developing a good orthography is to write words the way the native
speaker perceives the word to sound, not necessarily the way the native speaker actually
pronounces them. This is great for the immature reader. There is an added bonus,
however, and that is that when we write this way, we are also able to maintain a constant
word-image. Writing the way native speakers perceive the language to sound meets the
needs of both mature and immature readers with the same orthography.
Developing an orthography would be relatively easy if the pronunciation of the sounds
in the language was stable. Unfortunately, as we see in the Chumburung example above,
this is not the case. Sometimes, as in the Chumburung example just discussed, the changes
are below the level of native awareness (i.e., the native speaker isn’t aware that he is
saying the words differently). Here is another example of a change that the native speaker
isn’t aware of, this time some plural forms from English.
cap–s cab–z
bit
–s bid–z
buck
–s bug–z
Here we see that the native speaker of English pronounces the plural form as (voiceless)
s when it follows voiceless consonants and as (voiced) z when it follows voiced consonants.
In other environments, English has a contrast between s and z, as in sip and zip. However,
this contrast is lost when the sounds immediately follow a consonant and which of the two
sounds is produced is totally predictable from the environment. Although native speakers
are aware of the dierence between s and z, they are usually not aware (unless it is pointed
out to them) of the dierences in the above words.
At other times, however, the native speaker is well aware that he is saying sounds dier-
ently. A good example of this occurs in the English negative prex in-.
im–possible
in
–tolerant
il
–logical
ir
–reverent
In these examples, the native English speaker is well aware that he is saying an m before
possible and an r before reverent, etc. He is so aware of this dierence in fact, that he prob-
ably doesn’t even realize that these variants are all forms of the same prex, but rather
67
Appendix C: Morphology vs. Phonological Rules
he thinks of them as dierent prexes. So, some rules produce dierences that the native
speaker perceives (we’ll call these word rules), and other rules produce changes that the na-
tive speaker does not perceive (we’ll call these phrase rules). Since this is the case, in order
for an orthography to represent only sounds that the native speaker perceives, we need to
be able to distinguish between these two types of rules.
In this section, there are two data sets of English examples. The rst set demonstrates a
rule called English Flapping. This is a phrase rule and it produces changes that the native
speaker of English does not perceive. The second set demonstrates a rule called English
Trisyllabic Laxing (hereafter TSL). This rule is a word rule and it produces changes that the
native speaker of English does perceive.
English Flapping (Kenstowicz, 1994:195)
a) aɾom cf. atom
b) meeɾ-ing cf. meet
c) whaɾ is wrong? cf. What
Here are the relevant facts concerning English Flapping:
a) An intervocalic coronal stop changes to a sonorant coronal ap ɾ when it occurs
between two vowels, the rst of which is stressed.
b) While /t/ is a contrastive sound (phoneme) in the language, [ɾ] is not.
c) Native speakers do not realize that they are saying [ɾ] dierently from [t].
English Trisyllabic Laxing (Kenstowicz, 1994:196–197)
a) divine [aj] divin-ity [I]
serene [ij] seren-ity [ɛ]
profane [ej] profan-ity [æ]
b) vile [aj] vil-ify [I]
clear [ij] clar-ify [ɛ]
c) rite [aj] rit-ual [I]
grade [ej] grad-ual [æ]
d) tyrant [aj] tyrann-ous [I]
penal [ij] penal-ize [æ]
e) tyrant [aj] tyrann-ous [I]
fable [ej] fabul-ous [æ]
68
Bantu Orthography Manual
Here are the relevant facts concerning English Trisyllabic Laxing (TSL):
a) The rule shortens and laxes a long vowel when it is followed by two syllables,
the rst of which is unaccented.
b) not adequate. Forms like nightingale, stevedore, and ivory do not shorten the ini
tial vowel. The change is only conditioned by the addition of a sux.
c) Not all suxes trigger TSL (cf., brave/brav-ery, might/might-ily, pirate/pirat-ing).
d) Within the class of suxes that do trigger TSL, there are still idiosyncratic lexi
cal exceptions (e.g., obese/obes-ity).
e) Native speakers are aware of these changes.
Trying to determine which sounds a native speaker perceives and which sounds a native
speaker does not perceive can be extremely dicult since the matter can be very subjec-
tive. For this reason, I am including questions below that one can use to determine more
objectively which is which.
3
In order to help discover word rules, that is, rules that produce sounds that the native
speaker perceives, ask yourself the following questions:
1. Does a given rule apply across the board without exception, or are there
lexical exceptions? In other words, in the same grammatical and phonological
environment do some words undergo a particular rule while others do not? If
there are lexical exceptions to a particular rule, then it is important to spell
words that undergo that rule the way they sound after the rule has applied. In the
examples above of English TSL, not all suxes trigger TSL (cf., brave/brav-ery,
might/might-ily, pirate/pirat-ing). This means that there are whole classes of lexical
exceptions. Then within the class of suxes that do trigger TSL, there are still
idiosyncratic lexical exceptions (e.g., obese/obes-ity). This means that the TSL rule
produces changes that the native speaker perceives.
2. Does a given rule lack phonetic motivation? For most rules, we can see a phonetic
reason for why the rule’s output sounds the way it does. For example, sounds
often assimilate to certain qualities of other sounds. If however, a particular
phonological rule does not have any phonetic motivation whatsoever, then spell
the word the way it sounds after the rule has applied. Here is another example
from English.
electrik electris–ity
In this example, the nal k of ‘electric’ changes to s when the sux
–ity is added. There is really nothing in the phonological environment that would
induce a k to change to s before –ity, so it is obvious that the rule lacks phonetic
3
This section and the next owe a great deal to the theory of lexical phonology. For a detailed overview
of lexical phonology, the interested reader is referred to Kenstowicz (1994). For a clear presentation of the
dierences between rules that produce sounds that the native speaker perceives (viz. lexical rules) and
rules that produce sounds that the native speaker does not perceive (viz. postlexical rules), see Pulleyblank
(1986).
69
Appendix C: Morphology vs. Phonological Rules
motivation. This is a tip-o that the native speaker is aware of the change that the
rule makes. The implication for orthography is that we should spell the two words
as follows: elektrik and elektrisity. Notice that this does not permit one to maintain
a constant morpheme-image. It does, however, permit one to maintain a constant
word-image, which is what is important to mature readers. In this case, the word
elekrik will have a constant word-image, and the word elekrisity will also have a
constant word-image.
3. When a given rule applies, does it only apply when a prex or sux is added?
In other words, does the rule have to apply across a morpheme boundary? Note
that we are NOT talking about a word boundary. If the rule must apply across
a morpheme boundary, then the word should be spelled the way it sounds after
the rule has applied. Again looking at the English TSL example, we see that
specication of the environment of the change in purely phonological terms is
not adequate. Forms like nightingale, stevedore, and ivory do not shorten the initial
vowel. The change is only conditioned by the addition of a sux. Since the rule is
conditioned by a sux and would not otherwise apply, this tells us that the native
speaker is aware of the change that the rule produces. hence we should spell these
words with the changes.
In order to help discover phrase rules, that is, rules that produce sounds that the native
speaker does not perceive, ask the following questions:
1. When a given rule applies, is the new sound it produces one of the contrastive
sounds in the language? If the output of the rule is not a phoneme, then spell
the word the way it sounds before the rule has applied. In the English Flapping
rule above, the sound ɾ is not a phoneme in the language and so it should
not be written with the change because native speakers are not aware of the
noncontrastive sounds in their language.
2. When a given rule has applied, do native speakers think that the sound that results
is the same as or dierent from the sound that underwent the rule. If they think
it is the same sound, that is, they don’t realize that anything has changed without
having the change pointed out to them, then spell the word the way it sounds
before the rule has applied. Again in the case of the English apping rule, the
native speaker is not aware that he is not saying [t] in these forms. This is another
reason why we should write the form the way the native speaker perceives it to
sound (i.e., t) as opposed to the way he actually pronounces it (viz. ɾ).
3. When a given rule applies, is it necessary to refer to the internal structure of the
phrase in the rule’s environment? In other words, does the rule apply across a
word boundary? Note that we are NOT talking about a morpheme boundary. If
the rule applies across a word boundary, then the word should be spelled the
way it sounds before the rule has applied. This means that the word-image does
not change even though the word may be pronounced dierently in the context
of other words. For an example of this, recall the Chumburung example above,
repeated here for convenience sake.
70
Bantu Orthography Manual
Words spoken in isolation
ko ‘Ko (person’s name)’
kuŋu ‘head’
Words spoken in context
kofu kuŋu ‘Ko’s head’
The rule that changes /i/ to [u] applies across a word boundary. This tells us that the
native speaker is not aware of the change and so we should not write the change. This
permits the word ko to have a constant word-image and not be realized in two dierent
ways.
Following the above criteria helps to establish where word breaks should occur
4
and
eliminates the representation of low-level allophonic variation. It also eliminates the eect
of rules that apply across word boundaries and the eect of phrase initial and phrase nal
phenomena. This allows one to: a) maintain a constant word-image and thereby meet the
needs of mature readers, and b) write words the way the native speaker perceives them to
sound and thereby meet the needs of beginning readers.
A word of caution. Although the above principles present a way to rigorously determine
what should be written and what should not be written, it oers only a promising begin-
ning point for people to try. As with everything else in orthography design, try something
and then test it thoroughly.
Kenstowicz, Michael. 1994. Phonology in Generative Grammar. Oxford: Blackwell
Publishers.
Pulleyblank, Douglas. 1986. Tone in Lexical Phonology. Dordrecht: D. Reidel Publishing
Company.
Snider, Keith. 2001. Linguistic factors in orthography design. In Ngessimo M. Mutaka
and Sammy B. Chumbow (eds.), Research material in African linguistics: Focus on
Cameroon, 323–332. Köln: Rüdiger Köppe Verlag.
Venezky, Richard L. 1970. Principles for the design of practical writing systems.
Anthropological Linguistics 12:256–270.
4
Only those criteria which do not make crucial reference to word boundaries are useful in determining
where word breaks occur. Otherwise, this is a circular argument.
71
72
This procedural guide for linguistic analysis describes a strategy for orthography de-
velopment using a workshop approach. It is very practical in nature. Part 1, the Bantu
Orthography Manual, should be used along with it. This procedural guide has a target audi-
ence: the “Ordinary Working Linguist.” Chapter 1 explains the rationale of Kutsch-Lojenga’s
participatory approach to orthography development, and gives the big picture. Chapters
2–4 provide step-by-step instructions for carrying out a participatory approach via a series
of workshops.
Most of the charts to be lled in electronically were designed by either Karen Van
Otterloo or Constance Kutsch-Lojenga. The author deeply appreciates their very signicant
contributions to this manual based upon their familiarity with Bantu languages and their
experiences with Bantu language speakers. Susanne Krüger and John Duerksen provided
details and suggestions on word collection and data entry.
Native speaker insight is essential to the success of the participatory approach, which
has a spiraling,
1
informal curriculum. Native speakers are introduced to features of the
languages in their area before linguistic analysis begins, and as/after discoveries are made.
They are helped to see the impact of spelling choices upon writers and readers, ongoingly.
Optional lectures and templates are provided in appendices.
1
A curriculum which is additive and repetitive, constantly building upon what people learn, and revisiting
those concepts with greater depth is known as “spiraling.”
73
Develop an orthography for a Bantu language group, in partnership with speakers of that
language (laying a linguistic and educational foundation for the application of that orthog-
raphy to reading and writing).
• Community ownership/involvement in every stage of the orthography development
process
• The centrality of linguistic analysis to orthography development, inclusive of
phonology, morphophonemics, grammar and discourse (since these aspects of a spoken
language overlap and inuence one another on the surface, they will seem to compete
with one another for prominence in a writing system)
• Speakers’ perception should play a signicant role in orthography decisions. That
perception can be developed and enriched for those who take part in the orthography
development process
• Orthography-in-use as goal, and also as means for constant feedback and evaluation
• Questions of readability and writeability will be considered throughout the
development process
• Revisions will be ongoing, and will necessarily reect the political, educational and
social context of the writing system
The approach outlined below was developed by Constance Kutsch-Lojenga. Details and
suggestions have been added by Leila Schroeder and Karen van Otterloo. The general goal
of Kutsch-Lojenga’s approach is to allow participants to discover for themselves the pho-
nemes of their language, by systematically sorting words according to the various sounds
found in them. We have incorporated the goals of alphabet development and documenta-
tion of linguistic discoveries into the process described below.
Each topic is introduced generally, followed by detailed procedural instructions. Each
linguistic feature studied is captioned, i.e., Vowels, Week 1. Below each linguistic feature
caption are the instructions, including documentation of linguistic discoveries in Workbook
Form, i.e., Fill in Table 2, the Vowel Combinations Chart
Since there is a lot of variation between Bantu languages as to the rules which produce
surface forms of morphemes, the participatory approach to orthography development will
74
Bantu Orthography Manual
give signicant attention to morphological and morphophonological issues before writing
rules are determined.
This procedure was originally developed in East Africa, in the Swahili-speaking area,
and therefore the document contains many references to Swahili as the Language of Wider
Communication. However, the procedure can be adapted to other parts of the Bantu-
speaking world, where other languages function as the LWC).
The following section outlines a methodology for developing an orthography for a Bantu
language. Whether the focus of study is one language or a cluster of languages, the
approach will be the same.
vowels
consonants
tones
This study is needed
1. for the discovery of cases of complementary distribution:
• in terms of CV-combinations
• in terms of position of a C in the root structure
2. as necessary background material for the creation of exercises in primers
After establishing its phonemic inventory, the linguist wants to discover the lexical, or
word-internal, morphophonology of a language. Two aspects of morphology require study:
inection and derivation. For Bantu languages, the study of morphophonology, both at
morpheme boundaries within words and in clitic constructions, the main example of which
is the associative construction, is also important.
This information is needed so that sound changes in speech which are caused by phono-
logical processes can be reected well in writing. Most of the results of these word-internal
phonological and tonological processes will be written according to surface forms. The pho-
nological changes which each ax produces and/or undergoes needs to be written up in an
orthography guide (or “writers’ guide”). Once in a while, such phonological processes pro-
duce surface sounds which are not found in the inventory of underlying contrastive sounds.
These need very specic consideration with respect to orthography. Sometimes – especially
75
Chapter 1: Overview of Procedures
in the associative and other clitic constructions – questions on word boundaries need to be
taken into consideration together with the morphophonological sound changes.
Fortunately, Bantu languages are quite transparent in their morphological and mor-
phophonological structure, though there is great variation in the ways they handle them.
The people participating in this research should be made aware of every aspect of the
ndings immediately, as the process of research is going on. They should participate and
give their opinions and insights, some of which will help the research move faster, and pro-
mote learning for all. The approach reects a “learning by doing” philosophy.
Linguistic analysis for this approach begins with word collection, using native intuition
for initial spelling. This is followed by electronic data entry, and then moves to printing
and sorting of word cards based upon phonotactics and target-language speaker perception.
Alphabets are developed and modied as necessary. Finally, participants discover their
grammar together and, guided by Chapter 3 of Part 1 (Bantu Orthography Manual), devise
spelling rules which try to encompass their language’s morphology as well as its sound
system.
Native speaker participation with intuitive writing as the basis for the original
wordlist transcription does not imply that native speakers are asked to make suggestions as
to how they want to write the more linguistically and pedagogically complex phenomena.
The Manual provides a narrow list of grapheme and word break choices, and describes the
issues for each. Participants should, after a good presentation and discussion of the issues
involved, choose between the few options oered them. Testing of readability as well as ac-
ceptability will follow such decisions if questions remain.
76
Each topic is introduced generally, followed by detailed procedural instructions. Each
linguistic feature studied is captioned, i.e., Vowels, Week 1. Below each linguistic feature
caption are the instructions, including documentation of linguistic discoveries in Workbook
Form, for example, “Fill in Table 2, the Vowel Combinations Chart.”
Collect around 1,700 words, nouns, and verbs, as written by speakers of the language.
These will be used in the initial determination of symbols needed in the alphabet, as well
as being a resource for primer making and other language-development tasks. Speakers
of the language will represent the sounds of their language using the symbols of their
language of wider communication (LWC). By looking at inconsistencies in their tran-
scription and at underdierentiation (using a single symbol to represent two dierent
and signicant sounds), we and they gain a good idea of the areas which will require
study. Target-language speakers are likely to underdierentiate sounds—especially vowel
sounds—consistently. Native speaker intuition can be better than IPA transcription when
orthography is the ultimate goal, because the linguist will not know which phonetic dier-
ences are signicant, nor which sounds are the same to the target-language speaker. His/
her phonetic transcription will also yield a less than 100% accurate transcription, but it
will normally be inaccurate in a consistent manner.
The following description of the process has been gleaned from a detailed report written
by Susanne Krüger, based upon the rst series of word collection workshops in Mara region
of Tanzania in 2006.
Constance Kutsch-Lojenga’s original participatory approach did not include wordlist col-
lection via computers. It was simpler, with two goals in mind: native speaker insight and
intuitive orthography development accompanied by linguistic analysis. We have described
the use of Toolbox,
1
a language software tool, for the word collection and documentation
process, in an eort to gain the beginnings of a dictionary for each language, as a re-
source for primer development, and as a means of preserving spelling decisions. Toolbox
and FieldWorks, a similar software package, are able to give input directly to the Bantu
Phonology Tool. A software program other than Toolbox will require some procedural
adjustments.
Word collection will be done using the Africa Area 1,700-item wordlist, in Swahili.
Pitfalls of reliance upon a wordlist: many times the lexicon of the target is not an exact
match for the LWC gloss. People will try to make a t, and the result can be less-commonly
used loanwords, phrases, ambiguities of various sorts, and lots of blanks.
1
Field Linguist’s Toolbox is software that provides integrated data management and analysis for the eld
linguist, developed by SIL International to aid in linguistic data collection and publishing.
77
Chapter 2: Data Collection
Due to this challenge, an introductory activity using a semantic domain approach may
encourage participants to think more freely. For instance, if they have three ways to ex-
press the Swahili gloss in their mother-tongue, they may oer more interesting mother-
tongue vocabulary for the list (and less Swahili), whether or not it matches the Swahili
gloss exactly.
2
Try this with noun categories, such as “furnishings,” “hunting tools,” “musi-
cal instruments,” to get people thinking of their own rich vocabulary.
Whether the collection of vocabulary is done by wordlists or by semantic-domain study,
the speakers of the language should write the words rst of all themselves. This emphasiz-
es local ownership of the project from the start and also helps the linguist because people
will make systematic mistakes, rather than the random ones outside researchers would
make.
The specic vocabulary desired for Bantu wordlists: nouns in their singular and plural
forms (so as to be able to determine the root), and verbs, in their innitive form. Ignore
tone at this point.
Once the rst collection of vocabulary has taken place, one should weed out compound
nouns. Their phonological and morphological structure may exhibit irregular patterns.
They should therefore be studied later, in light of the inventory of sounds established.
All vocabulary collected should be divided into NOUNS, VERBS, and OTHER. In phono-
logical research, one rst needs to focus on the Noun Roots and the Verb Roots, i.e., mono-
morphemic units. It is important to correctly identify the word form of the collected lexical
items in the target languages especially when using a wordlist. The word form of the
translated lexical item might not be the same as the gloss word used for prompting (espe-
cially if the wordlist used is in a language from a dierent language family from the target
language.)
With Bantu languages, the citation form of nouns consists of a root with a prex. Basic
morphological study is needed at the onset, to determine the root structures, which will
form the basis of the phonological research. Early phonological analysis should take into
consideration the transparent forms of the common canonical root structures.
NOUNS: -CVCV (and -VCV and -CV will be treated later)
VERBS: -CVC-a (and -VC-a and -CV-a or -C-a later)
In addition to weeding out compound nouns, set aside derived forms, particularly longer
verbs with derivational extensions and also agentive nouns, often formed with the sux-i,
which may cause morphophonological changes in the root. These forms can be entered in
the electronic data recording system, but should not be printed out onto cards for sorting at
the Orthography I Workshop.
A local community can appoint an MC for the word collection workshop (in Tanzania it
is helpful to have participants elect a chairman and secretary at the beginning of the work-
shop). This person calls the meeting to order, organizes welcomes, devotions, and closing
2
Kutsch-Lojenga actually suggests using semantic domains for noun elicitation, and wordlists for verbs
(“Working document for Phonological Research in BANTU languages,” presented at Bantu Orthography
Working Group in Dallas, November 8, 2005).
78
Bantu Orthography Manual
remarks, introduces people, and generally leads through each day. This role can also be
taken by the word collection team leader.
The word collection team, charged with word collection, can perform the following roles
and duties:
The instructor is responsible to explain the workshop format to the group. He/she ex-
plains the methodology of the word collection to the participants and helps organize the
groups. This person is also responsible for administering a short Swahili spelling test at the
beginning of the workshop.
This job can be done by the team leader. A very uent, natural-sounding speaker of the
language of wider communication is valuable at this point, both for explaining the process
of word collection and for good pronunciation of the dictation test words.
All collected words are typed into Toolbox in order to store them and print out cards for
the orthography workshop. The computer scribe is responsible for entering the data, back-
ing it up every day, and printing out the cards after weeding out all doubles. This process is
time-consuming and the scribe will need the help of others at times.
This person is responsible to liaise with the hosts or the chairman regarding meals,
lodging, time frame, and any other needs or problems that arise. This is an important and
demanding role, taking up a considerable amount of time for the person involved. It should
be a fulltime role for someone on the team. If four groups are anticipated, meaning that
four facilitators are needed, then a fth person is needed primarily to work with the hosts
as contact person, and possibly to ll in for the computer scribe or facilitator.
When the team is pressed for time, participants may be divided into three groups instead
of four. During one word-collection workshop only three groups were formed, and the team
was still able to nish work in good time and spend more social time with the host and the
participants, building relationships.
Each group needs at least one scribe, a speaker of the target language who writes down
the collected words on a blackboard (so all target-language speakers can see and give in-
put) in an orthography as close to Swahili (or another language of wider communication)
as possible.
Scribes may initially be chosen according to their performance in a spelling test given
via the language of wider communication. However, a Swahili spelling test, for example, is
not always a good indicator of a good mother-tongue speller. Several times younger people
have been chosen as scribes because of their faultless performance in the spelling test, but
when called on to spell mother-tongue words, they may take a long time basically needing
to get the words spelled out for them by others. In these cases, the person suggesting the
spellings can be asked to take over after some time (to give the rst spellers a chance to get
79
Chapter 2: Data Collection
used to the job and in order to see if, with some time performance, they would get better).
Generally, the best spellers (and writers) “rise to the surface” within a few hours.
Another role that needs to be taken care of in each group is the role of discussion facili-
tator. This person needs to encourage the group to think creatively and “out of the Swahili
box” in order to get as many usable words as possible. The wordlist should not restrict the
writing down of related words. The facilitator should also make sure that all members of
the group have opportunities to voice their ideas and opinions.
This role was held by the team members. Initially, the idea was that in the course of the
workshop this role could be taken on by a member of the group. In some instances this
happened naturally. However, the team members must keep an eye on the proceedings the
whole time. It is not advisable for a group to keep working without the presence of a word-
collection team member.
The preferred group size is between 3 and 5 participants (as specied by Constance
Kutsch-Lojenga, personal communication).
The following daily schedule is proposed for each workshop:
:
10.00 am Registration, Welcome, Chai
11.00 am Introductions, Devotions
11.30 pm Introduction to workshop
12.00 pm Introduction to word collection, trial run, spelling
test
1.00 pm Lunch
2.30 pm Work in groups
4.00 pm Closing prayer
7.00 pm Dinner
:
8.30 am Devotions
9.00 am Word collection in groups
10.45 am Chai break
11.15 am Word collection in groups
12.45 pm Prayer time
1.00 pm Lunch
2.30 pm Word collection in groups
80
Bantu Orthography Manual
4.00 pm Closing prayer
6 or 7.00 pm Dinner for those not going home at night
The word-collection team meets every day after closing for debrief, typing in of words
and preparation for next day.
Sessions longer than about 1.5 hours are very hard for the participants and some break-
ing up of the work is helpful. The early nish each day allows those participants who stay
at home throughout the workshop to reach home in good time and allows the team to type
in words and prepare for the next day.
It should be possible to elicit 2,000 words in a four-day workshop using three groups.
Word-collection teams should begin with a semantic domain approach before attempting
to collect words via the wordlist. People nd it very hard to change from translating words
from a list to freely associating words. The wordlists, on the other hand, give more struc-
ture to the process, a sense of progress and completion.
When the wordlist process begins, each group is given part of the BD Standard 1,700-
item wordlist (a current soft copy is available upon request from the Bantu Department,
SIL). The group facilitator reads the Swahili word and the group comes up with a MT
translation and writes this down (on the blackboard). The facilitator then copies the word
exactly from the blackboard onto the blank wordlist. Participants may give more than one
word if there are multiple ways of translating the LWC word.
Information noted for every word is:
• Gloss in MT (Sg/Pl for nouns, Innitive for verbs)
• Swahili (or LWC) gloss
• Noun, verb, other
At the end of each workshop a “clean-up session” is held. During this session, a member
of the group (not the scribe) reads the mother-tongue words out loud to the whole group
and the group gives a translation into Swahili. During this time all the necessary infor-
mation on spelling, word break and elision or hyphenation have been noted. Corrections
supported by the whole group are marked and the nal version typed into Toolbox. It is
possible to use a printer (powered by the car through an adapter), printing only the mother-
tongue glosses from Toolbox. This list is easier to read than a handwritten one, and a print-
out of the LWC gloss is not helpful or necessary.
Any observations the word-collection team can make during the word collection about
correspondence between phonetics and the symbols used by the transcribers will be help-
ful. The team should be especially observant during the reread—when reading their own
writing, are people struggling with issues of vowel length, vowel quality, tone, etc? How of-
ten do they read a word with correct segments but the wrong tone, wrong vowel quality, or
wrong vowel length? What, if any, seems to be the “default” tone pattern people use when
guessing at the pronunciation of a word they do not recognize?
81
Chapter 2: Data Collection
In order to facilitate the collection of the 1,700-item wordlist and the subsequent creation
of a sorting card for each word, a Toolbox database and project have been created together
with wordlist documents.
During the word collection workshop, speakers of a language can record their words on
a printed wordlist form. A document of the 1,700-item wordlist with Swahili is available
(1700 swa-blank 0701).
Following the transcription of the words on the printed wordlist forms, computer scribes
can record the words in the Toolbox database. In the Toolbox Word Collection project, there
is a database of the 1,700-item wordlist. Each record contains various elds already lled
in (English gloss, Swahili gloss, wordlist index number, and semantic domain index num-
ber). The lexeme eld is blank and this is where the vernacular word is keyboarded.
Before beginning data entry, one needs to create an “empty” wordlist database for each
language. In the project, select the master wordlist database “Master Wordlist.db.” Select
the lter “001700List.” Then go to File/Export and select “Blank Wordlist Bantu lg.” The
lter should be set to the current database lter. Create a database for each language being
processed and change the lename appropriately.
At the end of a session, when the vernacular data have been veried and transcribed, the
Toolbox database can be lled in. If the database is sorted by the wordlist index number (\
is-1700), the user can proceed through the wordlist as it is found on the printed handouts.
For noun words, the singular is placed in the lexeme (\lx) eld and the plural form in
the plural eld (\pl). If there is only a plural form, this should be placed in the lexeme eld
and the plural marker removed.
For verbs, the “root” form should be place in the lexeme eld and the innitive form
placed in the vernacular paradigm eld (\pdv) following the innitive paradigm label (\pdl
inf) eld. An option is to place the innitive in the lexeme eld and ignore the paradigm
elds.
If the data in a record does not reect the language, it can be edited. For example, if the
Swahili gloss needs to be changed or modied, the text in the Swahili gloss eld (\gswa
OR \gr) can be edited.
For each record that will be useable in the orthography sorting process (nouns and verbs
only), the Card (\card) eld should contain a “yes.” If the word is not useable and does not
need to be printed on a card, the Card eld should contain a “no.”
If multiple people are entering data, each can have a separate database and the databases
can be merged at the end of the data entry session. The updated database can then be dis-
tributed to each computer scribe (for Toolbox v.1.5.1 and above).
82
Bantu Orthography Manual
When all the data has been keyboarded, the sorting cards can be printed. There are
lters in the Word Collection project for Nouns, Verbs, and Other. A Word document is cre-
ated. This can be reviewed and edited as needed before printing the cards.
To print Noun cards:
1. Set the lter to “00 ps nouns Cards”
2. Go to File/Export, choose “wordlist cards” export name. The lter should be
set to “Current Window’s records.”
To print Verb cards:
1. Set the lter to “00 ps verbs Cards”
2. Go to File/Export, choose “wordlist cards” export name. The lter should be set to
“Current Window’s records.”
To print other cards:
1. Set the lter to “00 ps other Cards”
2. Go to File/Export, choose “wordlist cards” export name. The lter should be set to
“Current Window’s records.”
Other printouts can be created in the Toolbox project.
File/export/Multi-Dictionary Formatter Standard dictionary type printout
Browse view List of elds
The following can be created from the Master Wordlist.db:
File/export/List English-Swahili (RTF) List with English and Swahili
File/export/List Swahili-blank (RTF) List with Swahili and (blank)
A limited number of word roots in Bantu languages. Very many words are semantically
related (they have the same root in noun and verb or other parts of speech) which results
in a high number of words with the same root but dierent morphological forms.
The diculty of changing approaches mid-way, between wordlists and semantic domain
elicitation
• The strong inuence of Swahili (LWC) in some areas
• Participants may give word-for-word translations of compound words in the LWC,
only coming back later with a traditional mother-tongue word
83
Chapter 2: Data Collection
• Participants may be thinking, since the LWC is used to conduct the workshop, that
it is the language of the workshop, so they see their task as translating, rather than
thinking and responding in mother-tongue. Krüger recommends more elicitation by
semantic domains, especially early in the process.
• Many of the participants speak Swahili very well and may nd it dicult to
remember the traditional words of their mother-tongue, once the more common
words are covered. People may become overly focused upon translating the exact
meaning of the word given in the list and waste time discussing semantic issues—a
worthy activity, but not for this workshop.
If people begin with lists, they will feel they have been exhaustive and that their work is
done, while there are still thousands of words which haven’t been elicited.
Mark passages in the wordlists where additional words can easily be
obtained (types of trees, animals, illnesses, or certain activities like working elds, build-
ing a house, or cooking special traditional food). Begin with one of these sections in the
wordlist. Ask participants to come up with additional words, especially at these marked
points. Participants will then be accustomed to the idea of freely giving any words they can
think of, using semantic domains.
Once collection of vocabulary has taken place:
Their phonological and morphological structure may exhibit irregular patterns, so they
should be studied later, after the inventory of sounds is established. Enter them all in the
electronic wordlist, but don’t use them for card-sorting. If the linguist chooses to provide
parts of speech for the words in the “other” category, this is ne, but for purposes of card
sorting for phonological analysis, they will all be placed in the “other” category.
It can be dicult at this early stage to recognize loanwords easily. Especially in
Tanzania, many words will be found that are identical with Swahili words. That does not
necessarily mean that they are loanwords from Swahili; they might simply be Bantu roots.
Sometimes it is possible to ask the participants whether the item in question is a mother-
tongue word or borrowed from Swahili. Another hint is when certain sounds (e.g., [l], [f],
[D], [T]) only occur in words known from Swahili. In all probability these will be loan-
words. When in doubt it is better to as they will quickly
become obvious throughout the phonological analysis at the workshop.
If the loanwords have already undergone phonetic assimilation to the system of the
mother-tongue, they should be left in the wordlist. Also, if there is no mother-tongue word
in the language to express the same concept, this indicates that the loanword will need to
be expressed in mother-tongue writing, whether it is borrowed or not. Save these words for
future reference but they will not be printed on cards.
In addition to weeding out compound nouns and loanwords, one should, at the initial
stage, also set aside derived forms, particularly longer verbs with derivational extensions
and also the agentive nouns, often formed with the sux -i, which may cause morphopho-
nological changes.
84
Bantu Orthography Manual
This means that this early phonological analysis needs to be carried out on a subset of
the vocabulary collected. All other words will need to be checked later, to see if their
phonological structure ts into the same inventory discovered.
Be sure, if several languages will be studied at the same time and place, to use dier-
ent colored cards for each language’s wordlist! As indicated in the software, NOUNS and
VERBS will be printed out on cards. LOAN and OTHER (parts of speech other than nouns
and verbs, as well as compounds) words will be printed out on paper only, for reference
(checking for any additional combinations of letters which may not occur in nouns or
verbs). If these occur, they will probably indicate labialization or palatalization of a simple
consonant already occurring elsewhere.
Cards will always be segregated by these categories. Use labeled plastic boxes or zipped
heavy duty plastic bags to keep them separate throughout Orthography I. In phonological
research, one rst needs to focus on the Noun Roots and the Verb Roots, i.e., monomor-
phemic units. Remind participants that we will never want to mix cards from these three
categories (unless a gloss is changed and a word is re-categorized). You may also want to
give participants a quick review of parts of speech.
Using the Bantu Phonology Tool (the Phonology Template Editor and Search Tool), gener-
ate a phonology report (PAW) to produce an inventory of all the sounds and letters of the
language as it is currently written.
Print it out, as a reference to be used when checking the inventory of phonemes of the
language (to ensure you haven’t missed anything) and when lling charts.
Bring these supplies to the Orthography I workshop:
• 2 sturdy containers per language group, labeled “Noun” and “Verb”
• 2 boxes of jumbo paperclips per group
• 1 box of big rubber bands per group
• large tablets of chart paper for all to use
• 1 or more boxes of felt pens for all
• 1 large chalkboard plus a small one for each group, with chalk and erasers
• notepaper
• post-it notes, small
While electronic record keeping is ecient and long lasting, laptops should not be visible
at the early stages of the workshops. Rather, the “cards” and the target-language speakers
should be front-and-center at this point, allowing the target-language speakers’ expertise to
be the focus. This also allows room for much more insight for speakers of the target lan-
guage, as a result of the sorting process and a more global beginning to a very analytical
process. This means that all changes, either to glosses or the spellings of words, must be
immediately written on the cards. As soon as these changes are recorded electronically, a
85
Chapter 2: Data Collection
tick on the card with a colored marker can indicate this has taken place, and a #1 can be
typed in the “card” eld for a given word in the database. Later on, if a second revision is
made, the 1 can be replaced with a 2.
Pre-sorting of cards before daily activities will be a must for the rst week or so, due
to morphophonological complexities aecting prex vs. root, and vowel vs. semivowel
identication.
The cards are a visual and tactile learning tool. They can be manipulated. Various charts
as seen in the worksheet section of the manual can be copied onto large chart paper and
placed on a table. Leave card-sized spaces on the chart, where cards can be placed by
participants.
Wrapping up, a procedural note: if you can get all your noun and verb entries corrected
in the Toolbox database about two days before everyone leaves, you can then check the list
on your laptop, with the participants watching/assisting. Side benet: the experts are there
with the linguist to not only correct spellings or clarify pronunciation, but to revise errone-
ous glosses. They learn a lot from the experience, too.
It may be worthwhile for the TAs to take the time before the workshop be-
gins, to have their cards sorted so that they have at least all the -CVCV roots sorted out, so
that the participants don’t have to deal with so many variables that rst day. And if they
have the cards printed up long enough in advance, the other cards can be sorted at leisure
over a period of days.
86
At this point, you are ready to begin phonological analysis. The analysis consists of four
parts:
1. consistency checks on what people have written
2. interpretation of the sounds
3. discovery of cases of complementary distribution (through phonotactic research)
4. nal inventory of contrastive sounds
The rst week will focus on the vowel system.
Read Part 1, chapter 2, “Linguistic Features,” Vowel section.
Presort for all the C-initial (CVCV) noun roots (V-initial roots will come later). In a Bantu
language, nding the roots involves looking beyond the initial noun class prex. This pre-
x is usually (V)CV-, but sometimes just (V)- or (V)N-. See Part 2, Appendix J for helpful
examples.
Most people who have had schooling in Swahili (and perhaps also via other languages)
will initially write their own languages with the ve well-known vowel symbols. As stated
in the Linguistic Features chapter of Part 1, the language may have ve, seven, or nine
vowels, or, very rarely, eight!
The general procedure is as follows: Once all -CVCV noun roots and -CVC- verb roots
have been identied,
1
the rst step is to sort the words by the rst vowel, and then check
for “same” and “dierent” sounds, listening to only one vowel symbol at a time. See if the
vowels as written on the cards have been written consistently. Is everything the target-
language speakers have written as <> the same sound; or is it a mixture of two sounds?
(e.g., [i] and [ɪ]). Is all they have written as <> really one and the same sound? Or does
their written <> represent two sounds: e.g., [e] and [ɛ], or [e] and [ɪ]?
This consistency checking procedure should be done separately for nouns and verbs.
First of all, noun roots have generally two vowels, whereas verb roots have only one vowel.
Second, doing it this way would immediately show any systematic dierences in vowel dis-
tribution between nouns and verbs. In fact, a language was once found in which the nomi-
nal system had an inventory of seven vowels, whereas the verbs displayed an inventory of
nine!
In a language where prex vowels are not inuenced by harmony with the roots, it is
good to begin analysis with an inventory of noun class prexes, since the participants must
be able to recognize them in order to nd the roots. However, in some languages, the qual-
ity of the vowel in some of the prexes is determined by the quality of the vowel in the
1
The sorting out of the CVCV nouns and CVC- verbs would really be better done ahead of time by the lin-
guists, I think. The participants won’t be ready for deciding which nouns are too long and especially decid-
ing about which are causative or passive verbs (Karen Van Otterloo, p.c.).
87
Chapter 3: Orthography Workshop 1:Phonological Analysis
root. The quality of vowel used in the transcription of the prex may actually be a good
clue as to the quality of vowel found in the root.
Also, in languages in which Dahl’s Law is active, the consonant of the prex may also
change from voiceless to voiced when the root begins with a voiceless C. Thus, the prexes
may have more than one form, even before C-initial roots. TAs should thoroughly familiar-
ize themselves beforehand not only with typical forms of each noun class marker, but also
with the phenomena represented within the language group which they will be assisting.
The proposed sequence of events is as follows:
Lexical ROOTS of nouns are the essential elements for the discovery of the sound sys-
tem: contrastive vowels and consonants and their allophones, as well as the tonal melodies
inherently belonging to the lexical roots.
Pre-sort nouns, looking only for those with -CVCV roots. TAs should reread
the Vowel Systems section, chapter 2. They should also look at Noun Class Prexes in
Appendix J.
Comparing the singular form with the plural is very helpful for root identication. Set
aside words with a prex ending in y or w, because the presence of a semi-vowel in the
prex is an indicator of a V-initial root. Example: in the cl. 7 Kifuliiru /ikitala/ ‘drying
platform’ there is a C-initial root: /ikitala/. Contrastively, /ikyusi/ ‘smoke’, has a V-initial
root: /iki-usi/. This causes the prex to be realized as /iky-/ instead of /iki-/.
Since Bantu languages normally have a noun-class system and the
citation forms of nouns and verbs consist of a root with its prex, it is important to rst
establish the basic forms of the noun-class prexes as they occur preceding the noun and
verb roots (see Appendix J for examples). At this stage, try to identify noun-class prexes
as they occur before a C-initial root, collecting all cards for nouns with the following root
structure:
-CVCV (and later, -VCV and -CV)
At a later stage, their morphophonological variants will be identied, as they
occur before V-initial roots (see Consonants Step 7). It may be helpful to the TA to pri-
vately
2
pencil in the tentative forms for each of the noun class prexes on the rst of the
morphology charts, , in Appendix J.
In order to identify the noun-class prexes in any typical Bantu language, use the widely
accepted numbering system for reference when lling them in on the Workbook Form,
Appendix L. See Appendix J for helpful examples and give various alternants as possibili-
ties, so people may recognize the prexes in their own languages. The importance of this
initial morphological research is to establish the roots, which should be underlined at this
stage, on the cards.
2
Avoid distracting participants from phonological analysis, at this point. Their noun classes prexes could
have many variants, and these should not be viewed as nal yet. Much depends upon the particular vowel
system they have, and whether the prexes contain long vowels, and so forth. At the end of the second
week, you’ll have enough information to complete the prex charting, together with the participants.
88
Bantu Orthography Manual
Pre-sort nouns with -CVCV roots in which V1 = V2 . This means sorting the noun
data into consistent combinations of vowels: One pile for -CiCi, another for -CeCe, -CaCa,
-CoCo, -CuCu, and so on.
A speaker of the language should read aloud through each pile, and all
should listen to determine whether or not the vowels that are written the same all have the
same quality.
If a pile contains two dierent vowel qualities (though both written using the same
symbol), they should be sorted separately, and a second reading should show consistency in
each set. We begin with <a> because this tends to be the most “neutral” vowel in a Bantu
language, the least likely to pose any problems. In this way, participants can get used to
the process without listening hard for allophonic variation or phonemic contrast.
Thus, we take a pile which contains words like eghesara, eghetanda, endanya, and omo-
hara and have someone in the group give the gloss and read the word while the others
listen. For each reading, the group must decide:
• whether both vowels in that root sound the same (e.g., does the rst <> in
<> sound the same as the second <>?
• whether the vowels in this word sound the same as the last word they read, i.e.,
are the <> sounds heard in <> the same as the sounds they heard in
<>, or are they dierent?
Once the pile of < nouns has been sorted in this way, go on to the <>
nouns and the <nouns, because the high vowels are easier to hear than the mid-
vowels. Then move on to the others. Finally, preserve all of your piles of nouns, with rub-
ber bands and post-it note labels.
Keep these stacks clipped and labeled for future reference. After sorting out
all the CaCa, CiCi, CoCo patterns, the OWL knows (or at least has agood rst guess at)how
many distinct vowels are contained within the data. If the stacks sort into seven piles,
there are probably seven vowels in the language, for example.If there are more than ve, a
temporary symbol should be chosen for the extra vowel, for use on the vowel chart for the
next step.
Make a chart as shown below, taking into account the vowels of the language.
Simply make columns for each vowel of the language across the length of the paper, and
rows for each vowel in the second position (with blanks for the identical V slots). Make the
boxes big enough to hold the stacks of cards. A partial example, below, for a hypothetical
7-vowel language: Pre-sort nouns with -CVCV roots in which V1≠V2.
Presort the nouns. Take the nouns with di-syllabic roots and sort them according to
vowel combinations, such as rst syllable: // and second syllable //, rst syllable // and
second syllable //, for each vowel in the language.
89
Chapter 3: Orthography Workshop 1:Phonological Analysis
This will show the basic vowel distribution patterns in lexical roots for the language
(distribution in the roots cannot be observed in verbs, since basic verb roots only have one
vowel). One wants to know the possible combinations of the 5, 7, or 9 vowels, and the com-
binatory restrictions.
Lay the chart paper across a table. Participants read presorted words
aloud and place them in their places.
V2 V1 a ɛ e i ɔ o u
a CaCɛ CeCa CiCa CɔCa CoCa CuCa
ɛ CaCɛ CeCɛ CiCɛ CɔCɛ CoCɛ CuCɛ
e CaCe CɛCe CiCe CɔCe CoCe CuCe
i CaCi CɛCi CeCi CɔCi CoCi CuCi
ɔ CaCɔ CɛCɔ CeCɔ CiCɔ CoCɔ CuCɔ
o CaCo CɛCo CeCo CiCo CɔCo CuCo
u CaCu CɛCu CeCu CiCu CɔCu CoCu
Once again, each set of data should be checked for “same” and “dierent” rst, before
deciding on the vowel qualities of each.
In a 5V system, sorting like this leads to a total of 25 combinations (5x5), of which 5
were dealt with in the rst part, those in which V1 = V2. So a maximum of 20 combina-
tions are left to check.
In a system of 7 vowels, this would lead to 49 combinations (7x7), of which 7 were dealt
with in the rst session, so a maximum of 42 combinations left. Most likely, not all 42 com-
binations are found. This study will show if a system of ATR vowel harmony is present in
the language (static = root-internal). See Part 1, chapter 2, Vowel and Vowel Harmony sec-
tions. Look for combinatory restrictions on the vowel combinations in the roots. If found,
the study of ATR vowel harmony will later have to be extended to the dynamic or mor-
phophonological part, namely the behaviour of axes in combination with roots.
A system with 9 vowels would lead to 81 possible combinations (9x9). However, it is al-
most certain that such a system displays ATR vowel harmony, and that one can divide the 9
vowels into two subsets of 5 (taking the vowel /a/, inherently [-ATR] also into the [+ATR]
set in order to check its occurrence in roots in combination with [+ATR] vowels). This
would add up to a total of 25 and 24 possible vowel combinations in roots, of which 9 were
already dealt with in the rst session, so a maximum of 40 more combinations of vowels in
roots are possible.
This study will probably conrm the initial hypothesis of the vowel inventory, if the
number of vowel combinations stays within the original set. If not, one needs to do further
investigation.
If the number of vowels is conrmed, the group should make immediate grapheme
choices, so everyone can get used to the orthographic vowel symbols and start correcting
the writing on the cards to reect their vowel system.
90
Bantu Orthography Manual
The following has been proven most useful in languages in eastern Africa, and responds
well to native-speaker intuition (for further discussion, see Part 1, chapter 2, Linguistic
Features, Vowel Systems section).
5V orthography: <>
7V orthography: either <> or <>
9V orthography: <>
Go to Appendix L, Workbook Form. Fill in Table 2
chart, as well as Table 3, Table 4
and Table 5
Sort verbs with -CVC- a roots/stems according their root vowel.
The innitive of a verb with a -CVC root will have a form like /oko-beb-a/. Sort by the
rst vowel of the root, as the nal vowel in innitive forms is inectional rather than
part of the root, and is generally /-a/. Each set of items with a particular vowel should be
checked for consistency, by people reading and listening. Again, if a pile is divided into
two, a second reading should show consistency in each set.
1. Discover the lexical forms of the verbal roots, and underline them.
2. Look for /(V)-ku-/ prexes, followed by -CVC-a. These are the typical canonical
verb pattern. Later, study -VC-a and any other forms.
3. If you nd instead /(V)-kw-/ followed by a vowel, put those in a dierent pile for
later, as they contain V-initial roots.
4. Go through each pile of these verbs and decide which pile of nouns the rst vowel
of the root in each one sounds like. Try to match a pile of verbs to each pile of
nouns, but without mixing them together.
should be lled in at this point, with target-lan-
guage speakers giving glossed examples for each, though nal symbols may not yet have
been chosen.
TAs should review Part 1, chapter 2, Vowels, Vowel Length section.
Go through the cards again with the group, listening for vowel length
on the roots, both contrastive and compensatory or conditioned.
You will likely nd both vowels which have underlying (non-conditioned) length and
those which are long due to their environment (i.e., they follow a semivowel or precede a
prenasalized stop). Vowels with conditioned length cannot be pronounced short by a na-
tive speaker. The lengthening is automatic. Thus, the length of the vowel is not contrastive
(cannot make a dierence in meaning).
91
Chapter 3: Orthography Workshop 1:Phonological Analysis
You will probably nd words with vowel length which cannot be explained by such
environmental conditioning. If this is the case, you will also nd words with short vowels,
in a phonetically similar environment. This will conrm the presence of contrastive vowel
length in the language. Looking through the /a/s for instance, when you nd a word with a
contrastively long vowel, pull out this card and mark the spelling change on the card (e.g.,
mugaata instead of mugata). Make a separate pile (one pile for each vowel) of the words
which have a long vowel. Move on to another vowel after completing the rst and separate
the long /i/s, /e/s, and other vowels in similar fashion. Native speaker perception of under-
lying length will sharpen as the process goes on.
Note, on a piece of paper (and on a small chalkboard) any minimal pairs (or near mini-
mal pairs) for length. Target-language speakers should probably do the writing on the
chalkboard.
Fill in . Also ll
in . Include a gloss for each word of each pair.
Save all of your stacks of cards, sorted by syllable pattern and part of speech, for use
next week.
Before you begin this week, read Part 1, chapter 2, Linguistic Features, Consonants
section. This will help you know what challenges to expect, and it gives recommendations
for grapheme choices. The mini-lecture for TAs only, Part 2, Appendix F, Lesson 5: Lesson
for TAs, with consonant spelling issues, should be given before this week begins.
The aim of the week is to establish the inventory of contrastive consonants and to study
their distribution within -CVCV roots. There are a larger number of consonants, and much
of the week will be spent on alphabet matters.
The same principles and approach previously used with vowels apply now to consonants:
• Nouns and Verbs will be looked at separately
• Starting point are initial consonants of the consonant-initial roots.
• The data should be sorted accordingly, and one rst listens for “same” or
“dierent,” then determines the actual phonetic quality of the consonants.
The approach is like that used for the vowels, except that with consonants it is not neces-
sary to look for words in which C1=C2. We simply look rst at each consonant symbol that
has been used in the C1 position of the root and sort words by the C1 symbol. Sort on a
large chart (see details below) as before, using only nouns the rst time and later doing the
same with verbs. Since we have kept the noun and verb cards completely separate, this will
be simple.
Fold a large, simple chart-paper grid, for use with each language group. Each box
on the grid should be big enough to hold a stack of cards and still have the label show in
the corner.
92
Bantu Orthography Manual
Print out the PAW
3
(Phonological Analysis Worksheet or “Bantu Phonology Report”),
which lists all segments and where they occur, according to the intuitive transcription
documented in word-collection.
Fill in the boxes on the chart paper with every letter or combination of letters identied
by the PAW. This should speed the learning process for the day, and ensure that no con-
sonants, digraphs, or consonant clusters are overlooked. For example, without this pre-
marking of all consonants, no one will notice that the <mb> and <ng> boxes are empty
during the C1 sort (so I changed that part). But they will notice that the NC boxes need to
be added when sorting by C2, later.
Large group activity: Part 2, Appendix C, Lesson 2: Helping Readers.
Comparing the PAW to the boxes at the end of this exercise will ensure that no conso-
nant has been overlooked accidentally from the phonemic inventory.
Note before sorting: morpheme breaks and syllable breaks many times don’t coincide in
Bantu languages! A nasal can be a morpheme in its own right, for example, but when it
occurs before a consonant, the nasal and consonant join to form a cluster, i.e., mb, nd, or
ng. In nouns, the Cl.9 prex usually shows up as the nasal in such a cluster. Thus, in a word
like embeho, the e-m- represents the prex, while the stem is -beho, and NOT -mbeho. The
root-initial consonant is b, and not mb.
.
Because NC combinations do not appear root initially, during this sort on the initial-C, it
is most likely that NC boxes will not appear on the chart (or if boxes are prelabeled with all
the segments found in PAW, will be empty). Leave room on the chart to add them in when
sorting is done by the second C. This should make it clear to all that NC segments are gen-
erally not found in root-initial position!
Go through each pile of disyllabic nouns (already sorted and paper-
clipped) by the rst vowel of the root. Disregard the prex, of course. Noun=prex+CVCV
root. As your group begins to ll the boxes on the chart with cards sorted by initial conso-
nant, write each consonant in the box as it is found (unless boxes have been prelabeled as
in step 3 above). The resulting stacks will give speakers a literal picture of the frequency of
occurrence of each phoneme in their language. The grid may look something like this (with
labels in one of the upper corners, so that they still show when the cards have been placed
in the boxes), with stacks of words on top of them, sorted according to the vowel which fol-
lows the C1:
b bh c mb t
d dw dy r ry
rw r f …
The cards being sorted for initial C are already sorted by the rst vowel, so when you
nish sorting the rst pile, the pile in which V1 is /a/, for example, all the cards on your
3
The PAW is part of the Bantu Phonology Tool, a piece of software which analyzes the phonology of a
language and produces a report. It can be found on the SIL website:
http://www.sil.org/computing/catalog/show_software.asp?id=116.
93
Chapter 3: Orthography Workshop 1:Phonological Analysis
consonant grid will be Ca- initial words. When you nish sorting each dierent pile for
C1, paperclip each pile, so that you can keep the ba- initial roots and ka- initial roots, etc.,
separate from the bi- initial roots and ki- initial roots, and so on. Leave each paperclipped
stack in its square on your chart paper consonant grid. When you have nished going
through all the CVCV noun cards, bundle all the cards on each square together with a rub-
ber band and keep them on the chart paper. Keeping them sorted by rst vowel as well as
by rst consonant allows easy evaluation of whether or not there are signicant gaps in
the phonotactic distribution—e.g., does a certain consonant sound occur only preceding a
certain vowel or set of vowels?
Next, go through all the words in each stack on the grid and listen to them together to
make sure that all the Cs which are written the same also sound the same.
When you nish going through all the CVCV root cards and listening to them to check
for consistency, you will have an inventory of C1 position consonants. The consonants
should immediately be lled in, in a chart with the main points of articulation.
Examples of each C1 followed by each vowel can now be written down in the
Workbook Form chart entitled . Record these examples before
moving on to the next step, which involves re-sorting the same cards by the second
consonant in the root.
A major principle in phonological analysis is “compare what is comparable.” In keeping
with this principle, the consonant inventory can only be determined by listening to words
containing the same consonant in the same position. This is why the second step is neces-
sary: establishing the inventory of consonants found in C
2
position.
Use the same chart paper grid you used for sorting root-initial conso-
nants, but this time to help participants sort the words by the word-medial, or C
2
conso-
nants. Since the words are not presorted by V
2
, after sorting the words by the C
2
consonant,
go through each pile and sort the words according to what vowel follows the C2 consonant.
For instance, in your stack of words in which C2 is r, paperclip together words like
ekekere, enkore, and ekehore in one pile, words like ekemori, insuri, and emboriin an-
other, and inshara and obhosara in yet another pile. Then read through, making sure that
all the Cs which are written with the same symbol sound the same to the ear of the target-
language speaker.
These consonants will be listed on the Workbook Form chart called
There is likely to be a greater number of consonants in this
root-medial position, because prenasalized consonants are now included (since the nasal
part is part of the root and not part of a prex). Also, keep track of all labialized and
palatalized Cs, so you can list the possibilities at the bottom of the chart.
The following steps should now be taken:
1. Check for specic gaps in the CV-distribution (i.e., the set of vowels that may
follow each consonant) and see if these gaps would lead to statements of
complementary distribution of specic consonants. (The distribution of [l] and
94
Bantu Orthography Manual
[r] should be checked separately, since their complementary distribution is often
determined by the preceding vowel.)
2. Check if there is any complementary distribution based on position within the
root, C1 or C2, e.g., [d] only ever occurring as C1 when preceded by a nasal
prex, and [r] only ever occurring as C2 in the root. Such a distribution would
suggest that the two are in complementary distribution.
3. Record examples of restricted occurrence or complementary distribution of
consonants in , at the bottom of the chart (and someone can record
them for the group on a chalkboard). These will prove very useful to literacy
personnel, later.
4. Replace the sorted noun cards in their container.
Help participants to see how restricted occurrences demonstrate allophonic variation. For
example, “soft voiced” consonants (fricatives) never will occur after a nasal. They only oc-
cur between vowels, while their allophonic variants, voiced stops, will always appear when
they are preceded by a nasal. These can often be shown to the group by alternations in the
root initial consonant of a class 9 noun when its nasal prex is replaced by a (V)CV- prex
in a diminutive or augmentative form. For example, the diminutive form of <>
‘goat’ is <> ‘little goat’ in Kifuliiru. This shows alternation between b and h,
conditioned by the nasal.
TAs should reread theConsonants section, chapter 2 of Part 1.
Now, take the CVC-a verbs, which were sorted earlier according to their
V and follow the same consonant discovery procedure as for nouns, using the same con-
sonant grid chart paper.
Sort rst for C1, paperclipping the stacks on each box of the grid
together to keep the Ca-initial verbs separate from the Ci-initial verbs, and so on.
When sorting for C1 is nished, do a readthrough. People will be listening to be sure that
every C they have written has the same sound as the others with that spelling. The read-
through is done to conrm that the sounds represented by a certain symbol are actually all
the same phoneme. Just paperclip them by vowel and rubber band each consonant group
on the grid. Label as needed with mini post-its.
(with group)If any additional Cs are found after the C1 sorting of verbs
(unlikely), add them to the chart of , before moving on to medial-
position consonants and make a tentative decision as to how they will be written, so any
spelling changes can be noted on the cards.
Give Part 2, Appendix D, Lesson 3: Consonant Symbol Choices (if
not given previously)
Examine C2 verb roots. You may use the same chart of consonant sym-
bols as the day before.
95
Chapter 3: Orthography Workshop 1:Phonological Analysis
When the group has nished sorting all the verbs by the consonant in C2 position, read
through each stack listening to the consonants sorted by the consonant in second position,
to verify that all of those written the same actually sound the same to the target-language
speaker. Again, if there are any further consonants discovered, add them to your
4
Also note any further co-occurrence restrictions there.
Check your PAW once again just to ensure that no letters have been overlooked. List
all the letters, including labialized and palatalized consonants, on scratch paper or chalk
board.
It might be helpful to circle the labialized and palatalized consonants while listing them,
because they will not be entered on your list of consonant phonemes. You will list them
on the Labialized Consonants and Palatalized Consonantssections of the workbook form.
These are consonant clusters, and as such will not be represented in an alphabet chart, but
the skill of reading them will be taught later in literacy materials. The and , “conso-
nants” in their own right, will probably be represented in any Bantu alphabet, because they
will occur in isolation, especially in verbal axation.
Study “Alphabet Chart Considerations” below.
Study the list of words from the “Other” category. Examine shorter and longer nouns
and verbs and words from other categories (pronouns, adjectives, adverbs, numerals, and
demonstratives, for example) to see if the vowels and consonants in these words can all be
written with the inventory of alphabet letters decided upon. Full analysis of morphophono-
logical issues will be delayed until the morphology workshop.
List any new consonants or clusters which you may (rarely) have found.
Fill in the inventory of contrastive consonants,
. Target-language speakers will enjoy choosing examples and
giving glosses.
Fill in the charts for , both word-initial and word-nal, in
the workbook form. Target-language speakers do not need to discuss these. Word cards are
a quick reference.
In distinguishing true vowels from consonants, study the following examples:
In the cl. 7 Kifuliiru drying platform there is a C-initial root:
. In smoke, there is instead a V-initial root: , which
causes the prex to be realized as instead of -. It may take a
bit of practice with the language to tell for sure what are the C-initial
roots, since some roots may not be immediately distinguishable: e.g.,
in Ngoreme, there are words such as well (of water), which
4
There is no need to sort by the second vowel here because these are CVC verb roots and the following V
will always be -a (as long as the TA’s have sorted out the passives and causatives and such.
96
Bantu Orthography Manual
apparently has a prex -, presumable cl. 7. However, there are also
words such as bellows which could be a cl. 9 or cl. 5 noun with
the augment - as its only prex (-), or it could be a class 7 noun
with an o-initial root (). To determine if these two nouns are
of the same class or a dierent class, singular form, while in other lan-
guages (probably including those in the Mara cluster) may have a dif-
ferent prex. Other classes should always have a dierent prex in the
plural. In the case of these two words, we see that has a plural
of , showing that it is indeed a vcl. 7/8 noun, whereas the plural
of is , showing that it is a cl. 9/10 noun. So once you
know the class in each case, you can see that both of these examples have
C-initial roots (Karen van Otterloo, pc). If such a questionable noun has
no singular/plural, ask the speaker to add a modier such as an associa-
tive phrase ( of others or something similar, and the class will
become apparent).
In a language where prex vowels are not inuenced by harmony with
the roots, begin with an inventory of noun class prexes, since the par-
ticipants must be able to recognize them by now, having underlined roots.
In at least some of the Mara languages of Tanzania, the vowel quality of
some prexes is determined by the quality of the vowel in the root. Thus,
the prexes may have more than one form, even before C-initial roots.
Take time in the seven-vowel languages to look at the quality of the vowels in the prex-
es and to see how these correlate with the quality of the vowel in the root. This will allow
participants to become more aware of what vowel harmony is doing in their language.
Sort all V-initial noun and verb cards separately. List VV sequences as
you go, e.g., ai, eo). A speaker of the language can write the list, using a chalkboard or
paper.
Look for the shorter roots rst: the –VCV roots. Go through the words looking for CyVCV
and CwVCV. Further sort these by the quality of the initial vowel.
Record your preliminary observations on chart.
Gather stacks of nouns, sorted by noun class. Review “Phonological conditioning at
morpheme boundaries” in Part 1, chapter 2, Morphophonological Processes.
Give participants a basic presentation on noun classes, tying it in
with Swahili and whatever they have learned about noun classes in school. (Optional: make
a printout for each participant, showing the standard numbering system along with the
sample prexes of Proto-Bantu and some existing Bantu languages. See Appendix J, Noun-
Class Prexes for Proto-Bantu (Gardner 2005:54).
If this was not done earlier (during root identication), Participants
sort through the -CVCV root nouns in their groups, nding words which correspond to
the classes they know in Swahili. Go through their cards with them, helping them match
up any prexes not found in Swahili with one of the classes in the numbered list. As they
gure these out, consult the lists of both standard and nonstandard noun class pairings
97
Chapter 3: Orthography Workshop 1:Phonological Analysis
(telling which singular goes with which plural). Record the prexes for each noun class on
Morphology Charts: ) and the form in
which they are found when they precede any vowel initial roots, if you’ve learned this.
Bring a laptop which has Publisher capabilities and be prepared to use the CD Art
of Reading.
This is the time to choose letters to represent each phoneme of the
language. Hopefully, most choices can reect the original, intuitive choices of the people.
Regard your alphabet decisions as tentative. It is not unusual to revise symbols later, after
more investigation.
A bit of time should be devoted to alphabetic sequence for each language. There are a
number of alphabetic issues which are not automatically obvious to the speakers of the lan-
guage. These concern the order of symbols not used in a language of wider communication,
and also the fact that we are generally listing graphemes for each phoneme in the alphabet
listing, including digraphs and trigraphs, and possibly including prenasalised consonants.
The literacy specialist or workshop leader could gradually work out a general alphabet
listing containing the sum total of all graphemes used for the alphabets of the dierent
languages. From that “total alphabet,” each language group might then deduce its own set.
People enjoy this exercise.
Sequence principles: special symbols usually follow the symbol they are most closely
related to, such as e...ɛ, i...ɨ, o...ɔ, u...ʉ.
Digraphs and trigraphs, for simple or for complex sounds, should follow the alphabetical
order, with secondary sorting on the second symbol, such as m - mb – mv – n – nd – ng –
ngb – nj – ny – nz.
98
Bantu Orthography Manual
Allow 2 to 4 hours per chart for their production on computer (see Part 2, Appendix A:
Alphabet Chart Template
5
), once keywords and pictures have been chosen by the group.
An alphabet chart is a pedagogical tool (but it is no substitute for a primer)
1. If a target-language speaker is already uent in a language of wider
communication, though, such a chart may be all he/she needs in order to begin
reading in the mother tongue, especially if the languages are related and they
have many corresponding sounds/symbols.
2. An alphabet chart should normally present the phonemes of the language with a
simple, direct correlation between sound and symbol.
3. It’s not that simple because sociolinguistic considerations should be reected in
the chart. For example, if most target-language speakers of a minority language
are literate in an LWC, literacy goals will involve transition from L2 to L1 (or back
transition).
• People may want to over-dierentiate sounds from their own language if they
are already distinguished in the LWC, as in /r/ and /l/ in Swahili. This may
actually be acceptable, for purposes of transitional reading.
• Auditory awareness of the sound/grapheme being taught is key. Usually, it is
best for the consonant being taught to occupy syllable-initial position in the
root, as people are most aware of sounds in this position.
• On the other hand, some people may want to show a distinction between their
language and the LWC by writing their unique sounds dierently. The Kwaya
people chose to write the soft sound of b, which they say is very dierent
from Swahili’s hard-sounding voiced bilabial. For this reason they insist on
writing their voiced bilabial (allophonic variants are [β] and [b]), which is
always a fricative unless it occurs after an <m>, as <bh>. This choice is
not recommended but may be necessary for community acceptance of the
orthography (see Part 2, Appendix D, Lesson 3: Consonant Symbol Choices, and
Consonants in Part 1, chapter 2, Representing Linguistic Features).
• The minority language may have absorbed a number of loanwords from the
LWC, and hence certain letters from that language, for sounds which are not
part of that language’s phonemic inventory. Ideally, one symbol should be
chosen to accommodate both. Take a look at the list of loanwords which was
printed out at the end of word collection. Example: The Tharaka language
never needs or uses the letter <s>, using <c> for the voiceless sibilant. But
when the New Testament was published, people wanted place names to be
spelled using the same place names found in a Swahili Bible, using the letter
“s,” so two letters for the sound need to be presented in the alphabet chart.
• Labialization (indicated by <w>) and palatalization (indicated by <y>)
can aect virtually any Bantu consonant. The letters w and y also occur alone,
allowing them to be taught as separate letters. Again, don’t teach mpw, gw, and
all the other labialized consonants as digraphs in the alphabet. They function
5
This is a Publisher 2003 le.
99
Chapter 3: Orthography Workshop 1:Phonological Analysis
more like consonant clusters. They can be presented very simply in a primer,
grouped with others, once people can recognize their constituent letters. Once
people can decode <k> and <w> individually, for example, a key word
such as ‘to plant’ can be introduced. This would be followed by an
exercise such as the following, where the primary member of the cluster is the
consonant, not the semivowel.
• Optimal key words are ones which are not only quickly recognizable, but words
in which the learner can easily hear the sound being taught. For this reason,
it is ideal for the consonant taught to occur in root-initial position and for a
vowel to occur repeatedly, especially in root-initial position.
An alphabet chart is a publicity tool
1. It can be a condence builder, showing the would-be reader that
reading the mother tongue is easy.
• For this reason, it may be best to keep the chart simple and
not teach double vowels as additional letters. In a seven-
vowel language, telling people they have 14 vowels can be
intimidating and necessitate visual crowding on the chart. If
you choose to show the 5 or 7 vowels of the language
on the chart rather than 10 or 14 and vowel length is
contrastive, be sure to include some picturable words
on your chart which
show this feature, as in the example here.
• The same is true for introducing any prenasalized stops
as separate “letters” on a chart (common in Bantu J
and E languages). It can be done, though the softer
[d] and its harder form in [nd] are actually allophones.
Since the prenasalized stops (i.e., mp, nd, nc) of Bantu
languages normally cannot be pronounced in isolation, they must be taught
as digraphs in a primer. Including them in the alphabet chart as well is an
option, but not a necessity. If they are not taught directly in the chart, be sure
to include some key words which have these complex sounds in them, as in the
example here.
• Since it is often hard to nd certain letters occurring word-initially, try to
nd words which exhibit multiple occurrences of the letter being taught. The
Tharaka language, used in the example above for the letter A, has very few
vowel-initial words, so it was decided to use words which contain only the
vowel being taught, when introducing each vowel. In Bantu words, one can
benet from multiple repetitions of a letter within a word, as in the example
key words kaana or mpempe, shown. For languages with noun-class prex
plus augment, even this may be impossible, but at least the noun prexes are
pronounced with less stress than the root and with less target-language speaker
awareness, so they don’t seem to interfere with the sound-symbol connection
being presented.
A a
kaana
E e
mpempe
100
Bantu Orthography Manual
2. The chart introduces people to some features of their language’s writing system,
but not all. Word break rules and features of the language’s morphology will
require further instruction, for would-be writers of the language.
Participants choose key words—picturable words, usually nouns—for their preliminary
alphabet chart. All illustrations must be approved by target-language speakers, to ensure
that they unequivocally communicate the intended meaning. The linguist helps people
choose words for best auditory discrimination of the sound being taught, as described
above. A note about illustrations: strive for immediate recognizability
1. Let target-language speakers choose all illustrations, to ensure that they
communicate well.
2. If no suitable illustration can be found from the Art of Reading CD to represent a
given word or letter, a participant may need to draw a picture and have it scanned
for inclusion in the alphabet chart, using the template.
Any unresolved phonological issues such as choice of some alphabetic symbols, ATR
harmony, semivowel/vowel problems, or vowel length should be noted in the manual
Workbook Form, as some deeper and more detailed analysis will be necessary. They may
require focused attention at the next workshop, including orthography testing. Inform the
literacy consultant or the linguistic consultant as to these issues. Provide examples.
(These can be done contiguously with card sorting, by subdividing a
language group for focus on dierent tasks)
1. Begin revising the electronic wordlists, based upon decisions made.
2. Record 3 to 4 short stories (less than 5 minutes each) on a sound recorder,
creating wav les.
3. Teach some target-language speakers to transcribe the stories, using Speech
Analyzer.
• Each story is written using black ink, one line per sentence or clause.
• Below each line of the story, in red ink, is a word-for-word transcription in a
LWC
• On a third line, if needed, transcription can be done in blue ink, with more
natural sentence ow in the LWC.
4. Transcription reminders:
• listen for pauses. Each one signals a new line on the paper.
• mark them with full stops
• leave room in the margins
101
Chapter 3: Orthography Workshop 1:Phonological Analysis
• use capital letters at the beginning of sentences (or phrases, at this point)
• leave three lines below each transcription
• choose one color of ink for the whole story
• it’s helpful to number the original sentences
• Check any LWC glosses which need revision to match the vernacular word on
the card/computer.
• Take care that transcription of the gloss is as close to the vernacular morpheme
order as possible. For example, if the verb in the vernacular has an object
marker in it, the word-for-word LWC (a Bantu language) translation should
have a corresponding object marker. People will need a little lecture here, with
examples of “free” and word-for-word translation. Use a language like English,
which is guaranteed to be quite dierent from theirs. “Once upon a time there
were three bears…: “Once upon a time” would literally be “Mara moja juu ya
wakati,” whereas, written freely, it would be “Hapo zamani” or something.
Try to nd a place where each language group can work without distraction from the
other groups, all of whom will be whistling like a ock of birds. Two or three target-lan-
guage speakers are an optimal number. If there are more, some may be best used at this
time for recording or transcribing stories in another room or choosing illustrations for an
alphabet chart.
While complete tonal analysis is impossible at this time, a number of basic questions can
be answered now:
• Is tone on noun prex stable (i.e., always high)?
• If it changes, does it change only when followed by a V-initial stem, or even when
followed by a C-initial stem?
• Tone on roots: are there rising and falling contour tones on the surface? In what
syllable positions?
• Can a contour be found only on a long vowel, or on short vowels as well?
In most Bantu languages in Central and Eastern Africa, the basic tone system consists of
two underlying tones, High and Low. However, there are exceptions here and there.
Since tone is a relative feature and always perceived in relation to tone in other words,
the tone of a particular word is usually elicited within a frame. A frame for verbs could be
the imperative followed by an adverb, like “______ well!” This would work for most intransi-
tive verbs and others which don’t require a stated object. A frame may be essential in West
Africa, where oating tones cause perturbations on the tones of adjacent words. Since this
initial charting activity is just giving the big picture though, a frame is optional
Use chart paper to track tone patterns on verbs, like the one for nouns below, but
without the prex column.
102
Bantu Orthography Manual
(while others are telling or transcibing stories)Set aside all verbs which
don’t have C-initial roots.
Someone in the group reads each word and whistles all verb innitives of a particu-
lar syllable structure. Divide the melodies into consistent groupings. It is most likely that
verbal innitives will fall into two tonal classes at most, and perhaps only one. The longer
a verb stem is, the easier it is to hear the dierent melodies, so include not only the CVC-
FV verb roots, which were sorted already for segmental phoneme discovery, but also the
C-initial verb roots with extensions, (i.e., okorareka, okoraghora, okobharoka).
If the root is C-initial, have someone read it and whistle its tonal melody, so that each
syllable (of the whole word, not just the root) gets a whistle and make piles of “same” and
“dierent” melody. This will automatically result in the sorting of two-, three- and
four-syllable verbs into separate piles, since HHHL is not the same as HHHHL. If there are
dierent tonal classes for verbs, you may have two piles of ve-syllable innitives with
dierent melodies.
Do not make separate piles for dierences in tempo! Ignore the length of vowels, listen-
ing only for dierences in pitch pattern among verbs with the same number of syllables.
The rst possible expectation would be to have two groups: H-tone verbs and L-tone
verbs (whereby the H or L tone belongs inherently to the root, but may be realised on the
root itself, or later, or sometimes even earlier). Another possibility is predictable high pitch
on the antepenultimate mora.
Listen to the whistles and try to mark down the tonal pattern of each group of verbs.
With practice participants often recognize the tonal group to which a verb belongs, with
lots of whistling and listening. Pitch traces are the best way to record the patterns. Most
Bantu languages have downdrift, so that a H tone that occurs after a L tone will have a
lower pitch than a H tone that preceded the L, and a pitch trace transcription shows this
more accurately.
If there are tonal classes for verbs, note this on the chart of the Workbook Form.
For a rst impression of nominal tone in most Bantu languages, just elicit nouns in iso-
lation. If you are in a language area in which oating tones or H spreading are common,
though, use a frame. Try a verb plus noun frame, such as “Look at the ______.” This would
result in a verb preceding the noun. If no perturbations seem to occur on adjacent words,
try a simpler frame such as a noun plus demonstrative, “That _____.”
There is a slight complication with preceding frames for languages where there is a
vocalic augment. The augment of the noun will coalesce with the nal vowel of the preced-
ing word. Also, in languages where the adjective has an augment as well, it may be hard to
pronounce without the augment coalescing with the nal vowel of the noun. Instead, using
a following associative phrase may be helpful. Regardless, when augments are involved,
elicit the isolated noun rst, and then use the frame.
Sort all nouns of the basic -CVCV root structure by noun class, and design a chart
for nominal tone. Make the boxes big enough to “hold” stacks of word cards, sorted by mel-
ody (written with pitch traces at the top of each box) and number of syllables in the root.
103
Chapter 3: Orthography Workshop 1:Phonological Analysis
Sort the nouns by
class, and then by roots.Stick to -CVCV and -CVCVCV roots. Then,
1. Go through each stack using the same procedure as for verbs, reading and
whistling each word. Only sort by melodies in which the pitches show a dierent
pattern, and between dierent numbers of syllables. The speakers of the language
should get into a rhythm of speaking the word (held up or written for all to
see), then whistling it, then creating groups of words which have the same tonal
melody.
2. The TA writes a pitch trace for each pattern in a box of a chart like the one below,
made on large chart paper. One participant should be given responsibility for
keeping a particular stack or stacks neatly piled and/or clipped together. Based on
the checking system explained above, list the tonal melodies for nouns. If there is
enough data, this should be done preferably per noun class (so that one can ensure
the full range of possibilities is found in every noun class).
3. As cards are matched with a pitch trace, stacks are formed.
4. Groups listen for the pattern of words within each stack, for consistency of
melody.
6
Most patterns (or all) should show up in all the noun classes. However, by keeping the
dierent noun classes in separate piles, you will have a pile of cl. 1 nouns with a LLHL pat-
tern and a pile of cl. 3 nouns with a LLHL pattern, a pils of cl. 5 nouns with a LLHL pat-
tern. Have a spot (“square”) for each pattern, keeping the cards of each class paperclipped
together on that spot.
Most Bantu languages will have four or fewer classes of tone patterns in nouns, and the
same patterns will be found in all the noun classes. However, there will probably be some
exceptional words that do not t easily into any group. Do not expect to have a thorough
analysis of tone at the end of this exercise, but it is hoped you will be able to document
some basic tonal melodies.
7
6
If there is an occasional multi-syllable word that does not t any patterns, it should be set aside for later
evaluation. It might easily prove to be a compound word or a phrase.
7
In some countries, such as Tanzania, where the inuence of another language is very strong, the lan-
guage may be in the process of losing tonal contrast. Be aware. If you nd no contrast between words in
isolation, it may be useful to test the words within some frames, i.e., short phrases, to see if there is a con-
trast which emerges in context.
104
Bantu Orthography Manual
Syllables At the top of each square below, (as needed) write pitch traces for
sorting cards into tonal patterns. Additional columns may be added
to the right for causative/passive if the melodies are dierent.
Prexes + 1
syllable,C-
initial
Ex: [
]
Put cards here
Ex: [ ]
Put cards here
Prexes + 2
syllable,C-
initial
Prexes + 3
syllable,C-
initial
Prexes + 4
syllable,C-
initial
Prexes + 5
syllable,C-
initial
Prexes +
2syllable
V-initial
Prexes +
3syllable
V-initial
Prexes +
4 syllable
V-initial
Prexes +
5 syllable
V-initial
At the end of the process, keeping the dierent noun classes in separate piles, the re-
sult will be a pile of cl. 1 nouns with a LLHL pattern and a pile of cl. 3 nouns with LLHL
pattern, a pile of cl. 5 nouns with the same pattern, and so on. Each class will be paper-
clipped, and all of the same melody and number of syllables will then be stacked and
rubber-banded together.
Do the same exercise for nouns as previously done for verbs. The most probable nding
will be two tone groups: H-tone verbs and L-tone verbs (whereby the H or L tone belongs
105
Chapter 3: Orthography Workshop 1:Phonological Analysis
inherently to the root, but may be realized on the root itself, or later, or sometimes even
earlier). Another possibility will be predictable high pitch on the antepenultimate mora.
Based on the checking system explained above, list the tonal melodies for nouns. If there is
enough data, this should be done preferably per noun class (so the full range of possibilities
is explored for every noun class). Similarly, the tonal classes for verbs can be established.
These activities are primarily to be done by a TA.Keep a complete list of minimal
pairs: nouns which are segmentally identical but which have dierent tonal melodies, and
similarly verbs which have dierent underlying tones.
Since there are generally four dierent noun-tone patterns, and only two dierent verb-
tone classes, there are generally more lexical minimal pairs in the verb system than in the
noun system. The function of tone in the grammar is generally much more important.
It is very important to nd out where the tone functions in the grammar. One can try
to nd out if it functions in particular domains which have tonal contrast in other Bantu
languages, typically:
• 2sg. vs. 3sg. subject prex u- with L and H tone underlyingly (but sometimes,
because of tone shift, not even realised on the morpheme itself, but a couple of
syllables to the right)
• 3sg. vs. 2pl. object prex -m- with L and H tone respectively
• Verb tenses, esp. 1
st
degree past vs. 3
rd
degree past
• Relative clauses
Note the number of tone classes for nouns on the section of the
Workbook Form, with examples.
List the tonal melodies for nouns. If there is enough data, this should be done preferably
per noun class (so that one can ensure the full range of possibilities is found in every noun
class). There is room in the tone worksheet to note any observations or impressions about
such features and processes as:
• Depressor consonants
• Tone spreading
• Tone shifting
• Downdrift and/or downstep
Bring laptops today. This can be a day for bringing closure to either tone classes,
alphabet charts, or text recording and transcription. It is also an excellent time for a few
target-language speakers to assist the TA in correcting spelling and glosses in the electronic
data, referring to the corrected word cards.
If you can get all your noun and verb entries corrected in the
Toolbox database about 2 days before everyone leaves, you can then check the list on your
laptop, with the participants watching/assisting. Side benet: the experts are there with
106
Bantu Orthography Manual
the linguist to not only correct spelling or clarify pronunciation, but to revise erroneous
glosses. The participants learn a lot from the experience, too.
Assign someone from each language group to prepare, together with the rest of the
group if possible, a 10-minute report on the linguistic features of their language (i.e., how
many vowels do they have, anything interesting about their noun class prexes, their
consonants, tone, or length…) and interesting features of their alphabet. The report will be
presented on the following day.
Finalize the 500-wordlists and alphabet charts for each group. Print out
8
at least one
copy of each, and allow one or two people within each group to check through them, while
the others work on tone or story transcription. It is suggested that charts and lists give
credit to the people who checked and developed them, with a listing of the names of the
workshop participants somewhere on the papers.
This is the day for:
• review of what has been learned: a representative of each language group should
make a short presentation to the large group about the linguistic discoveries
relevant to their language.
• closing ceremony.
• distribution of a limited number of alphabet charts (one per participant) and
wordlists, which then may be viewed by the larger language community after
people leave.
Before alphabet charts are distributed, instruct participants: “Target-language speakers
need practice reading their own language and may not at rst like all the spelling choices
we have made here. This is natural. Some things will not be written quite the way they ex-
pected. But if many people have trouble reading the same thing, write that word down on a
separate piece of paper, and tell us next time. If people say, ‘We don’t say it like this. It has
a longer sound,’ what will you (the former workshop participant) tell them?” Be sure they
are prepared to answer such questions.
Distribute, read together, and discuss the following instruction sheet, translated into the
language of wider communication. It will accompany the wordlist and alphabet chart.
8
Use Lexique Pro to print the wordlists. Someone will need Publisher 2003 in order to print the alphabet
charts.
107
Chapter 3 - Orthography Testing
Show both the chart and the wordlist to at least 15 people. When you do,
1. Ask them to sign their name on the back of the list.
2. Show the alphabet chart rst. If there are any new symbols on it, make sure
people understand them.
3. List any problems with alphabet chart pictures (or suggestions), if any.
4. List (later, when they’re not looking) any words which are often a problem for
people to read.
5. If someone to whom you showed the list speaks another dialect of your mother-
tongue, please note this next to their signature.
6. Were there any dialect-related problems anyone had, either with the ABC chart or
the wordlist?
7. Record any complaints.
8. Evaluate before you come back:
• Did most people like the ABC chart?
• Did most people like the spellings of words in the list?
• Do most people like the idea that their language is now written?
108
Morphophonological processes in Bantu languages involve:
(C)V – V hiatus resolutions: hetero-syllabication (e.g., formation of two syllables from
one), elision, and glide formation (often two or three of these solutions are used in
one language)
C – V morphophonology
C – C morphophonology (particularly in cl. 9/10 noun-class morphology, and 1sg
subject/object prex morphology)
In Bantu languages, the study of morphophonology should be divided into two parts:
Nominal morphology and morphophonology
Verbal morphology and morphophonology
The topics under nominal morphology and morphophonology are:
Noun-class prexes of nouns in their citation forms
Class prexes used with adjectives
Class prexes used with numerals
Class prexes used with possessive pronouns
The associative markers
Class prexes used with demonstratives
Class prexes used with any other noun modiers
The topics under verbal morphology and morphophonology are:
Inectional
• subject prexes (for participants and for classes)
• object prexes (for participants and for classes)
• relative prexes (for participants and for classes)
• T/A markers
• Negative markers
• any other inectional elements particular to a language
Derivational
• Verb – Verb:
– productive derivational suxes: causative, passive, applicative, reciprocal,
for example
– combinations of these derivational suxes
– changes in verbal aspectual suxes because of the use of (combinations of)
derivational suxes
• Verb – Noun: agentive nouns derived from verbs
• Verb – Adjective: adjectives derived from verbs
Once clear and systematic solutions have been found for these various morphophono-
logical issues, charts can be drawn, particularly of the class prexes functioning with the
various nominal modiers, which often contain a lot of morphophonology. Record not only
orthographic conventions chosen, but the actual pronunciations, when they diverge from
109
Chapter 4: Orthography Workshop 2: Morphophonology
spellings! It is important that a record be left behind for any future linguistic research. For
example, the Kwaya language of Mara Tanzania has two charts for associative marker: one
reecting the spelling decisions, and the other representing the phonology of the language.
Some charts may later be distributed, or hung up on the walls of a translation oce, so
that people frequently refer to them and learn to write correctly.
By way of example, see the following chart. It portrays the complete set of possessive
pronouns for all classes in the Yaka orthography. The language is YAKA (C10), CAR.
1sg 2sg 3sg 1pl 2pl 3pl
1
2
3
4
5
6
7
8
9
13
14
19
With the information obtained so far, a writers’ guide can gradually be developed. The
writers’ guides are not a phonological justication for spelling decisions. They present the
agreed-upon spelling, with examples, to provide guidelines for correct writing for present
and future generations. Examples of writers’ guides are given in Appendix 1, and others
may be available from your entity.
110
Bantu Orthography Manual
The following elements should denitely be found in a Writers’ Guide:
• choice of alphabet letters
• word boundary decisions
• capital letters, punctuation rules
Literacy Lecture, Part 2, Appendix G, Lesson 6: Goals and Products of Workshop 2
and charts described below and pictured right. Read Part 1, chapter 2: Morphophonological
Processes, Principles of Tone Marking and Word Boundary Principles.
Large wall chart copies of each language’s noun classes, with
prexes and an example noun for each. These charts will be used almost daily as a
reference.
Poster-size checklist of charts to be lled and consultant-checked during the 3-week
workshop, showing all language groups’ progress (see below).
rules for vowel length, Dahl’s Law, noun phrase constituents,
adjectives.
Chart adjectives which begin with consonants, and those which begin
with vowels, for each noun class.
Put big wall charts up, attaching 2 blank chart papers to the right of the noun
class charts. Linguist demonstrates how to ll in the rst 4 rows of one language group for
all, before dividing into groups.
Chart numerals, 1–10
Fill in chart.
(Ngoreme) omoti gamwe
(Kwaya) omugongo gumwi
(Simbiti) omote gomwe
(Ikiizu) ʉmʉti gʉmwɨ
Some noun class prexes will not be retained beyond the number ve. The enumerative
for class 10 is always a noun.
111
Chapter 4: Orthography Workshop 2: Morphophonology
As a review and a learning experience, a linguistic consultant charts
samples from all, on ip chart, as above: Noun Classes 2 and 3 with numbers 1, 2, 6, and 8
in separate chart pages.
Read Word Boundary Principles section of Part 1, chapter 2.
1. Linguist explains how “near” and “far” work in Swahili, with a chart:
(singular) huyu yule
(plural) hawa wale
2. Linguist demonstrates the same thing for another Bantu language. See
chart, as a formatting guide for
near to speaker, near to hearer, far from both, and previous referent ( but known
about mutually).
Chart demonstratives on a large wall chart,
aligned with noun class prexes.
Copy the wall chart onto
chart.
Chart on
chart paper, with the group.
Look for Emphatic forms to show up, and note them if they
occur.
Continue in the same manner with interrogatives and with
all of noun phrase charting below.
Fill in electronic charts for each of the fol-
lowing, after lling in blank chart paper with participants
in each language group. Be sure to record any spelling
112
Bantu Orthography Manual
decisions with examples and any unresolved spelling issues (with examples) on the Spelling
Rules page following the chart.
TAs should read Part 1, chapter 2, Morphophonological Processes.
“What is a word?” Part 2, Appendix H, Lesson 7: Word Break with
locative examples (and their 3 noun classes) from Niha.
Three possible orthography choices are oered: 1) writing the underlying locative, with
vowel changes written only on the following noun with a space between the two words; 2)
writing the vowel change onto both the locative and the following noun, allowing vowel
length to be accurately reected; and 3) writing the vowel change only on the locative,
dropping it from the prex position on the noun and using a hyphen.
People break into groups to chart their locatives, written disjunctively in
one of these ways. Experimentation is good before a decision is reached.
Record both the spelling decisions and the phonological processes on the
charts. This can be done by someone assigned the task.
A consultant gives a description of the procedure for
lling in the chart of associatives. She gives Bantu examples of the morphophonological
changes across the word break. Options for writing them do not include hyphenation (lit-
eracy consultant provides all the possibilities). We don’t want associative markers to look
like locatives, stuck onto the beginning of the noun. Also, since in Swahili the associative
is written disjunctively, people should not object to writing them separately from the nouns
they precede. People will not usually question the suggestion. The waIkiizu (E402), how-
ever, chose to just write the associatives and the locatives in their underlying forms. After
all, they have 7 vowels, and the variant speech forms are many indeed. They also are con-
vinced that the associative is always wa, despite its morphing before every V-initial noun.
Record both the spelling decisions and the phonological processes on the
charts. This can be done by a person assigned the task.
113
Chapter 4: Orthography Workshop 2: Morphophonology
Linguist or TA presents several possible forms of relative phrase. In each
case, an element in the verb phrase coordinates with the noun at the head of the phrase.
The phrase functions like an adjective modifying the noun, so you still need a “real” verb
later in the sentence.
Post a progress chart like the following up on a wall, minus the Xes, for all to
see. Groups may cross o charts after they have been “handed in” electronically to the
consultant/s. Consultants circle the Xes on the wall after charts have been checked with a
target-language speaker.
Wrapping up Charts
Chart Needed Ngoreme Kwaya Ikizu Simbiti
Nouns: Demonstrative forms- other x x x
Demonstrative/Demonstrative
Demonst. Phrases
x x x
Interrogative
x
x
x x
Pronominal forms -1
x
x
x x
Pronominal forms -2
x
x Help
x
Pronominal forms -3
x
x
x x
Adjective
x
x
x
Associative Constructions
x
Help
x
Diminutives and Augmentatives
x
x x
Enumeratives
x
x x
Locatives
x
x
x
Possessive pronouns
x
x
Verbs: Copular forms
Copular forms part 2
x
Morphophonological
processes…Perfective
Verbal Mood: Imperative plural
x
x
x
Verbal Mood: Imperative singular
x
x
x
Verbal Mood: Subjunctive
x
Verbal Mood: Subjunctive 2
x
x
114
Bantu Orthography Manual
Verbal Prexes: Subject.
x
x x
Verbal Prexes: Subject 2
x
Verbal Prexes: Object
x x x x
Verbal Prexes: Object Relat.
x x
Verbal Prexes: Subject Relat.
x
Verbal Tense: Future 1
x x x
Verbal Tense: Future 2
x x
Verbal Tense: Future 3
x x
Verbal Tense: Future 4
--- x
Verbal Tense: Past 1
x x x
Verbal Tense: Past 2
x x
Verbal Tense: Past 3
x x
Verbal Tense: Present or unmarked
x
Verbal Tense: Present 2
--- --- --- ---
Verbal Tense-Aspect/other
see Past
1-3
Verbal Tense-Aspect Anterior
x x x see Past
1-3
Verbal Tense-Aspect Past Anterior
x see Past
1-3
Verbal Tense-Aspect Progressive
x x x
Verbal Tense-Aspect Present Persistive
x x
Verbal Tense-Consecutive
x
Vowel Harmony in Extensions
x
Conditional (if it exists)
x
Regret?
x
Stories (Hadithi)
x x x x
Adjectives (Vivumishi)
x x x x
Associatives (-a Unganifu) (optional)
x
Noun Classes (Ngeli za Maneno)
x x x x
: Review Word Boundary Principles in Part 1, chapter 2, Linguistic Features. This
week has three goals for each language group:
1. lling and checking of all remaining morphology charts
2. review of spelling rules (i.e., word break rules, writing of grammatical vowel
length, spelling of copulas and focus markers, grammatical or lexical tone
marking) decided upon as part of the charting process
115
Chapter 4: Orthography Workshop 2: Morphophonology
3. editing an oral text which was transcribed at the previous workshop, to be
published
Charts to be checked, with linguist (she/he has the master set of all charts from this
point):
• Associative
• Locative
Items to be published:
• Noun class charts, copied o from worksheet and corrected if needed
• Adjective charts
• Verbal prexes: subject or object
• Stories
How chart checking should be done:
• Linguist sits at the front table with a laptop. People bring their assigned chart to
the linguist on a ash drive. Bring a target-language speaker along. TA is optional.
• Linguist asks questions and types phonemic information and linguistic details
as needed on each “chart” given to him/her. Linguist highlights any spelling
questions which arise, consulting the Orthography Manual, testing, and/or
discussing challenges with the literacy specialist or consultant.
Literacy consultant: Make sure to get alphabet charts checked o, vowel books (for
7-vowel languages) approved. Assist with orthography questions and story editing issues.
Make sure you have an edited story from each group by two days before the workshop
ends. Collect all four items for publishing.
Follow the rules the group have already chosen and edit their stories.
Now try reading them a couple of times. If rules need to be changed or added, call a con-
sultant over. If they do, some charts will have to be rewritten.
Encourage groups to turn their charts in daily this week and to have them checked by
your linguistic consultant working with a target-language speaker, to ensure that the charts
accurately represent the pronunciation, not just the orthographic conventions chosen, of the
linguistic phenomena recorded. The literacy consultant should oer alternatives and testing
suggestions for any unresolved orthography issues as they arise.
Check these charts with linguist as time allows:
• Imperative
• Noun class charts, copied o from worksheet, and corrected if needed
• Adjective charts
• Verbal prexes: subject
• Use of the augment (questionnaire)
• Copular forms
• Imperative (singular?)
116
Bantu Orthography Manual
• Relative phrases
• Vowel coalescence chart, if not already completed
• Print all charts, copy all electronic charts available, backing them up onto the
workshop leader’s computer
• Print all ABC charts which need a reprint
• Publish any booklets which may be needed
(nd in Appendix I, Lesson 8, page 137) Review the
writing rules for one language group, beginning with a sample sentence or two from one
of their stories. Outline a simple poster chart, which will be lled in by each group, listing
rules for spelling:
• Associatives
• Locatives
• Copulas and other clitics in the language (when analyzing, pay particular attention
to varying functions or possible similarities or contrasts with a focus marker /n/ and
with the conjunction /na/)
• Tone
• Other
Groups now begin listing the rules they have already prescribed,
as a review.
Groups begin editing their stories (for large groups the stories have been written out on
chart paper, copied exactly as they were transcribed during Orthography 1). The entire
group should be able to see the stories and editing should be a group process. Go through
the stories at least twice, making sure that all locatives and associatives have been noted.
Ask dierent people to read the story aloud. If several people have problems at the same
place in the story, note where they stumble and inquire as to the reason. Get the story
checked by a consultant and prepare it for publishing.
If time allows, participants may select an illustration for the story from the Art of
Reading CD (see www.ethnologue.com/LL_docs/intl_illustrations.asp).
During closing ceremony, the chosen story is read to the whole
group, and each participant is given a copy of the 3 charts and the story, plus any edited
alphabet charts or vowel booklets.
117
Title in local language
Alfabeti za Ki...
yagakana
A a B b
baaba
118
Learning Objectives for this lecture:
1. People know what tangible results to expect from the workshop
2. People understand that choice of grapheme to represent a sound varies with the
needs of a language (so sociolinguistic factors will not have a larger inuence on
grapheme choices than optimal for reading/writing)
We are here to begin developing writing systems for your languages. We hope you will
leave here with an alphabet chart and a list of words written in your language. This work
will not be easy. Your biggest challenges will be:
1. Helping writers by making spelling easy for them, with simple rules
2. Helping new readers by using one symbol for each sound in your language
whenever possible. We use alphabets to do this.
Show *sample alphabet charts (i.e., for related languages). *Chalkboard needed
We hope that each of you can leave here after three weeks, with a chart like
one of these, which helps people be aware of all of the sounds which make a
dierence in meaning for your own language. Sometime during the second week
of the workshop you will begin to choose pictures of things which everyone in
your language group can recognize, to represent the sounds of your language. An
ABC chart like these, but with pictures of words from your language, will help
people learn to read it someday.
You will be working very hard for three weeks on things which are new to you,
yet also familiar. They are familiar, because you use these sounds and meanings
every day. They are also new and strange, because you have not had to think
much about how they would look if they were written down.
Writing systems teach people many things about their language, things they have
never had to think about until they are written down. (Distribute sample alphabet
charts.) What can you learn about a language (and a culture) by looking at its
alphabet chart? (Discuss briey.)
See the Tharaka alphabet chart (for example). How many vowels does this
language have? Do they just have short vowels, or long ones too? Look closely.
Any other observations? Some of the Tharaka “letters” are very long. Why have
we put pictures in the chart, rather than just words?
119
Appendix B
Look at Suba. This language is more closely related to yours. Are most of the
items familiar to you? The Suba alphabet chart teaches the long and short vowel
sounds separately, though they actually have the same sounds and distinctions as
Tharaka vowels.
Look at the letters.
• Suba has the letter L. Does Tharaka? No, because in Kitharaka the two letters
<r> and <l>do not make a dierence in the meanings of their words.
• On the other hand, Tharaka has the letter û and the letter î, 7 vowels in all.
Does Suba? No, Suba doesn’t need them because they only have ve vowels
which make a dierence in the meanings of their words.
• Which alphabet has a z?
• Which one has an s? Why do you think Tharaka chose c instead of s? There are
problems with their choice for spellers and people who read more than just
Kitharaka, but English or Swahili. We will talk about such things later.
• Which letters do you think are easiest to read? Why?
• Which letters might be the hardest, both to spell and to read?
We will be talking, next time, about the things which make an alphabet easiest to
use, both for recognizing dierent letters and for writing them in words. As we
study your languages together, we will all talk more about the things you must
consider in choosing letters for your language’s alphabet.
3. Helping new readers by breaking up the sounds into words and sentences.
Example: sinunuimkatemjinikwasababuhaufai
Exercise: Break into words—Sinunui mkate mjini kwa sababu haufai.
When we talk to our friends, we usually drop some of the sounds in words,
knowing they will understand us. In English, when we talk fast, we say “G’bye.”
But if we speak slowly, we are actually saying “Good Bye.” That is the way we
need to write it: “Good Bye.” We native English speakers say, “Dijuseeimatth
store?” But when we write, we break these sounds of fast speech into slow speech,
and really help the reader and writer to understand: “Did you see him at…” It is
the same for you. In Swahili, you may say, “…..” when, if you think about it, or if
you write it, it is actually two words: Swahili example:
4. Helping readers by making reading easy for them:
• Symbols which are easy to distinguish, (Example: what if all of our letters were
tall, like this: lltt!!ɺf)?
• Shorter words, when possible
• Giving the reader all the information he needs to be sure of the meaning of a
word (including tone, the correct vowel, length, etc.)
(Examples of contrastive vowel length and contrastive tone)
What happens to your understanding of these sentences? What makes a
dierence in your understanding?
• Mti ambao unafa
120
Bantu Orthography Manual
• Mti ambao unafaa
5. Helping readers understand what the writer wants to tell them (not just helping
them sound out words so they can be pronounced). We use alphabets to help us
do this. We choose letters to represent sounds. But a writing system is more than
sounds. I could teach you the Spanish alphabet, A, B, C, Ch, D, E, F, G, H, I, J, K…
but if you could repeat the sounds after me and sound them out in words, would
you be reading? (Example: Quisiera leer mi idioma.) Did you understand? No, of
course, you still need to get meaning from words in order to be truly reading.
Another goal we have for each language group is to record some of your stories before
we leave. You might want to be thinking of an interesting one which your best storyteller
can tell, later this week.
121
Learning objectives: people will start thinking about readability/writeability issues in
choosing graphemes:
• Visual discrimination of symbols
What is dicult here for a new reader? Find the letter which is dierent in
each row. (Chalkboard)
αααoα (Can you easily nd the o?)
uunuu
bdbbb
pbbb
uuûu
t
• Simple sound-symbol connections (n vs. ng’)
• Comprehension challenges (see examples below)
In order to write what we say, we must
1. Break up the stream of speech : sinunuimkatemjinikwasababuhaufai
Divide the stream into words: Sinunui mkate mjini kwa sababu…
2. Choose from the ancient Roman symbols available—the letters we need in order
to write the sounds of our languages.
These letters are a code developed by the Romans thousands of years ago, for their
language, Latin.
A B C D E F G H I J K L M N O P Q R S T U V W X Y Z.
The Spanish needed some dierent letters, so their alphabet looked like this:
A B C Ch D E F G H I J K L Ll M N Ñ OP Q R RR S T U V W X Y Z
3. Notice that there are very few digraphs in these alphabets, because there is almost
a 1-to-1 correspondence between 1 symbol and 1 sound.
4. We can use those letters in any way we want, so they work for our language.
Because all of you already know how to read, you can easily learn how to
pronounce some new letters, or say them dierently when you read a dierent
language with them. For instance, with the Spanish language I could tell you that
122
Bantu Orthography Manual
• Ñ stands for “ny,” and when you read these words, you will say the “ny”
sound: say
ni ña – ‘girl’
a ra ña – ‘spider”
• Spanish uses a double l for a ‘y’ sound can you “read” these words?
Llamo
amarillo
llora
• Now, here’s another dierence in the Spanish alphabet: they use “qu” instead
of k, before e or i. Can you “read” these the Spanish way?
Queso –‘cheese’ and
Quiero – ‘I love’?
Now you all are ready to “read” Spanish. Yo quiero
queso amarillo, para mi niña. (I want yellow cheese for my little girl)
You were not really reading, of course, because you couldn’t understand what you
said! But it does show you that once people have learned how to convert letters
into sound, the rest is easy, especially with understanding!
5. Here is an important point about writing languages down: We help readers when
we write down the way people talk, but in slow speech.
If we wrote the way native English speakers speak English, we would lose many
things which help a reader understand. For example, we often say: “Waijaduthat?”
and “gimmithiapple.” You must read them aloud, in order to gyuremaut—gure
them out.
We need writing systems which show slow speech, not words crowded together,
so the reader always can recognize whole words.
For example: na antû and people not nantû
ni wee it is him not nwee
6. Problem for readers: remembering a long series of sounds or syllables:
okwifunyongora is harder than a word like rigori.
7. Readers have to tell the dierence between all these letters. We may be able to
help them by choosing symbols which are not too hard to tell apart.
1
α-o u-n b-d p-b u-û f-t nt-nf ng-nj ch-h bh-sh ng’-ng
What are the challenges for the reader below? (Similar appearance of most of the
graphemes)
nduru
nturu
nguru
njuru
1
This section is taught later, in Lesson 4, in expanded form
123
Appendix C
8. The hardest letters for people to read are the digraphs (two or more letters for
just one sound). Your language will need many of these, but whenever you
can eliminate some of these and choose one letter from the Roman alphabet, I
recommend it.
For example, if you are writing bh and gh but you are not using the b and g at all,
I recommend using the b and g. If you do this, you can simply tell the reader that
in your language it makes a soft sound.
9. There may be a problem with this recommendation, if people really want their
language to be written like Swahili, with gh meaning the softer sound of g. But
after they try reading and writing with the longer spellings, they may complain
that it takes too long to write them this way, and it is harder to read the words,
especially when they are long.
What are challenges for writers?
• Long words
• Complicated spelling rules they must remember
• Words or morphemes that change spelling.
How do we help writers?
Give them spelling rules that are simple. In English, we always write some words the same,
even though they are pronounced dierently using phrase-level rules. For example,
We always say “The” as /thə/ before a consonant: “thə dog.”
But before a word beginning with a vowel, the same word is pronounced /i/. So we say
“thi apple.” Since it is actually the same word in the mind of the speaker, we make things
simple for both the reader and the writer and spell “the” the same. A native speaker will
almost always pronounce them according to the speech rules of his language. Let’s try it
with a native English speaker reading this list:
• the pencil
• the house
• the purple cow
• the orange
Did you hear the dierence? It is natural to any native English speaker, even a child,
though maybe not to you.
What are the challenges for new readers (review the list above)?
1. They’re working on matching sounds with symbols
2. They’re trying to move their eyes from left to right, remembering a series of sounds.
3. They put them together into wordsThey will sound out “Okusoma na
Okwandika Ekikwaya” and they may say, “This is not the way we say it. We say,
‘Okusomanokwandika Ekikwaya’.” What will you tell them?
124
Bantu Orthography Manual
What about mature readers?
Our goal is also helping them, because they will be the people who can actually read
something harder like the New Testament.
Mature readers recognize whole words, not just syllables or letters. If you write everything
connected, with some letters missing, as in rapid speech, these more advanced readers may
come to you complaining that it’s hard for them to read quickly and with understanding.
They are forced to read slowly and out loud, in order to understand. They will ask you to
make separate words of these things, so they can “see the meaning.”
125
Learning objective: People will demonstrate an increased awareness of the rationale
behind grapheme choices. They will make decisions which reect not only their attitudes
toward other languages and the sounds of their language, but the needs of beginning and
advanced readers.
1. ABCs…on Chalkboard: This is the ancient Roman alphabet, with a few
modications. They had 26 sounds in their language and used one letter for each
sound, a direct match.
2. What we try to do in any alphabet is to match up sounds and letters. You can use
one symbol per sound if your language has just as many vowels and consonants as
the Romans had. It doesn’t! All of your languages have more than 26 sounds.
3. Are there any simple alphabet letters we don’t need at all? (Erase Q, X, and V
from big board…)
4. What sounds do you feel you have in your language which aren’t represented
here? Invite TAs from the four groups to come up and write on their chalkboards
their unique letters!
• Ikizu has added two more letters for the vowels you need (Add ʉ and ɨ to the
Ikizu chalkboard).
• Simbiti has a vowel diphthong: what is it? Ai.
• Most of you have added several letters for the prenasalized stops, because each
one of them has its own sound: All of have(Add mb to the chart.)
• Most of you have (Add nd to the chart.)
• We have , etc…(Add ng to the chart.)
A B C D E F G
H I J K L M N
O P Q R S T U
V W X Y Z
126
Bantu Orthography Manual
5. What are some other symbols that we (or some of us) must add to our charts? Ny,
rr, ng’…
6. Are there still some Roman alphabet letters you are not using at all (i.e., and
)? (Are and really needed, or can you use a simple and a simple? In
answering these questions for yourselves, here are some things to think about:
Pro-B or G Pro-Bh or Gh
Shorter words for reader Show dierence from Swahili
Easily recognizable letters It’s what people expect
Faster writing Seeing the dierence between bh and h, for
example, can be hard for readers
Put on chalkboard, with some Kwaya writer (and another Kwaya reading them,
top to bottom):
Akagheera abhoobha okubhebhea
Akageera abooba okubebea
Ibhibhurughusi e kebhoroghwe ebhebhoroghwe
Ibiburugusi ekeborogwe ebeborogwe
You can all hear that the and when they stand alone, are soft sounds, softer
than what you hear in Swahili.
The and are your languages’ shy, quiet sounds, which become brave only
when a friend joins them (a nasal friend such as m or n)! They are actually the
same sound, which you will see when you arrange the consonants on your charts
at your tables: you will nd that the prenasalized hard sounds only occur after an
m or n, and the soft versions only occur when they are alone.
You can see that the roots are actually the same, in these examples. They
represent the same sound, but harden after a nasal:
Orubharu ‘ubavu’ sembaru ‘mbavu’
oroberi, ‘kijiji’ zemberi ‘vijiji’
Oribhubu ‘rigisi’ gimamba ‘hoja’
127
Appendix D
Write on the board and have people all read the following aloud. You can see
as you read across, that native speakers cannot help themselves. The soft letter
hardens after a nasal.
ighi (or ighi) ighi
agha agha
eghe eghe
ubhu ubhu
ibhi ibhi
ebhe ebhe
Can you hear the bravery of those sounds when they join a nasal? They are
the same sound, but they harden when they have their friend with them. It is
impossible for you to pronounce the g with a soft sound after m or n. (They are
the same sound and should be written with the same letter in both places.)
Conversely, when you speak Kikwaya, you will never say “Ariga” (with a hard
sound for g). When you read Kikwaya it should be the same. Call for a volunteer
Kwaya speaker. Instruct him to pronounce the bs and gs softly when he reads
them:
• Ariga oriwo omunibhi umwi abheye na omgasi waye na abhana bhaye na
omkasi waye owemirimu.
• Ariga oriwo omunibi umwi abeye na omgasi waye na abana baye na omkasi
waye owemirimu.
If you were to instruct native Kwaya speakers that they are reading their language,
not Swahili, they would probably very quickly learn to pronounce the words the
way Kwaya speakers naturally speak, making them soft when they are alone, and
making them hard after a nasal.
Should people write <> ‘fathers’ or , then?
Conclusion: Many Bantu languages have a soft sound for or when they are alone.
The sound becomes harder when a nasal is with it.
• Native speakers of your language will very quickly learn to say the soft sound of B
and G when they read aloud, because it is natural for them to do this. But—
• If people really don’t like writing just b or g, you may want to choose the digraphs,
or a symbol like β or Bh. We don’t want people rejecting their alphabet!!
128
Bantu Orthography Manual
Repeat choices:
Shorter words for reader how dierent from Swahili
Easily recognizable letters It’s what people probably expect
Faster writing Seeing the dierence between bh and h,
for example, can be hard
So this is your choice: B, β, or Bh, G, or Gh
129
As we have said, the challenges for readers and writers are:
1. long words: okwifunyongora
2. spelling rules: Only write a vowel which sounds long as long when it …
3. words and meaningful parts of words which change spelling (Example: ku-pasua,
ku-tengeneza, kulala, kwenda)
Some of the things we do to make reading easier:
1. Break up the stream of speech (Example on chalkboard):
Sinumuimkatemjinikwasababuhaufai
We don’t write it like we say it, because it is too hard for the person looking
rather than listening, to make sense of it!).
Instead, we break up that stream of speech into separate words, even though
when we speak it usually all sounds like one big loooong stream of speech. We
make this look like: Sinunui mkate mjini kwa sababu haufai.
2. We try to make these separate words look the same every time the reader sees
them. In English words, for example, we have question words. Almost all of them
start with the letters <wh>. This helps people recognize them as soon as they
see them written. We have The <>
beginning immediately signals a reader that this is a question word! So a simple
spelling rule helps readers get meaning quickly when they read.
3. We try to keep the spelling of words the same, even though their sound may
change when they are spoken fast in a sentence. For example, Americans may say,
“Waijaduthat?” (Write on chalkboard). We write it like this: Why did you do that?
But when we are really good readers reading aloud, we still say: “Waijaduthat?,”
not “Why did you do that?” And “gimmethupencil,” not “Give me the pencil.”
So, we spell words the way we speak when we’re talking slowly. We need writing
systems which show slow speech, not words crowded together, so the reader
always can recognize whole words.
Tharaka example: na or ni? In Kitharaka, na means ‘and’. Ni means ‘it is’.
The Tharaka people thought that if they could write just the way they speak, they
would combine words like this: na antü would be written <nantü> ‘and people’.
na antû ‘and people’ not nantû
ni wee ‘it is him’ not nwee
If both na and ni are written attached to the word which follows them, with the
vowel dropped, the reader will not be able to tell whether the little word was na
or ni. Also, if they are written together with the word which follows them, you are
130
Bantu Orthography Manual
combining a verb with a noun, which are two distinct parts of speech. The reader
can’t easily tell whether the writer meant to say “it is people” or “and people,”
because both would be written <nantû>. Can you see the solution? Writing them
both fully, as separate words. Will this be a problem for the reader? Only for a
beginner. The beginner will quickly learn to get meaning and naturalness when
reading fast, because he/she will recognize the meaning and be used to speaking
his own language naturally.
Try this with an example from some language spoken at the workshop.
131
1. Small words, especially grammatical ones, should always be written the same,
when possible (i.e., when native speakers aren’t acutely aware of a phonemic
distinction). Write them disjunctively when possible. If they’re attached
phonologically, try a hyphen with people.
2. Apostrophes should represent thing when possible. English uses them for
vowel elision and also for possessives, which confuses highly literate people all the
time. or (apostrophe problem).
• Multiple uses for it
• Its position, detached from the line of letters. Hyphen is sitting down with the
vowels…
Since we’re stuck (sociolinguistically) with ng’ in Eastern Africa, we should avoid
other apostrophes dangling in the air, and let them just be part of this trigraph.
3. Lake Region vowels—the devoiced Vs in 2
nd
syllable of CVCV roots—due to their
lack of stress, they may lose some of their contrastiveness, like English does in
unstressed syllables, where everything becomes [ə].
4. Any questions about distinguishing palatalized consonants from CV sequences?
(in Linguistic Features, Consonants, in Part 1, chapter 2) Consider phonotactic
rules of the language. Strive for spelling consistency and consider morpheme
preservation issues. Look at the Tharaka examples showing what often happens
to Vs after palatals. Tharaka Vowel “Coalescence” chart. Writing of the palatal
sound as a <y> versus as an <i>or<î>may depend on what the original V was
underlyingly, more than upon some phonological rule. Preserving a distinction
which would be otherwise lost, if both Vs, for example, in a 7-vowel system.
5. Palatal Nasals, p. 31-34. If your language has palatal, palatalized, and palatalized
palatal nasals, they probably need to be distinguished in the writing system
(especially if they occur frequently or their spelling aects people’s recognition of
axes within words).
6. Labialized consonants, same. [kwe] versus semivowel following a V, like [kuwe].
See p. 35-40. A phonological distinction phonologically may signal a crucial
distinction morphologically. Listen for tone changes, and ask your linguistic
consultant! She knows the morphemic clues. Here’s a phonological clue:
If the vowel after the possible semivowel sounds long, then it really is a
semivowel.
7. See Part 1 whenever you have questions, but most symbols will be obvious and
intuitive.
132
Instructional Objectives:
• Participants will be able to list their accomplishments from the last workshop
• They will be able to explain the rationale behind their spelling decisions from the
last workshop
• Participants will see the value of studying their morphology in making spelling rules
• They will be able to list what they expect to produce by the end of this workshop
Let’s review what we learned and accomplished last time, and then we will discuss the
plans for the next three weeks.
What we did last time:
• Week 1: Vowels:
– Vowel inventory (and Chart of Vowels lled, with example words in Part 2,
Appendix L, Orthography Workbook Form)
– Vowel combinations (chart lled; see Part 2, Appendix L, Orthography
Workbook Form)
– Long vowels where do they occur? ( occurrence in disyllabic word roots; see
Part 2, Appendix L, Orthography Workbook Form)
– We studied two kinds of words. What are they? Nouns nomino and verbs
vitenzi.
– Noun roots are identied (lling noun class charts, just like Swahili has). Put
up a sample noun class chart for one language group.
– Check spellings of wordlists, using cards and giving ourselves spelling rules
based upon certain rules for Bantu languages. Can you remember what
they were, for all of us? When do you automatically write a vowel as short?
(Before prenasalized stops: <nd, nj, ng, nz, mb …>). Does this apply to
<ny> and <ng’>? Why?
• Week 2: Consonants:
– We said that it is good to use only one symbol for each sound in your
language, but that sometimes this is impossible, both for writing and for
reading. For this reason most of you chose to write your soft consonant
sounds with one letter, such as <b> or <g>.
– We said that some sounds, actual letters in your alphabet, can harden or
soften, even though they are the same phoneme or letter in your alphabet.
This is what happens with
oRoot-initial consonants
oRoot-medial consonants
– We learned the possible locations of each consonant we noticed, after
studying the consonants in dierent positions in a word, that some change
in pronunciation. For instance, you never nd a [β ] after an /m/. It always
133
Appendix G
becomes a [b] sound in your languages, after an <m>. Example: You can
see that the roots are actually the same, in these examples. They represent
the same sound, but harden after a nasal:
Simbiti: ‘ubavu’ ‘mbavu’
Ikiizu: ‘kijiji’ ‘vijiji’
Kwaya: ‘rigisi’ ‘rigisi’ pl.
– Native speakers don’t usually notice this change. You may be aware of it
because your [β] is dierent from Swahili’s [b]. Readers of Kikwaya or
Kingoreme, though, will know which way to pronounce it correctly for
Kikwaya or Kingoreme.
– Choose key words and pictures for alphabet charts
– Revised spellings of your wordlists
– We included w and y in all of your alphabet charts—but now we will look at
them much more closely! I will explain this later.
• Week 3
– Tone study begun
– Noun class charts partially lled (show Ngoreme chart)
– Alphabet charts produced
– Short stories recorded, with transcription and word-by-word Swahili gloss
and phrase-by-phrase free translation
These will be your biggest challenges:
Our goals for this workshop:
• Find out people’s response to the wordlists
• Fix up our alphabet charts
• Discover some spelling rules and document them
• Together we will need to write down grammatical words and particles and list them
on charts. This time you will make your own charts, as well.
• Test the readability, comprehensibility, and acceptability of certain spelling choices
• Stories: Correct some errors in accurately transcribing what storytellers said. If we
have time, edit the spelling and punctuation of one story, to be printed out at the
end of the workshop
• We will pay a lot of attention to the places where meaningful parts of words come
together, because they make a big dierence to writers and readers. The linguistic
consultant will be helping us identify these parts.
The pastor walks into his oce and nds three people in it. He nds a note from
the church secretary on his desk. What is the problem with the message he reads?
– Wanazima vitabu. The secretary intended to say, “They are borrowing…”
What do you have to do to make the meaning clear?
– Mbwa wake anacha nyumba. Can you just write one <a> in these words?
oWanaambia: We can give it a completely dierent meaning by changing
134
Bantu Orthography Manual
only one tiny vowel: now it says wanaimbia.
oIn Kiikizu: ɨkɨrɨro (kilio) ɨkɨrɨɨro (mboga)
oIn Kwaya: risina (bulb) risiina (name)
Is some information missing? (And what happens to your understanding of these
sentences when just one letter is added?)
What makes a dierence in your understanding?
Did one little letter make a big dierence in your understanding of a whole word?
Break this word into parts: wa-na-ambia. (Now substitute wanaimbia.)
When viambishi come together, we often have spelling challenges. Karen will
explain these in detail, as we look at nouns rst, and later as we study verbs.
– You will see during this workshop that your vowels sometimes change and
become consonants (w or y) in some situations. Look at Noun class 1 prex
for kiNgoreme. For example, in Kingoreme, what vowel here became a w?
Noun class 1 prex omu-
prex + vowel becomes orthographic
result
example
word
Gloss
omu
*
+ a ‣ omwa omwame ‘bwana’
omu + e ‣ omwe omwenekere ‘mwenyeji’
*Danny Foster adds (pc), “We don’t know if [the] underlying [form] is omo or omu. I’ve chosen
the high vowel. The prex vowel is subject to vowel harmony rules when the root begins with
a consonant.”
Literacy Consultant’s job:
• Help nd solutions to reading-spelling challenges
• Help document what we learn
• Make good alphabet charts and other literacy materials
Linguistic Consultant’s job:
• Helping you discover the grammar of your language
• Helping you discover as many of your grammatical words as possible
• Uncovering the rules you aren’t even aware of, which make things work the
way they do in your language (like we did last time, in looking at your noun
class prexes) and what happens to them to make them take dierent forms…
Participants’ job:
• Give us examples from your language when we ask for them, as you did before
• Try reading things dierent ways (i.e., with a variety of spellings) and give me
feedback on which way makes understanding and spelling easiest for you
• You can make charts of how your grammar works after we have made the
discoveries. The more you learn about your grammar, the better you will spell and
135
Appendix G
read your language and the better you will explain spelling rules to others! We will
all make one big chart with each language group, to start with. Later, each of you
can ll in a chart with the same information—not yet.
Review the following orthography principles, if needed
Remember, your goals in developing your writing system:
1. Helping writers by making spelling easy, with simple rules
2. Helping new readers by using one symbol for each sound in your language
whenever possible
3. Helping new readers by breaking up the sounds into words and sentences. Example:
sikununuaviazimjinikwasababunilikosawakati Exercise: Break into words—
Sikununua viazi mjini kwa sababu nilikosa wakati.
4. Slow speech. When we talk to our friends, we usually drop some of the sounds in
words, knowing they will understand us. In English, when we talk fast, we say,
“G’bye.” But if we speak slowly, we are actually saying “Good Bye.” That is the way
we need to write it: “Good Bye.” We native English speakers say, “Dijuseeimatth
store?” But when we write, we break these sounds of fast speech into slow speech,
and really help the reader and writer to understand: “Did you see him at…” It is the
same for you. What does this (write it) mean to you? <Shkm>. Can you tell this is
“shikamoo”? Readers, unlike listeners, need all of the clues we use in slow speech to
help them understand.
5. Helping readers by making reading easy for them:
• symbols which are easy to distinguish (Example: what if all of our letters were
tall, like this: lltt!!ɺf?)
• shorter words, when possible
• giving the reader all the information he needs to be sure of the meaning of a
word (including tone, the correct vowel, length, etc.)
(Examples of contrastive vowel length and contrastive tone?)
• Helping readers understand what the writer wants to tell them (not just helping
them sound out words so they can be pronounced) (Example: Quisiera leer mi
idioma.) ¨Did you understand? No, of course, you still need to get meaning from
words in order to be truly reading. You’re not just making sounds. Have you
noticed that a really advanced reader can read without making a sound? He
recognized several letters and sometimes whole words at once.
6. Alphabets are just a tool, not the ten commandments from God, handed down
in stone in English! The letters do not even represent the same sounds for all
languages. You get to decide what sound each letter stands for, for your language.
Example: In Swahili, <ch> is the [ch] sound. In French, <ch> is the sh[š] sound.
Chien – dog. In Czeckoslovakian, they write it with a <cz>. The [ch] sound in
French is written this way: <tch> . In Swahili, [y] is written <y>, but in Spanish,
it is usually written <ll>, as in Me llamo.
136
Bantu Orthography Manual
Of course, we also want to make it easy for people who already read Kiswahili to
learn to read their mother-tongue, so we keep many Swahili symbols and sounds
the same in the mother tongue.
It was noted before that w and y are called ‘semi-vowels’. They have an inuence upon the
vowels that follow them by making them sound long, when in fact there is only one short
vowel present. For example, watho ‘law’ will sound as if a long aa sound is present. But we
know that the root of the word is -atha, to which is added the (abstract) noun class prex
û, which changes to w before a. Since the underlying form only has one vowel, only one
vowel will be written. This is still pronounced long.
The same applies to the other semi-vowel y. When this occurs before a single vowel, that
vowel will sound as if it is long. For example, yake ‘his’. This is made up of î + -ake. It
sounds long, but there is only one a vowel present in the underlying form. Here are some
other examples:–
When two identical vowels follow a semi-vowel, then both vowels will be written. This is
a long syllable and will sound long with or without the presence of the semi-vowel. For
example, ‘you built’ (distant past) is waakire. Here the dierent parts making up the word
are: û + a + aka + ire. Here we see that one vowel is the past tense marker, and the
other a is the rst letter of the word ‘build’. Other examples are:–
û +a + anjîîria = waanjîîria î + a + amba = yaamba
û + a + athaga = waathaga î + a + akire = yaakire
Also note that two vowels that join together to form one long vowel will also both be
written after a semi-vowel
wa + ûgu = woogu
ya + ûgu = yoogu
So basically, the rule is a simple one: where there is a SV syllable which only has one
vowel in the underlying structure, then only one vowel will be written, but where there
are two vowels in the underlying form, then two vowels will be written. In pronunciation
it is possible that both will sound the same.
137
Participants will make good decisions as to word-break rules,
regarding clitics and other grammatical words or particles which may be aected by
phrase-level rules in their pronunciation. They will understand the options presented to
them, and their choices will reect their perception of their language as well as readability
and writeability issues.
• chart listing word break
principles, in language of wider
communication (the chart
is partially covered, at the
beginning, so that all that shows
is the title: What is a Word?
• ipchart with paper, chalkboard,
or whiteboard (for examples)
When people rst begin to read, they sound out, letter-by-letter, syllable, without always
understanding what they read. Why? Because they are working very hard just to make
sounds which match letters.
A beginning reader looks at <> ‘grandchildren’ in Kikwaya, and by the
time he reaches the end of the word, he may not even remember what it all said, to put
the syllables together to make words. He’s just seeing letters and making sounds. That is
normal.
But if a reader gets lots of practice, he can in a few months start sounding out words faster
and putting those sounds together into words. Later, a really uent reader, such as the
people in this room, can just glance and see a whole word at once. This is especially true
for words which are frequently used. Shorter words, such as nyumba, mtu, wa, kubwa, mti,
ni, nani, je, etc., are used so often that a practiced reader recognizes them as a whole word.
He is not sounding out
Let’s think rst about what makes a word. A word:
1. It has meaning all by itself, though some words have more distinct meaning than
others. Nouns, like and, have lots of meaning on their own, and
when we think of them we have a picture in our minds. That’s why we used nouns
for your alphabet charts. (Talk about adjectives and numbers and the way they
can be picturable and isolatable.)
2. Most adjectives and pronouns also have distinct meaning, and you can still picture
them a bit: simple pronouns like , number words like , and even roots
of verbs, like - or -
1. It has meaning all by itself.
2. It can move around and be put in dierent
places in a sentence
3. Other words can come in between words
138
Bantu Orthography Manual
3. Most Bantu words, especially verbs, contain grammatical indicators which must
be there, in order to make sense. We will look at them next week. That means that
you have a part of a word which carries most of the “wordness.” In order to really
make sense in a Bantu language, you have to add something to the verbal root,
making something (mzizi) such as <>. Just briey, what information
do you get from this verb and its grammatical attachments? Many more things
can be added to a Bantu verb, giving much more information, but this much is
essential. This much grammatical information is usually essential to a Bantu verb
in order for it to make sense, so it is all part of one word (an ax, not a word).
4. Can the in communicate any meaning by itself? No, it needs the rest
of the word in order to make any sense at all.
5. A word can move around in a sentence. For example, take the word You
can put it at the front of a sentence: . You can put it after
a verbal phrase: . The in anacheka cannot
move from that exact place in the verb. It must be right there at the front, telling
the listener or the reader who is doing the laughing, and it has no meaning on its
own. As you probably know by now, in is an ax, not a word.
6. Words can be written separately when other words can come in between them.
For example, in Swahili, yain a sentence like Nyumba ya mdogo wangu ‘the house
of my younger sibling’, is it possible to put another word between and ?
Yes, you can say So it should be written
separately because it is possible.
7. Sometimes you have very little words in Bantu languages; they carry their own
little bit of meaning, but people always say them with a noun. The following is a
locative example from Nyiha. People recognize them as separate words because
the nouns nyumba and mbwa can stand all by themselves. It’s easy to recognize
and by themselves. But a spelling problem always arises when
the noun starts with a vowel, and you put the locative before it! Then you get
spellings like the noun becoming and Two words
have been written as one, because the vowels at the end of the locative changed
and so did the vowels at the beginning of the noun <>. Now the noun is
merged with another word and it’s hard for the reader to recognize the noun.
8. Locatives and associatives for your languages will probably turn out to be small,
separate words. Other words can intervene between them, they have meaning of
their own, and they’re movable.
What should we do to resolve the writing problem above? The Niha have choices:
1. Write the locative the same everywhere and write the noun the same everywhere.
Then you have <. Native speakers will
eventually learn to say the words naturally in uent reading: /
/, because this is how they talk.
2. Write the locative attached to the noun it describes, with a hyphen:
etc.
139
Appendix H
3. Write the locative combined with the vowel prex from the noun, like this:
This makes the noun look dierent, harder to recognize
now because its preprex has been removed. It also makes the locative harder for
the reader to recognize because the and are no longer there as
a clue.
4. Write all the changes as people say them, with the locative attached to the noun it
preceded: <>. It makes the noun harder for people to read, because it is no
longer standing alone, and it always looks dierent because it has these locatives
stuck onto it in such a way that it’s hard to cut apart the locative from the noun—
but it’s written the way it sounds, making it easy for a new reader to sound out,
letter by letter. It will slow reading for a uent reader.
5. The rst two options are the best. Today we will be looking at locatives,
associatives, and possessives and tackling these problems for writing. (You
actually have the same issue with your prexes on your nouns.)
Any questions?
140
• People remember their writing rules.
• They can apply them to editing a story.
• They test their rules by reading the story aloud.
• They note any unnoticed linguistic elements which need to be documented, while
editing.
• A handout chart for each group to put up, like the Ngoreme one at bottom.
• Review the concepts we taught earlier, re: what makes a word and the reason for
word breaks.
• Look at their charts to see what spelling rules they have established already.
List them, for:
• : Example from Ngoreme:
ko-mugabo mo-mugabo
ko-moona mo-moona
ko-monto mo-monto
•
• Copulas—do they want to write the copula as attached, but to write the na
conjunctive separately?
Focus marker /n-/ is written adjoiningly. There seems to never be an underlying
vowel with it. Since they have the copula n- which is very snugly attached to
the following noun or adjective, do they always want to write it adjoiningly?
Or do they want to write it with an apostrophe after it? It will look just like the
conjunction if they write them both at the front of a word (usually a noun) with
no vowel after either one.
• Augments/diminutives—nothing unusual
• Negatives—written as part of the verbal phrase (tu)
141
Appendix I
• Other, including tone: extensions (causative and passive are interesting—they are
just i and u, respectively)
• Vowel length? Ask us when problems arise, and we’ll talk! At issue: simplicity of
spelling rules versus morphology. Kwaya example: woone.
-rogota -agura -engeri -ikara -ogu -u
‘-tambaa’ ‘-ongezea’ ‘-kaa’ ‘-sikia’
twakarogota ‘twaokota’ twakaagura twakeengeri twagaikara twakoogu
turarogota ‘tunaokota’ turaagura tureengeri turaikara turoogu
nturogotire ‘tumeokota’ ntwagurre ntwengeriiri ntwikaire ntoogure
ntwarogotire ‘tuliokota’ ntwaagurre ntweengeriiri ntwaikaire ntoogure
tukarogota ‘tukaokota’ tukaagura tukengeri tugaikara tukoogu
tokurogota ‘tukiokota’
n-tu-rogot-ire
n-tu-ogu-ire => ntoogure
n-tu-a-rogot-ire => ntwarogotire
n-tu-a-ogu-ire => ntoogure
Sheria za Kuandika Kingoreme
1. Mahali: (ko-mugabo, mo-mugabo, mo-moona,
ko-moona, ko-monto)
2. –a unganifu: (wo bo go gyo) and (wa, ba,
gwa, gya) and (we, be, gwe, gye)
3. Kitenzi shiritisho (ni): (before C? before V?)
4. Negatives: ??
5. Vowel length? Write short before prenasalized
consonants and after semivowels. Violate this rule
only: “When you must show something which makes
a grammatical dierence in a verb…”
142
1 sg mʊ- jʊ- ʊ-
2 pl ba- bá- bá-
3 sg mʊ- gʊ- ʊ-
4 pl mɪ- gɪ- ɪ-
5 sg i- lɪ- lɪ-
6 pl ma- gá- á-
7 sg kɪ- kɪ- kɪ-
8 pl bi- bi- bi-
9 sg N- jɪ- ɪ-
10 pl N- jɪ- í-
11 sg lʊ- lʊ- lʊ-
12 sg ka- ká- ká-
13 pl tʊ- tʊ- tʊ-
14 sg/pl bʊ- bʊ- bʊ-
15 sg kʊ- kʊ- kʊ-
16 locative pa- pá- pá-
17 locative kʊ- kʊ- kʊ-
18 locative mʊ- mʊ- mʊ-
19 pi- pí- pí-
20 gʊ- gʊ- gʊ-
21 gi- gí- gí-
22 ga- gá- gá-
23 locative ɪ- ɪ- ɪ-
These are a synthesis of Schadeberg 2003:149 and Katamba 2003:104.
143
Appendix J
mʊ- ba-
Ø Ø
mʊ- mɩ
i- ma-
kɩ- bɩ-
N- N-
lʊ N-
ka- tʊ-
bʊ- ma-
kʊ- ma-
pa- --
kʊ- --
mʊ- --
pi- --
gʊ ga-
gɩ- --
ɩ --
144
Bantu Orthography Manual
The chart below (Katamba 2003), in addition to the simplied Proto-Bantu chart of noun
class prexes, may help the linguist recognize the noun class prexes for a given language.
1 Human beings
1a Proper names
Kinship terms
Personications
2 Regular plural of class 1
2b Regular plural of class 1a
3 Natural phenomena
Body parts
Plants
Animals
4 Regular plural of class 3
5 Natural phenomena
Animals
Body parts
Collective nouns
Undesirable people
Augmentatives
Derogatives
6 Regular plural of classes 5 and 14
Mass terms and liquids
Time references
Mannerisms
Modes of action
7 Body parts
Tools, instruments, and utensils
Animals and insects
Languages
Diseases
Outstanding people
Amelioratives
Derogatives
Augmentatives
Curtatives (shortness, stoutness)
Mannerisms
8 Regular plural of class 7
9 Animals
People
Body parts
Tools, instruments, and household
eects
10 Regular plural of class 9
11 Long, thin entitities
Languages
Body parts
Natural phenomena
Implements, utensils, and other
artifacts (“artefacts”)
12 Augmentatives
Derogatives
Diminutives
Amelioratives
13 Regular plural of class 12
14 Abstracts
Collectives
15 Innitives, a few paired body parts
16 Location terms
17 Location terms
18 Location terms
19 Diminutives
20 Derogatives
Augmentatives
Diminutives
Amelioratives
Mannerisms
21 Augmentatives
Derogatives
22 Plural of class 20
23 Location terms
145
by
Participants of Duruma Writers’ Workshop
together with
Constance Kutsch-Lojenga
at Kinango, Kenya
(updated version: November 1994)
mu + i → thief
mu + e → clever person
mu + a → child
mu + o → saviour
mu + u → creator
a + i → thieves
a + e → clever people
a + a → children
a + o → saviours
a + u → creators
mu + i → body
mu + e → month
mu + a → tree, sp.
mu + o → re
mu + u → river
mi + i → bodies
mi + e → months
mi + a → trees, sp.
mi + o → res
mi + u → rivers
ø + i → wave
ø + e → mango
ø + a → fruit of eggplant
ø + o →
ø + u → ower
146
Bantu Orthography Manual
ma + i → waves
ma + e → mangos
ma + a → fruits of egg pl.
ma + o →
ma + u → owers
chi + i → mortar
chi + e → tool
chi + a → hanging basket
chi + o → tr. house length
chi + u → frog
vi + i → mortars
vi + e → tools
vi + a → hanging baskets
vi + o → tr. house lengths
vi + u → frogs
lu + i → spleen
lu + e → blade
lu + a → grinding stone
lu + o → sharp point
lu + u → winnowing tray
u + i → song
u + e → razor blade
u + a → food?
u + o → brain
u + u → our
ku + i → to sing
ku + e → to go
ku + a → to say
ku + o → to save
ku + u → to ask
147
Appendix K
N + b → mb goat
N + v → Mv rain
N + d → nd he-goat
N + dz → ndz elephant
N + j → ɲj path
N + g → ŋg drum
N + gb → ŋmgb crocodile
N + z → nz groundnuts
N + l → nd tongues
N + p → pH rat
N + t → tH girae
N + ts → tsH leopard
N + tʃ → tʃH hide, skin
N + k → kH chicken
N + kp → kpH octopus
N + ß → pH rhinoceros
N + m → m cassava
N + n → n coconut
N + ɲ → ɲ star
N + ŋ → ŋ cow
N + f → f kidney
N + s → s watch, clock
N + ʃ → ʃ shirt
N + h → h yeast
N + i → ɲi spleens
N + e → ɲe razor blades
N + a → ɲa grinding stones
148
149
Appendix L
Acknowledgments
Dr. Constance Kutsch-Lojenga was my original advisor for documentation of the phonology
of each language to be studied. After that important planning stage, Karen van Otterloo
added to the charts we designed, providing many more, especially for studying and
recording morphology. Finally, I have attempted to include a simple means of recording
spelling decisions and their rationale in the documentation process, with inspiration from
Danny Foster, Training Coordinator for SIL, Uganda-Tanzania Branch. I’m grateful to all
three of you!
Leila Schroeder
150
Bantu Orthography Manual
153
Vowels ................................................................................................................................153
Table 1: Vowel Inventory ..............................................................................................153
Table 2: V1 - V2 combinations chart ............................................................................154
Table 3: Long Vowel occurrence in disyllabic word roots ............................................155
Table 4: Vowel Length ..................................................................................................156
Table 5: Vowel Harmony ..............................................................................................157
Consonants .........................................................................................................................158
Consonants: Root-initial ................................................................................................158
Consonants: Root-medial ..............................................................................................159
List of examples for each consonant phoneme ..............................................................160
Double articulated stops ...............................................................................................162
Labialized Consonants ...................................................................................................162
Geminate consonants ....................................................................................................162
Palatalized Consonants .................................................................................................162
Syllable structures ..............................................................................................................163
Non-Word Final Syllables..............................................................................................163
Word nal syllables ......................................................................................................163
Tone ...................................................................................................................................164
Tone: Verbs ...................................................................................................................164
Tone: Nouns ..................................................................................................................165
Noun Classes ......................................................................................................................165
Noun class prexes with Consonant-Initial Roots .........................................................165
Non-standard Noun Class Pairings ................................................................................169
Noun Class Prexes with Vowel Initial Roots ...............................................................170
Morphophonology ..............................................................................................................182
Vowel Coalescence (word-internally) at morpheme boundaries ...................................182
Vowel Elision ................................................................................................................183
Noun Phrase .......................................................................................................................183
Demonstratives .............................................................................................................183
Demonstrative forms: other ..........................................................................................187
Interrogative Forms.......................................................................................................189
Pronominal Forms 1 ......................................................................................................191
151
Appendix L
Pronominal Forms 2: Quanticational ..........................................................................193
Pronominal Forms 3: Set pronouns and related forms ..................................................195
Adjectives (Qualiers) ..................................................................................................197
Adjectives (Quantiers) ................................................................................................199
Associative Constructions .............................................................................................201
Diminutives and Augmentatives ...................................................................................203
Numerals .......................................................................................................................209
Locatives .......................................................................................................................211
Possessive pronouns .....................................................................................................216
Use of the Augment on Nouns ......................................................................................218
Verbs (Verbal Axation) ...................................................................................................219
Copular Forms ...............................................................................................................220
Morphophonological Processes In Verb Stems With Perfective Final Sux .................222
Verbal Tense/Aspect: Imperative Plural .......................................................................225
Verbal Tense/Aspect: Imperative Singular ....................................................................227
Verbal Tense/Aspect: Simple Subjunctives ...................................................................229
Verbal Tense/Aspect: Subjunctives – Distal ..................................................................231
Verbal Prexes: Subject ................................................................................................233
Verbal Prexes: Object Prexes ....................................................................................236
Verbal Prexes: Object Relative ....................................................................................239
Verbal Prexes: Subject Relative ..................................................................................241
Verbal Tense/Aspect: Future 1 ......................................................................................243
Verbal Tense/Aspect: Future 2 ......................................................................................245
Verbal Tense/Aspect: Future 3 ......................................................................................247
Verbal Tense/Aspect: Past 1 .........................................................................................249
Verbal Tense/Aspect: Past 2 .........................................................................................251
Verbal Tense/Aspect: Past 3 .........................................................................................253
Verbal Tense/Aspect: Past 4 .........................................................................................255
Verbal Tense/Aspect: Present - Unmarked ....................................................................257
Verbal Tense/Aspect: Present 2 ....................................................................................259
Verbal Tense/Aspect: Other: __________________________ .........................................................261
Verbal Tense/Aspect: Anterior ......................................................................................263
Verbal Tense/Aspect: Past Anterior ..............................................................................261
Verbal Tense/Aspect: Present Progressive ....................................................................263
152
Bantu Orthography Manual
Verbal Tense/Aspect: Present Persistive .......................................................................269
Verbal Tense/Aspect: Consecutive ................................................................................271
Vowel Harmony in Extensions ......................................................................................273
Chart Checklist ...................................................................................................................275
153
Appendix L
Name of language ______________________________________________
Please read chapters 1 and 2 of Part 1 before beginning the process of orthography
development. Chapter 3 will guide you in choosing of symbols and word breaks.
_________ ______ ______
________
List symbol choices below. Add rationale for any non-Roman symbols, at bottom of the
page.
Sample
154
Bantu Orthography Manual
This chart shows the possible vowel combinations in disyllabic word roots. List all the
vowels of the language across the top and side. The vowels listed on the left side are those
which come rst; those along the top come second.
Write the rst ve or seven vowels down the left column, and the same vowels across the
top. The sample words written in each box will have a Swahili gloss. They should all have
disyllabic roots, with series of 2 vowels. The rst table gives a partial example from a
hypothetical language.
V2
V1
a e i o u
a
litata
mbo
kitate
kikapu
kitati
kiazi
kitato
kiatu
kitatu
kifaru
e kiteta
hotuba
kitete
ndege(namna)
kiteti
samaki
(namna)
luteto
mti (namna)
kitetu
mto
i
o
u
155
Appendix L
Write long vowels in rst column. Give one example for each vowel of the language,
with CV following the long vowel syllable and another example in which an NCV syllable
follows. Follow the orthography rules that have been decided upon for vowel length. Make
the two examples as similar as possible, e.g., i… omu-siiya… omu-simbe.
156
Bantu Orthography Manual
Vowel length: Is it contrastive (see Part 1, chapter 2, Vowel Length)?
Yes ___________ No _____________
157
Appendix L
Vowel harmony: Yes ________ No _______________
Give examples:
Look for examples of a verb in the given verbal extension, with i, e, etc., in the stem, to see
what other vowels appear in the following syllables
kutenda
kuimba kupenda
kutendea kuimbia kupendea
kutendesha kuimbisha kupendeza
kutendwa kuimbwa kupendwa
kutendeka
kutendana
kutendua
kutenduka
kutendama
atendire
158
Bantu Orthography Manual
(see Part 1, chapter 2, Consonants)
Consonants of _______________________________________
Write the grapheme chosen for your language in the appropriate box of the chart below.
Manner of articulation Voicing/
nasalization
Labial/
labio-
dental
alveolar Palatal/
alveo-
palatal
velar/
glottal
Implosives voiceless
prenasalized
voiced
prenasalized
Stops
and aricates
voiceless
prenasalised
voiced
prenasalised
Fricatives voiceless
prenasalised
voiced
prenasalised
Sonorants nasal
oral
Consonants which can be labialized:
Consonants which can be palatalized:
Limited distribution of vowels following these consonants, CVCV:
159
Appendix L
Manner of articulation Voicing/
nasalization
Labial/
labio-
dental
alveolar Palatal/
alveo-
palatal
velar/
glottal
Implosives voiceless
prenasalized
voiced
prenasalized
Stops voiceless
prenasalised
voiced
prenasalised
Fricatives voiceless
prenasalised
voiced
Prenasalised
Sonorants nasal
oral
List any consonants which only occur root-medially, with examples:
Describe any complementary distribution of consonants in the two positions:
160
Bantu Orthography Manual
Write the letter on the left, and on the right, a sample word with underlined syllable in
focus. Each syllable has the vowel /a/
p kupanga to rent
161
Appendix L
162
Bantu Orthography Manual
(see Part l, chapter 2, Double articulated stops)
(see Part 1, chapter 2, Labialization and palatalization)
(see Part 1, chapter 2, Geminate consonants)
(see Part 1, chapter 2, Labialization and palatalization)
163
Appendix L
Root initial syllable structures which occur within words
(Prexes are irrelevant. Ignore them.)
baata duck
164
Bantu Orthography Manual
(see Part 1, chapter 2, Principles of Tone Marking in an Orthography)
What is the tone of the innitive prexes:
on CVC or longer verbs? Augment________________ Prex ________________
on -CV verb stems? Augment________________ Prex _________________
on -VCVC verb stems? Augment________________ Prex _________________
Describe tonal melodies on C-initial verbs, including basic patterns and any conditioned
variation(s) of those patterns due to long vowels or other phonological phenomena. Give
examples, with pitch traces of each pattern (use a separate page if necessary).
Note whether H is found
a) on rst syllable of verb stem__________________________________________________________________
b) on a specic syllable position (i.e., penultimate or antepenultimate)__________________________
c) other
In the same context as the last question, note whether H “spreads” onto following
syllables._____________
If so, how far does it spread?
a) One syllable?__________
b) One mora?_____________
c) Two syllables?________
d) Up through a certain syllable position (e.g., antepenultimate or
penultimate)?_______________________________________________________________________________________________
Note whether there are any dierent patterns, or if the normal patterns found on the
V-initial verbs are merely “shifted leftward” one syllable.
165
Appendix L
What is the tone of prexes on CVCV (or longer) nouns?______________________________________________
Give examples, with pitch traces, of each:
How many tone patterns are found in nouns?_______ Give examples, with pitch traces, of
each.
Are there any dierent patterns found only in a certain noun class? If so, describe them
below.
This section lists only those noun classes that exist in the language. Please record prexes
before a .
m mtoto child
Mswahili Swahili person
wa watoto children
Waswahili Swahili people
166
Bantu Orthography Manual
167
Appendix L
168
Bantu Orthography Manual
169
Appendix L
------- bheesina namesake
------- nyabhutama type of snake bhi bhinyabhutama type of snakes
------ wasongo chameleon ga gawasongo chameleons
170
Bantu Orthography Manual
Write the vowel list for your language in the “vowel” column. Now go back to the “prex”
column and ll in the prexes for noun class 1. Chart the changes which take place when
each vowel meets that noun class prex.
mû + i ‣ mwi mwiki bride
mû + î ‣ mwî mwîtîgua trustworthy person
mû + e ‣ mwe mwekûrû woman, lady
mû + a ‣ mwa mwarimo teacher
mû + u ‣ muu ----------- -----------
mû + û ‣ mûû mûûgî clever person
mû + o ‣ mwo mwokozi saviour
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
171
Appendix L
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
172
Bantu Orthography Manual
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
173
Appendix L
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
174
Bantu Orthography Manual
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
175
Appendix L
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
176
Bantu Orthography Manual
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
177
Appendix L
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
178
Bantu Orthography Manual
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
179
Appendix L
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
Notes regarding class 9/10 prexes: these may or may not be identical to each other,
depending on the language. Nasals in these prexes may be dropped out either preceding
a certain phonological class of segments (in which case the lling of the noun class chart
in the manual will make the environments clear) OR just in random places not denable
in any phonological terms, in which case the language has a class 9a/10a which will
probably include a good number of the loanwords found in class 9/10. Also, a class 9 noun
which begins with Augment (- or -) may either have a vowel initial root (since the
- is the usual form of the 9/10 prex preceding a vowel initial root) or actually have
an <ny> initial root, since the nasal of the prex also drops out preceding a root which
begins with a nasal consonant (<m>, <n>, <ny>, or <ng’>). In order to tell whether
the <ny> is actually part of the root, elicit the diminutive form (“Ask what a very small
...... would be called) and if the <ny> still shows up with the diminutive prex, then the
root should be interpreted as <ny> initial.
180
Bantu Orthography Manual
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+
+ ‣
+ ‣
181
Appendix L
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
+ ‣
182
Bantu Orthography Manual
+ i ‣
+ ‣
+ ‣
+ ‣
+ ‣
Vowel Coalescence chart (see Part 1, chapter 2, Vowel Coalescence):
This chart shows the changes that take place when two vowels occur next to each other
across morpheme boundaries. List all the vowels of the language across the top and side.
The vowels listed on the left side are those which come rst; those along the top come
second.
183
Appendix L
(see Part 1, chapter 2, Vowel Elision (Vowels)
List examples of adjoining words which will be written as one, with an apostrophe to
represent a vowel which will not be written.
1 mtu huyu huyo yule
2 watu hawa hao wale
184
Bantu Orthography Manual
Class 1: mtu yule mtu -----
Class 2: watu wale watu---
Class 1
Class 2
Class 3
Class 4
185
Appendix L
Class 5
Class 6
Class 7
Class 8
Class 9
Class 10
Class 11
Class 12
Class 13
Class 14
Class 15
Class 16
Class 17
Class 18
Class 19
Class 20
186
Bantu Orthography Manual
(Please give a few examples)
187
Appendix L
Record any other demonstratives found in the language for every noun class. There will
probably be more emphatic forms than will t on this page. Just record a couple to show
the basic morphology involved:
Notes:
188
Bantu Orthography Manual
(Please give a few examples)
189
Appendix L
List here the question words corresponding to the following glosses:
Is there a Question marker that is used before an interrogative clause, like je in Swahili? If
so, list and give an example with a gloss:
190
Bantu Orthography Manual
(Please give a few examples)
191
Appendix L
Record pronouns of the following type for every noun class:
---------
---------
----------
----------
192
Bantu Orthography Manual
(Please give a few examples)
193
Appendix L
Record pronouns of the following type for every noun class:
------ ------- ---- --------
------- ------- ---- --------
---------
---------
194
Bantu Orthography Manual
(Please give a few examples)
195
Appendix L
Record pronouns of the following type for every noun class:
-------- --------
-------- --------
--------- ---------
--------- ---------
196
Bantu Orthography Manual
(Please give a few examples)
197
Appendix L
mtu mrefu mwerevu m-, mw-
watu warefu werevu wa-, w-
198
Bantu Orthography Manual
(Please give a few examples)
199
Appendix L
1 mtu ----- yeyote ye-
2 watu wote wowote wo-
200
Bantu Orthography Manual
(Please give a few examples)
201
Appendix L
mwalimu wa mtoto
walimu wa mtoto
202
Bantu Orthography Manual
(Please give a few examples)
203
Appendix L
mûkûrû older man kamûkûrû kîmûkûrû
(mw)ana child kamwana kîmwana/kîana
(mu)ntû person kamuntû- kîmuntû
204
Bantu Orthography Manual
205
Appendix L
206
Bantu Orthography Manual
207
Appendix L
208
Bantu Orthography Manual
(Please give a few examples)
209
Appendix L
1 mtu mmoja -- -- -- m-
2 watu -- wawili watatu wa-
210
Bantu Orthography Manual
(Please give a few examples)
211
Appendix L
(Sample from Niha)
nyumba pa=nyumba ku=nyumba mu=nyumba
ikwi pikwi kwikwi mwikwi
mbwa pa=mbwa ku=mbwa mu=mbwa
212
Bantu Orthography Manual
213
Appendix L
214
Bantu Orthography Manual
215
Appendix L
(Please give a few examples)
216
Bantu Orthography Manual
mtu
watu
217
Appendix L
(Please give a few examples)
218
Bantu Orthography Manual
1. Does the augment (i.e., pre-prex, the initial vowel of the nominal prex) usually
appear in nouns used in isolation (i.e., nouns listed without any context)? YES/NO
(Circle one.)
Give two or three examples (one with each dierent augment vowel) with glosses:
Note any general exceptions, e.g., a certain class in which the augment is used
dierently (you may cross reference this with another chart if desired):
2. If the augment is generally used on nouns in isolation, can it be removed? If so, can
people describe any dierence in meaning between the form with and without the
augment?
3. Is the augment used in the noun phrase on words other than nouns, e.g., adjectives,
numerals, etc.? To determine this, translate the following:
Mti mrefu:
Ni mti:
Ni mti mrefu.:
Mti ni mrefu.:
Mti mrefu mmoja ulianguka.:
Waliangusha miti yote.:
Hawakuangusha miti yoyote.:
219
Appendix L
Inectional
subject prexes (for participants and for classes)
object prexes (for participants and for classes)
relative prexes (for participants and for classes)
T/A markers
negative markers
any other inectional elements particular to a language
Derivational
Verb-verb:
Productive derivational suxes: causative, passive, applicative, reciprocal
Combinations of these
Changes in verbal aspectual suxes because of the use of (combinations of)
derivational suxes
Verb-Noun: agentive nouns derived from verbs
Verb-Adjective: adjectives derived from verbs
220
Bantu Orthography Manual
Notes:
221
Appendix L
(Please give a few examples)
222
Bantu Orthography Manual
In the rst part of the chart, for each consonant in the phonemic inventory of your
language, nd a verb stem ending with that consonant. List the consonant in the rst
column, the innitive and its gloss in the second and third columns, and a verb form
with perfective ending (usually can be elicited using me- tense in Swahili, e.g., ameketi,
etc.) in the fourth column. Note in the last column any changes that occur in the stem-
nal consonant. For example, in Kifuliiru, in a verb stem ending with <k>, the <k> is
changed to <s> in the perfective form: kubiika ‘to put’, abiisiri ‘he has put’.
For the second part of the chart, nd a verb ending with each of the extensions listed and
note any changes in the extension in the perfective forms. In some extensions, the nal
consonant may drop before the “perfective” ending. Vowel coalescence may then take
place.
223
Appendix L
Notes:
224
Bantu Orthography Manual
(Please give a few examples)
225
Appendix L
Mark the placement of any H, rising, or falling tones on all forms.
“Stem” below refers to all parts of the verb to the right of any object prex. E.g., The
innitive kushiba has a two-syllable stem, while kula has a one-syllable stem, and kukalia
has a three-syllable stem. If there are two (or more) tone classes of verbs in your language,
include one of each. If a verb with a long vowel exhibits a slightly dierent surface tonal
pattern, include such a verb in the table below, along with the negative form of one verb
of each syllable count.
226
Bantu Orthography Manual
(Please give a few examples)
227
Appendix L
Mark the placement of any H, rising, or falling tones on all forms.
“Stem” below refers to all parts of the verb to the right of any object prex. E.g., The
innitive kushiba has a two-syllable stem, while kula has a one-syllable stem, and kukalia
has a three-syllable stem. If there are two (or more) tone classes of verbs in your language,
include one of each. If a verb with a long vowel exhibits a slightly dierent surface tonal
pattern, include such a verb in the table below, along with the negative form of one verb
of each syllable count.
228
Bantu Orthography Manual
(Please give a few examples)
229
Appendix L
Mark the placement of any H, rising, or falling tones on all forms.
“Stem” below refers to all parts of the verb to the right of any object prex. E.g., The
innitive kushiba has a two-syllable stem, while kula has a one-syllable stem, and kukalia
has a three-syllable stem. If there are two (or more) tone classes of verbs in your language,
include one of each. If a verb with a long vowel exhibits a slightly dierent surface tonal
pattern, include such a verb in the table below, along with the negative form of one verb
of each syllable count.
230
Bantu Orthography Manual
(Please give a few examples)
231
Appendix L
(e.g., go and do x - uende ukafanya....) Mark the placement of any H, rising, or falling
tones on all forms. “Stem” below refers to all parts of the verb to the right of any object
prex. E.g., The innitive kushiba has a two-syllable stem, while kula has a one-syllable
stem, and kukalia has a three-syllable stem. If there are two (or more) tone classes of
verbs in your language, include one of each. If a verb with a long vowel exhibits a slightly
dierent surface tonal pattern, include such a verb in the table below, along with the
negative form of one verb of each syllable count.
232
Bantu Orthography Manual
(Please give a few examples)
233
Appendix L
Using your list of example nouns from each noun class, elicit a simple subjunctive verb
form with a subject marker that agrees with each class. It may be necessary to switch verbs
occasionally for reasons of semantic non-compatibility with subjects of certain classes, but
make an eort to keep the verb in each column the same if possible (i.e., ideally, use only
three dierent verbs for this chart - one verb for each column). A passive or “neuter” verb
form, if a suitable one can be found, may be more likely to be semantically compatible
with multiple subjects, e.g., “may x be found,” or “may x appear,” etc. Using a simple
subjunctive form ensures that the subject prex appears right next to the verb root, e.g., 1
st
singular forms in Swahili might be that/may I arrive, that/may I go,
that/may I pray, etc.
234
Bantu Orthography Manual
If there is any vowel coalescence going on between a subject prex and a vowel-inital root,
nd a verb to represent each vowel which can be found initially in a verb root, and ll the
entire paradigm (i.e., one verb which begins with <i>, another verb which begins with
<e>, another beginning with <o>, etc.). Use another copy of this chart if necessary.
If any variation in the form of the prexes is conditioned by the following (verb root-
initial) consonant, make note of that below, pointing out an example (above) for each
variation. Such conditioned variants will normally only involve a prex that consists
only of a nasal consonant (usually rst person singular, e.g., rst person singular prex
- appears as - when it occurs directly preceding a verb root beginning with a bilabial
consonant.
Notes:
235
Appendix L
(Please give a few examples)
236
Bantu Orthography Manual
Fill this chart using transitive verbs with rst plural subject so that each gloss is “we ...
[ourselves],” “we ....... [you (sg)],” “we .... [him],” etc. The exception to this pattern is that
in the elicitation of rst person object markers, a second person plural subject should be
used (“you ...me,” “you......us”). The RFLX row should be done with rst person plural
subject—“we .... ourselves.” For each vertical column, use only one verb root, and do not
change tenses. The form should be either a subjunctive or a simple past, corresponding to
the Kiswahili tulijipiga, tulikupiga, tulimpiga, etc.
Fill the entire paradigm for a verb representing each initial vowel you nd in verb roots
(i.e., one verb which begins with i, another verb which begins with e, etc.). If any variation
in the form of theprexes is conditioned by the following (verb root-initial) consonant,
make note of that below, giving an example for each variation. Such conditioned variants
will normally only involve a prex that consists only of a nasal consonant (usually rst
person singular).
Note any object prexes that seem to have H tone. An H tone with a rst person singular
object of the shape N- will often be realized on a preceding or syllable.
237
Appendix L
If it is possible to have more than one object prex in the verb, give some examples: we
gave it (food) to him, we gave it (food) to it (e.g., dog, cow, etc.). If this is not possible, see
if it is possible to have two objects if one of them is reexive, if one of them is rst person
(singular or plural), if one of them is second person (singular or plural).
Notes:
238
Bantu Orthography Manual
(Please give a few examples)
239
Appendix L
Using your list of example nouns from each noun class, elicit a simple past verb form with
a rst person plural subject (we) with a relativized object that agrees with each class.
These will be equivalent to phrases like (the person)
whom I saw, (the people) whom we saw, (cl.)
which we saw, ... etc. In a language other than Swahili, the relativized object may or may
not need to be repeated just before the verb stem. In order to elicit these, it may be helpful
to put the clause into a sentence such as: This is the .........(insert noun of the appropriate
class) which we saw. List below only the part of the sentence translatable as “whom/which
we saw.”
Notes:
240
Bantu Orthography Manual
(Please give a few examples)
241
Appendix L
Using your list of example nouns from each noun class, elicit a simple past verb form
with a relativized subject agreeing with each person and class. These will be equivalent to
phrases like I who fell, you who fell, he who fell, ...
(the tree) which fell, etc. List these in the rst column. In the second column,
list the same forms but in simple future tense I who will fall, etc.. In the
third column, list negative forms: , , etc.
Notes:
242
Bantu Orthography Manual
(Please give a few examples)
243
Appendix L
Mark the placement of any H, rising, or falling tones on all forms.
“Stem” below refers to all parts of the verb to the right of any object prex. E.g., The
innitive has a two-syllable stem, while has a one-syllable stem, and
has a three-syllable stem. If there are two (or more) tone classes of verbs in your
language, include one of each. If a verb with a long vowel exhibits a slightly dierent
surface tonal pattern, include such a verb in the table below, along with the negative form
of one verb of each syllable count.
Notes:
244
Bantu Orthography Manual
(Please give a few examples)
245
Appendix L
Mark the placement of any H, rising, or falling tones on all forms.
“Stem” below refers to all parts of the verb to the right of any object prex. E.g., The
innitive has a two-syllable stem, while has a one-syllable stem, and
has a three-syllable stem. If there are two (or more) tone classes of verbs in your
language, include one of each. If a verb with a long vowel exhibits a slightly dierent
surface tonal pattern, include such a verb in the table below, along with the negative form
of one verb of each syllable count.
Notes:
246
Bantu Orthography Manual
(Please give a few examples)
247
Appendix L
Mark the placement of any H, rising, or falling tones on all forms.
“Stem” below refers to all parts of the verb to the right of any object prex. E.g., The
innitive has a two-syllable stem, while has a one-syllable stem, and
has a three-syllable stem. If there are two (or more) tone classes of verbs in your
language, include one of each. If a verb with a long vowel exhibits a slightly dierent
surface tonal pattern, include such a verb in the table below, along with the negative form
of one verb of each syllable count.
Notes:
248
Bantu Orthography Manual
(Please give a few examples)
249
Appendix L
Mark the placement of any H, rising, or falling tones on all forms.
“Stem” below refers to all parts of the verb to the right of any object prex. E.g., The
innitive has a two-syllable stem, while has a one-syllable stem, and
has a three-syllable stem. If there are two (or more) tone classes of verbs in your
language, include one of each. If a verb with a long vowel exhibits a slightly dierent
surface tonal pattern, include such a verb in the table below, along with the negative form
of one verb of each syllable count.
Notes:
250
Bantu Orthography Manual
(Please give a few examples)
251
Appendix L
Mark the placement of any H, rising, or falling tones on all forms.
“Stem” below refers to all parts of the verb to the right of any object prex. E.g., The
innitive has a two-syllable stem, while has a one-syllable stem, and
has a three-syllable stem. If there are two (or more) tone classes of verbs in your
language, include one of each. If a verb with a long vowel exhibits a slightly dierent
surface tonal pattern, include such a verb in the table below, along with the negative form
of one verb of each syllable count.
Notes:
252
Bantu Orthography Manual
(Please give a few examples)
253
Appendix L
Mark the placement of any H, rising, or falling tones on all forms.
“Stem” below refers to all parts of the verb to the right of any object prex. E.g., The
innitive has a two-syllable stem, while has a one-syllable stem, and
has a three-syllable stem. If there are two (or more) tone classes of verbs in your
language, include one of each. If a verb with a long vowel exhibits a slightly dierent
surface tonal pattern, include such a verb in the table below, along with the negative form
of one verb of each syllable count.
Notes:
254
Bantu Orthography Manual
(Please give a few examples)
255
Appendix L
Mark the placement of any H, rising, or falling tones on all forms.
“Stem” below refers to all parts of the verb to the right of any object prex. E.g., The
innitive has a two-syllable stem, while has a one-syllable stem, and
has a three-syllable stem. If there are two (or more) tone classes of verbs in your
language, include one of each. If a verb with a long vowel exhibits a slightly dierent
surface tonal pattern, include such a verb in the table below, along with the negative form
of one verb of each syllable count.
Notes:
256
Bantu Orthography Manual
(Please give a few examples)
257
Appendix L
Mark the placement of any H, rising, or falling tones on all forms.
“Stem” below refers to all parts of the verb to the right of any object prex. E.g., The
innitive has a two-syllable stem, while has a one-syllable stem, and
has a three-syllable stem. If there are two (or more) tone classes of verbs in your
language, include one of each. If a verb with a long vowel exhibits a slightly dierent
surface tonal pattern, include such a verb in the table below, along with the negative form
of one verb of each syllable count.
Notes:
258
Bantu Orthography Manual
(Please give a few examples)
259
Appendix L
Mark the placement of any H, rising, or falling tones on all forms.
“Stem” below refers to all parts of the verb to the right of any object prex. E.g., The
innitive has a two-syllable stem, while has a one-syllable stem, and
has a three-syllable stem. If there are two (or more) tone classes of verbs in your
language, include one of each. If a verb with a long vowel exhibits a slightly dierent
surface tonal pattern, include such a verb in the table below, along with the negative form
of one verb of each syllable count.
Notes:
260
Bantu Orthography Manual
(Please give a few examples)
261
Appendix L
(Describe here)
Mark the placement of any H, rising, or falling tones on all forms.
“Stem” below refers to all parts of the verb to the right of any object prex. E.g., The
innitive has a two-syllable stem, while has a one-syllable stem, and
has a three-syllable stem. If there are two (or more) tone classes of verbs in your
language, include one of each. If a verb with a long vowel exhibits a slightly dierent
surface tonal pattern, include such a verb in the table below.
Notes:
262
Bantu Orthography Manual
(Please give a few examples)
263
Appendix L
Mark the placement of any H, rising, or falling tones on all forms.
“Stem” below refers to all parts of the verb to the right of any object prex. E.g., The
innitive has a two-syllable stem, while has a one-syllable stem, and
has a three-syllable stem. If there are two (or more) tone classes of verbs in your
language, include one of each. If a verb with a long vowel exhibits a slightly dierent
surface tonal pattern, include such a verb in the table below, along with the negative form
of one verb of each syllable count.
Notes:
264
Bantu Orthography Manual
(Please give a few examples)
265
Appendix L
Mark the placement of any H, rising, or falling tones on all forms.
“Stem” below refers to all parts of the verb to the right of any object prex. E.g., The
innitive has a two-syllable stem, while has a one-syllable stem, and
has a three-syllable stem. If there are two (or more) tone classes of verbs in your
language, include one of each. If a verb with a long vowel exhibits a slightly dierent
surface tonal pattern, include such a verb in the table below, along with the negative form
of one verb of each syllable count.
Notes:
266
Bantu Orthography Manual
(Please give a few examples)
267
Appendix L
Mark the placement of any H, rising, or falling tones on all forms.
“Stem” below refers to all parts of the verb to the right of any object prex. E.g., The
innitive has a two-syllable stem, while has a one-syllable stem, and
has a three-syllable stem. If there are two (or more) tone classes of verbs in your
language, include one of each. If a verb with a long vowel exhibits a slightly dierent
surface tonal pattern, include such a verb in the table below, along with the negative form
of one verb of each syllable count.
Notes:
268
Bantu Orthography Manual
(Please give a few examples)
269
Appendix L
Mark the placement of any H, rising, or falling tones on all forms.
“Stem” refers to all parts of the verb to the right of the innitive prex. E.g., The innitive
has a two-syllable stem, while has a one-syllable stem, and has
a three-syllable stem. If there are two (or more) tone classes of verbs in your language,
include one of each. If a verb with a long vowel exhibits a slightly dierent surface tonal
pattern, include such a verb in the table below, along with the negative form of one verb
of each syllable count.
Notes:
270
Bantu Orthography Manual
(Please give a few examples)
271
Appendix L
Mark the placement of any H, rising, or falling tones on all forms.
“Stem” below refers to all parts of the verb to the right of any object prex. E.g., The
innitive has a two-syllable stem, while has a one-syllable stem, and
has a three-syllable stem. If there are two (or more) tone classes of verbs in your
language, include one of each. If a verb with a long vowel exhibits a slightly dierent
surface tonal pattern, include such a verb in the table below, along with the negative form
of one verb of each syllable count.
Notes:
272
Bantu Orthography Manual
(Please give a few examples)
273
Appendix L
Choose simple -CVC- verbs for the chart below. The V in each verb should represent each
dierent phonemic vowel in your language. Choose a verb of each vowel to represent each
verbal extension (derivational sux) listed below. If possible, use a single verb for each
column. However, various verbs may need to be used in a single column in order to nd a
root that is semantically compatible with the sux.
Watch for any vowel harmony things that may be happening. The extension may
harmonize with the stem, or sometimes in the causative, the stem harmonizes with the
extension.
The examples below are not all naturally occurring forms in Swahili, but are intended only
to model the sux.
274
Bantu Orthography Manual
u gloss gloss
List appropriate vowels in the following ll-in-the-blank statements. The rst blank in each
statement may have two or more vowels listed, but only one vowel should be listed in the
nal blank of each statement. E.g., the rst statement may look like this:
In a verb with a stem vowel of ___i, u, or a__________ the vowel of a front vowel sux is
realized as______i________.
(A language with ve vowels will need to ll only the rst two in each group. Languages
with seven vowels will probably have three dierent forms of each type of sux.)
In a verb with a stem vowel of _________________ the vowel of a front vowel sux is realized
as______________.
In a verb with a stem vowel of _________________ the vowel of a front vowel sux is realized
as______________.
In a verb with a stem vowel of _________________ the vowel of a front vowel sux is realized
as______________.
In a verb with a stem vowel of _________________ the vowel of a back vowel sux is realized
as______________.
In a verb with a stem vowel of _________________ the vowel of a back vowel sux is realized
as______________.
In a verb with a stem vowel of _________________ the vowel of a back vowel sux is realized
as______________.
275
Appendix L
5 Table 1: Vowel Inventory
5 Table 2: V1 - V2 combinations chart
6 Table 3: Long Vowel occurrence in disyllabic
word roots
7 Table 4: Vowel Length
8 Table 5: Vowel Harmony
9 Consonants: Root-initial
11 Consonants: Root-medial
12 List of examples for each consonant phoneme
14 Double articulated stops
14 Labialized Consonants
15 Geminate consonants
15 Palatalized Consonants
16 Non-Word Final Syllables
17 Word nal syllables
18 Tone: Verbs
19 Tone: V-initial verbs
19 Tone: Nouns
20 Noun class prexes with Consonant Initial
Roots
24 Non-standard Noun Class Pairings
25 Noun Class Prexes with Vowel Initial Roots
39 Vowel Coalescence (word-internally) at
morpheme boundaries
39 Vowel Elision
276
Bantu Orthography Manual
40 Example phrases for this/these (huyu/hawa)
and that/those (near) (BASIC)
42 Example phrases for that/those (far) and that/
those (near addressee)?
45 Demonstrative forms: other
47 Interrogative Forms
49 Pronominal Forms 1
277
Appendix L
Complete Consultant Checked
Interrogative
Pronominal forms -1
Pronominal forms -2
Pronominal forms -3
Adjective
Associative Constructions
Diminutives and Augmentatives
Enumeratives
Locatives
Possessive pronouns
Verbs:
Copular forms
Copular forms part 2
Morphophonological processes…Perfective
Verbal Mood:Imperative plural
Verbal Mood:Imperative singular
Verbal Mood:Subjunctive
Verbal Mood:Subjunctive 2
Verbal Prexes: Subject.
Verbal Prexes: Subject 2
Verbal Prexes: Object
Verbal Prexes: Object Relat.
Verbal Prexes: Subject Relat.
Verbal Tense: Future 1
Verbal Tense: Future 2
Verbal Tense: Future 3
Verbal Tense: Future 4
Verbal Tense: Past 1
Verbal Tense: Past 2
Verbal Tense: Past 3
Verbal Tense: Present or unmarked
Verbal Tense: Present 2
Verbal Tense-Aspect/other
Verbal Tense-Aspect Anterior
Verbal Tense-Aspect Past Anterior
Verbal Tense-Aspect Progressive
Verbal Tense-Aspect Present Persistive
Verbal Tense-Consecutive
Vowel Harmony in Extensions
Conditional (if it exists)
Regret?
278
Anderson, Stephen C. and Christian Alomofor. 2005. Awing Orthography Guide.
Cameroon: SIL.
Bastin, Yvonne. 2003. The interlacustrine zone. In Derek Nurse and Gérard Philippson
(eds.), 501–527.
Bird, Steven. 1998. Strategies for representing tone in African writing systems: A critical review.
Edinburgh: University of Edinburgh.
Bird, Steven. 2000. Orthography and identity in Cameroon. Notes on Literacy 6(1–2):3–34.
Botne, Robert. 2003. Lega. In Derek Nurse and Gérard Philippson (eds.), 422–445.
Bresnan, Joan and Samuel A. Mchombo. 1995. The lexical integrity principle: Evidence
from Bantu. Natural Language and Linguistic Theory 13:181–254.
Burquest, Donald A. 1993. Phonological analysis: A functional approach. Dallas: SIL.
Edmondson, Jerold A. and Donald Burquest. 1998. A survey of linguistic theories. Dallas: SIL.
Eira, Christina. 1998. Authority and discourse: Towards a model for orthography selection.
Written Language and Literacy 1(2):171–224.
Gardner, William L. 2001. Consonant mutation in Shona languages. In Thomas Crowell
and Janet Harthan (eds.), Working papers 2, 51–59. Nampula, Mozambique:
Sociedade Internacional de Linguística.
Gardner, William L. 2005. Overview of Bantu languages. Tutorial paper, Fuller Theological
Seminary.
Gowlett, Derek. Zone S. In Derek Nurse and Gérard Philippson (eds.), 609–638.
Gudschinsky, Sarah. 1973. A manual of literacy for preliterate peoples. Dallas: SIL.
Hyman, Larry M. 2003. Segmental phonology. In Derek Nurse and Gérard Philippson
(eds.), 42–58.
Katamba, Francis. 2003. Bantu nominal morphology. In Derek Nurse and Gérard
Philippson (eds.), 103–120. .
Kithinji, Stephen, Nick Lunn, and Leila Schroeder. 1999. Tharaka orthography. Manuscript.
Kroger, Paul. 1990. Lexical phonology and the rebirth of the phoneme. Notes on Linguistics
50:11–24.
Kutsch-Lojenga, Constance. 1993. The writing and reading of tone in Bantu languages.
Notes on Literacy 19(1):1–19.
279
References
Kutsch-Lojenga, Constance. 2003. Bila. In Derek Nurse and Gérard Philippson (eds.), 450–
473.
Leitch, Myles. 2000. SHOEBOX Babole dictionary: Computerized Babole-French research
dictionary. Libreville: SIL.
Leitch, Myles. 2003. Le systeme babole. In F. Idiata (ed.), Les classes nominales et leur
semantisme dan les langues bantoues du nord-ouest, Munich: Lincom Europa.
Maddieson, Ian. 2003. The sounds of the Bantu languages. In Derek Nurse and Gérard
Philippson (eds.), 15–41.
Mattingly, I. G. 1992. Linguistic awareness and orthographic form. In R. Frost and L. Katz
(eds.), Orthography, phonology, morphology, and meaning, 11–26. Amsterdam: North-
Holland.
Moe, Ronald. 2003. Developing an orthography for a Bantu language. Manuscript.
Moe, Ronald and Samuel Mubbala. 2006. Lugwere writers’ guide. Entebbe: SIL.
Nurse, Derek and Gérard Philippson. 2003. The Bantu languages. London: Routledge.
Nzogi, Richard. 2005. Lugwere orthography statement. Manuscript.
Odden, David. 2003. Ruji-Ruvuma (N10, P10-20). In Derek Nurse and Gérard Philippson
(eds.), 143–163.
Reeder, JeDene. 1998. Vowel harmony in Pagibete verbs. Manuscript.
Roberts, David. 2008. Thirty Years of Tone Orthography Testing, Journal of West African
Languages 200 XXXV.1-2.
Schadeberg, T. C. 1999. Katupha’s law in Makhuwa. In J. M. Hombert and L. M. Hyman
(eds.), Bantu historical linguistics: theoretical and empirical perspectives, 379–394.
Stanford: CSLI Publications.CSLL
Schadeberg, T. C. 2003. Historical linguistics. In Derek Nurse and Gérard Philippson (eds.),
143–163.
SIL International. 2002. Developing an orthography. In LinguaLinks version 5. CD-ROM.
Dallas: SIL.
SIL International. 2002. How to make spelling choices. In LinguaLinks version 5. CD-ROM.
Dallas: SIL.
Smalley, William A. (ed.) 1964. Orthography studies. London: UBS.
Snider, Keith L. 1992. Grammatical tone and orthography. Notes on Literacy 18(4):25–30.
Snider, Keith L. 1998. Orthography design and development. Paper presented at the SIL
280
Bantu Orthography Manual
Literacy Consulting Skills Seminar, Yaounde, Cameroun. August 29–29, 1998.
Snider, Keith L. 2001. Linguistic factors in orthography design. In Ngessimo M. Mutaka
and Sammy B. Chumbow (eds.), Research material in African linguistics: Focus on
Cameroon, 323–332. Koln: Rudiger Köppe Verlag.
Snider, Keith L. 2005. You can have your cake and eat it, too: An orthography that meets
the needs of both mature and immature readers. Lecture notes for an orthography
class at the University of North Dakota, June 29, 2005.
Ssemakula, Mukasa E. 2005. A Primer on Speaking and Writing Luganda, http://www.
buganda.com/Luganda.htm#elem) (accessed April 27, 2005).
Stegen, Oliver. 2005. Vowel length in eastern Bantu orthographies. Paper presented at the
SIL Bantu orthography meeting Dallas, November 7–12, 2005.
Van Dyken, Julia R. and Constance Kutsch-Lojenga. 1993. Word boundaries: Key factors in
orthography development. In Rhonda Hartell (ed.), Alphabets of Africa, 3–20. Dakar:
UNESCO.
Van Otterloo, Karen and Roger Van Otterloo. To appear. Determining word boundaries.
Dallas: SIL.
