Writing Yorùbá: Ajami, Crowther, and the Orthography Committees

Yorùbá has been written in two scripts. The older is Ajami, Yorùbá written in modified Arabic characters by Muslim scholars, documented from the late nineteenth century 1. The one now standard is Roman, developed by Church Missionary Society missionaries and Yorùbá Christians from the 1840s, formalised at a Lagos conference in 1875, and revised by government committees in the 1960s and 1970s 23. The current system uses subdots to distinguish letters and accents to mark tone, and both are structurally necessary rather than ornamental.

Understanding this history matters because the writing system was designed by particular people for particular purposes, and some of those purposes were evangelical. The script is not neutral, and knowing who made which decision lets a reader see where the language's own logic ends and a translator's theology begins.

Before Roman script: Ajami

Ajami is the general term for African languages written in modified Arabic script, a practice found across Muslim West Africa for Hausa, Fulfulde, Wolof, Kanuri, Nupe, and Yorùbá among others 1. It followed the spread of Islam and of Arabic literacy, and in the Nigerian region it expanded considerably after the founding of the Sokoto Caliphate in the early nineteenth century 1.

For Yorùbá specifically, the surviving documentary record is later than the Roman-script record might lead one to expect. The oldest documented Yorùbá Ajami material dates to the late nineteenth century, with early examples attributed to scholars such as Badamasi b. Musa Agbaji around 1891 1. Ilọrin, with its Islamic scholarly establishment, is a principal centre, and manuscript collections from the Ilọrin Emirate have been catalogued and published 1.

Two points need care here. First, that Arabic-script literacy existed among Yorùbá Muslims well before 1891 is not in doubt, but the surviving Yorùbá-language Ajami texts are later than that literacy, so claims about a much older Yorùbá Ajami literary tradition run ahead of the documented evidence. Second, Ajami is not a historical dead end: recent decades have seen new Yorùbá Ajami texts in religious writing, short stories, advertisements, and political pamphlets, and the tradition is better described as continuing than as extinct 1.

Ajami's technical difficulty is that Arabic script marks consonants and long vowels well and short vowels only through optional diacritics, while Yorùbá needs a seven-vowel distinction plus three tones. Yorùbá Ajami writers extended the script to cope, but no single standardised Yorùbá Ajami orthography was ever agreed, which is one reason it did not become the general written medium 1.

The missionary Roman orthography

The Roman orthography emerged from the Church Missionary Society's work in the 1840s and 1850s. The central figure is Samuel Àjàyí Crowther (c. 1806-1891), a Yorùbá man captured in the slave trade as a boy, freed by the British naval squadron, landed at Freetown, educated in Sierra Leone, ordained, and eventually consecrated as the first African Anglican bishop 4.

Crowther published a Yorùbá vocabulary and grammatical account in 1843, followed by A Grammar of the Yoruba Language in 1852 and an expanded A Vocabulary of the Yoruba Language the same year 45. The 1843 work is generally described as the first grammar and vocabulary of an African language produced by a native speaker of that language, which is a substantive achievement and should be recognised as such alongside the criticisms that follow in translation-and-bias.

Crowther did not work alone or from nothing. C. A. Gollmer proposed an orthography in 1847, which fed into the CMS's 1848 Rules for Reducing Unwritten Languages to Alphabetic Writing in Roman Characters, the general missionary template applied across many languages 3. The early orthography was inconsistent, and Crowther's own 1852 grammar did not mark tone systematically 4.

The 1875 conference

The Church Missionary Society convened a conference on Yorùbá orthography on 28 and 29 January 1875 at the CMS Faji Mission House, Lagos, with Crowther among the prominent African participants 3. The conference resolved a range of inconsistencies in existing spelling practice and settled the conventions that would stand for most of the next century, including the subdotted letters ẹ, ọ, ṣ and the use of acute and grave accents for High and Low tone with Mid left unmarked 3.

The 1875 system remained in use for nearly a hundred years 3. It is the reason a Yorùbá Bible, a Yorùbá hymnal, and a Yorùbá newspaper from the colonial period are all legible to a reader today with only modest adjustment.

The reform committees

Pressure for revision built in the mid twentieth century. Àjàyí (1960) called for a new conference to reform the orthography for school use 3. Ayọ̀ Bamgbóṣé, later Nigeria's first professor of linguistics, published Yoruba Orthography through Ibadan University Press in 1965, a linguistic appraisal of the existing system with concrete reform proposals, arguing from the difficulties native speakers were experiencing in using it 36.

Bamgbóṣé then chaired the Western Nigeria Ministry of Education's Committee on Yoruba Orthography from 1966, and edited its report, which confirmed the underdots on the open vowels ẹ and ọ and on the fricative ṣ, and settled the marking of High, Low, and Mid tone 6. The current standards were established following the 1974 Joint Consultative Committee on Education report 37.

Contention did not end there. Disagreement continues over spelling conventions, word division, and above all over how consistently tone marks should be required in ordinary printed Yorùbá 3. This is a live language-planning debate, not a settled question.

Why the diacritics are not optional

There are two distinct kinds of mark, and conflating them causes confusion.

Subdots are part of the letter. The dot under ẹ, ọ, ṣ distinguishes separate phonemes: ẹ [ɛ] from e [e], ọ [ɔ] from o [o], ṣ [ʃ] from s [s]. Writing ese for ẹsẹ̀ is like writing bit for beat, a different word, not a casual spelling of the same one.

Tone marks carry lexical and grammatical meaning. As set out in yoruba-phonology-and-tone, the acute marks High, the grave marks Low, and Mid is unmarked. Because Yorùbá tone distinguishes words, removing tone marks removes information the reader cannot always recover.

The scale of the resulting ambiguity has been measured. Working with a one-million-word Yorùbá corpus, Iroro Orife found that 85 percent of all words carry diacritics, that 32 percent of unique undiacritized word types correspond to two or more distinct diacritized forms, and that 64 percent of unique undiacritized monosyllabic words have multiple diacritized forms 8. The average number of alternatives per undiacritized word, a measure called lexical diffusion, was 1.47 8. Because Yorùbá verbs are predominantly monosyllabic, the ambiguity falls heaviest on exactly the words that carry the action of a sentence 8.

Concrete examples from that work show the problem 8:

• gba stands for gbà (spread), gba (accept), and gbá (hit)
• ilu stands for ìlú (town), ilu (opener), and ìlù (drum)
• oko stands for ọkọ́ (hoe), ọ̀kọ̀ (spear), and ọkọ̀ (vehicle)
• sa stands for sá (run), ṣá (fade), and ṣà (choose)
• n stands for n (I) and ń (the continuous aspect marker)

The common defence of untoned writing, that fluent readers disambiguate from context, is true but limited. It works for familiar running prose and fails for names, unfamiliar words, poetry, liturgical text, dictionary entries, and anything a machine must process. It also fails for exactly the reader this corpus addresses, someone rebuilding literacy in a language they speak but were never taught to read.

Yorùbá in Unicode and on devices

Every character Yorùbá needs exists in Unicode, but the encoding has a structural awkwardness worth understanding, because it explains why Yorùbá text online is so often broken.

Unicode provides precomposed characters for ẹ (U+1EB9) and ọ (U+1ECD) and ṣ (U+1E63). What it does not provide is a single precomposed character for a subdotted vowel that also carries a tone mark, such as ẹ́ or ọ̀ 9. Proposals for such precomposed characters were declined, on the general Unicode principle of not adding new precomposed forms that can be built by combination 9. So ẹ́ must be encoded as a base character plus at least one combining mark.

Three practical consequences follow.

First, dead-key input is awkward. A conventional keyboard dead key maps to a single code point, and since ẹ́ has none, standard dead-key schemes cannot produce it directly 9. Dedicated Yorùbá keyboard layouts solve this, but the user has to install one.

Second, the same visible word can be several different byte sequences. The underdot may be encoded as combining dot below (U+0323) or, in text produced by older or non-standard tools, as one of several visually similar combining marks, and combining marks may appear in non-canonical order 9. Two words that look identical on screen then compare as unequal in software, breaking search, sorting, and deduplication. The remedy is Unicode normalisation, applying NFC or NFD consistently across a corpus, which is standard practice in Yorùbá language technology work.

Third, rendering depends on the font. Fonts that do not properly position a combining acute over an already-subdotted letter will produce collisions or misplaced marks. Fonts with good combining-mark support handle it correctly.

The result is that most Yorùbá text on the internet is plain ASCII with no diacritics at all, not because writers reject the orthography but because devices and applications made it difficult 8. This is a technology-adoption problem rather than a linguistic one, and it is being worked on: automatic diacritic restoration, treating undiacritized text as a source language and diacritized text as a target, now achieves error rates below 5 percent using neural sequence-to-sequence models, with pre-trained models and datasets released openly 8.

What to take from this

The Yorùbá writing system is about 180 years old in its Roman form, was shaped substantially by Yorùbá people including Crowther, was formalised at a mission conference in 1875, and was revised by Nigerian linguists working through government committees in 1966 and 1974. It is well designed for the language: it marks the seven-vowel contrast and the three-tone contrast with economy.

Its main practical failure is that the marks are widely omitted, which produces a generation of Yorùbá speakers who cannot comfortably read their own language and assume the fault is theirs. It is not. The orthography is learnable in an afternoon, and learning-yoruba-resources sets out how.