Orthography development is an important goal in language documentation (Seifart 2006): it is often the most tangible outcome of a language documentation project for speakers, and it facilitates the creation of other outputs, such as transcribed documentary corpora, teaching materials, dictionaries, grammars, and text collections. Nonetheless, the challenges faced in orthography development are not always explicitly discussed in the literature on language documentation (though see Cahill 2018; Jones & Mooney 2017; Lüpke 2011; Seifart 2006). This paper explores three challenges faced in developing an orthography for the Enggano language (IPA /eŋgano/, ISO 639-3 Code eno) an endangered language of Indonesia spoken by approximately 1,500 speakers on Enggano Island, Sumatra. The primary aim of the paper is to reflect on the ongoing process of orthography development, providing a model for how to navigate such challenges in future language documentation projects. A secondary aim is to present the Enggano orthography, providing some motivation for the decisions that were reached.

The paper is structured as follows. Section 2 provides some background on the Enggano language and the history of its documentation. Section 3 reviews the existing literature on orthography development, and discusses major factors known to affect orthographic choices. Section 4 presents three case studies of challenges in Enggano orthography development, specifically the challenges of representing central vowel phonemes, nasal vowel phonemes, and allophonic glides. Finally, section 5 concludes with a presentation of the symbols and conventions of the Enggano orthography developed during the current documentation project, and reflections on the lessons learned for future orthography development projects.

The Enggano language is spoken on Enggano Island, situated off the southwest coast of Sumatra, Indonesia. Enggano Island is the southernmost of the Barrier Islands chain, which also includes Simeuluë, Nias, and the Mentawai Islands (Figure 1).

Most scholars now agree that Enggano is an Austronesian language (Arka et al. 2022; Dyen 1965; Edwards 2015; Nothofer 1986; Smith 2017, 2020).1 However, there is ongoing debate as to whether Enggano forms a subgroup with the Barrier Island languages and Batak languages of Sumatra, as claimed in Nothofer (1986), Smith (2017), and Billings & McDonnell (2022), or whether it is a primary branch of Malayo-Polynesian, as claimed in Edwards (2015). There are six main villages on Enggano, all located on the north coast of the island, as shown in Figure 2.

These include Enggano-speaking populations as well as non-Enggano speaker populations. As a whole, the language can be considered endangered, but the degree of endangerment varies (Arka et al. 2022). The language is most vital in the central villages: Meok, Apoho, and Malakoni. In northern and southern villages, non-Enggano populations are higher, which has resulted in increased influence of Indonesian, the national language of Indonesia, more code-mixing, and higher rates of language shift towards Indonesian in those villages.

Enggano has a relatively long history of documentation. The earliest records were collected during the Dutch colonial era in the mid-nineteenth century, and consist of word lists (Boewang 1854; Helfrich 1893, 1916; Helfrich & Pieters 1891; Oudemans 1879, 1889; van der Straaten & Severijn 1855; von Rosenberg 1855; Walland 1864; as well as the Holle list, collected in 1895 and published in van der Noord 1987) and some texts collected by Helfrich (1916). Based on a seven-month stay between 1937 and 1938, Hans Kähler produced the most comprehensive documentation to that point, consisting of a sketch grammar (Kähler 1940), a text collection (Kähler 1955, 1957, 1958, 1960a, 1960b, 1961, 1962, 1964, 1975), and a dictionary published posthumously as Kähler (1987). More recent linguistic work includes an unpublished word list collected by Bernd Nothofer in 1986 which formed the basis for historical work (Nothofer 1986); materials produced by government agencies in Indonesia, including the word list and grammar sketch produced by Pusat Pembinaan dan Pengembangan Bahasa [Center for Language Development and Cultivation] (Kasim et al. 1987; Nikelas, Rasyad & Semi 1994); the dictionary produced by Kantor Bahasa Provinsi Bengkulu [Regional Office for Language in Bengkulu Province] (Riswari et al. 2021); and word lists/example sentences in several linguistic theses (Butters 2021; Wijaya 2018; Yoder 2011).

In order to understand the structure of contemporary Enggano and address the debates surrounding its historical position in the Austronesian family, the authors of this paper are involved in an ongoing research project, “Enggano in the Austronesian family: Historical and typological perspectives”.2 The project has two major strands, a documentation strand and a historical linguistics strand. The aim of the historical linguistics subproject is to assess the position of Enggano within the Austronesian family, while the documentation subproject’s goal is to produce a grammar of contemporary Enggano, an annotated documentary corpus, and teaching materials to aid in the process of language revitalization. In order to achieve these aims, it is necessary to have a standard orthography for writing the language.

Unfortunately, the previous literature does not provide a standard means of writing Enggano. Early records did not use a common orthography, as illustrated in Table 1.3 Moreover, there is little information on whether the orthography in these early word lists reflects a phonological analysis of the language or simply the compiler’s perception of words, which may have been influenced by their native language phonologies. Thus, it is unclear whether these are alternative spellings of the same form, or reflective of dialect differences and hence different pronunciations.

Kähler’s work is based on a phonological analysis of the language as set out in the grammar (Kähler 1940: 83), the text collection (Kähler 1975ː III–IV), and the dictionary (Kähler 1987: iii–iv). However, as discussed by Schmidt in the foreword to Kähler (1987) and summarised in Edwards (2015: 60), the orthography is not consistent across the different materials. This is illustrated in Table 2.

The materials differ not only in the symbols they use—e.g. ⟨u⟩, ⟨ə̄⟩, and ⟨ə⟩ for the central vowel /ɨ/—but also in the conventions adopted, e.g., whether to represent nasal vowels, as in the grammar and dictionary, or not, as in the 1975 texts.4

More recent works have tended to use either Indonesian orthography, which is practical but (as we will see) not necessarily suited to Enggano phonology, or the IPA transcription system, which is consistent and allows users to accurately represent the sounds of words, but is not particularly practical as it includes symbols that are not familiar to speakers, such as ⟨ɨ⟩ and ⟨ʔ⟩. Consequently, there is no orthography in the existing literature that accurately represents the sounds of Enggano in a way that is simple and easy to use.

Within the community, there is no standardized way of writing Enggano that is universally accepted and adopted. Instead, there are multiple competing orthographies in use. There is an ongoing project to translate the Bible into Enggano, and those closely involved in this project have suggested symbols such as ⟨ė⟩ to represent the schwa vowel /ə/. Other speakers who are literate in Indonesian tend to adopt Indonesian conventions when writing Enggano, or combine these with regional practices, such as the use of ⟨eu⟩ for the central vowel, which is a digraph used in Sundanese (Wibawa et al. 2018) and Acehnese (Durie 1984).5 Adopting the conventions of the national language, Indonesian, is certainly the simplest option, and simplicity is desirable in an orthography (Seifart 2006). However, Enggano phonology differs from Indonesian phonology, and hence Indonesian orthography is not perfectly suited to the structure of Enggano, as we discuss in Section 4. Thus, the community does not have a single standard orthography that could be adopted in the documentation project either.

Consequently, as in many documentation projects (see Lüpke 2011), an important aim of the project is to develop a standard orthography for Enggano. Following Seifart (2006), we want the orthography not only to reflect the structure of Enggano, but also to be widely accepted and adopted—both in the community and by other researchers working with contemporary Enggano data. This will help to avoid the potentially divisive effects of multiple competing orthographies (cf. Lüpke 2011: 318). In order to do this, we need to take both linguistic and extra-linguistic factors into account and ensure that important stakeholders in the Enggano community are involved in the orthography development process. As an additional challenge, our work also took place during the COVID-19 pandemic, which imposed its own constraints on the process.

Coulmas (2003: 35) defines an orthography as a set of graphic symbols (“graphemes”), the rules for how they are used, and the sounds they represent. Ideally, the orthography should not only reflect the structure of the language but also be practical and easy to use. Unfortunately, these desiderata are often in conflict, and consequently the orthography developer must balance the advantages and disadvantages of the different options. Following Seifart (2006: 287), decisions in two major areas must be made when developing an orthography in the context of minority/endangered languages: orthographic depth and adopting existing conventions.

Orthographic depth is the extent to which an orthography reflects pronunciation (Katz & Frost 1992). A shallow or phonemic orthography has one letter per sound, or rather one grapheme per phoneme, since allophonic variation and suprasegmental features are often not represented. In the context of endangered languages, where there may be many semi-speakers or learners, Seifart (2006: 283) suggests that a phonemic orthography may be particularly practical, since it clearly distinguishes important sound contrasts. This allows learners to understand how to write a word based on how it sounds and how to pronounce a word based on how it is written. A deep or morphemic orthography, however, preserves the graphic identity of words so that the same morpheme is written with the same letters even if it is pronounced differently in different contexts, as is the case for English reduce vs. reduction (Seifart 2006: 278). A “deeper” orthography can be easier for fluent speakers to read, and it is particularly useful in contexts where there is significant dialect variation and where languages undergo significant morphophonological alternations.

The second important decision is whether or not to adopt the existing conventions of a dominant language, which can make the orthography easier for speakers to learn. Endangered languages almost invariably exist in multilingual environments, where speakers are literate in the orthography of a dominant language (Lüpke 2011). Hence, adopting existing conventions may increase uptake and usability. Following this principle, we adopted an alphabetic system, as is used in Indonesian, rather than a syllabary or logography (see e.g. Coulmas 2003 on different types of writing systems). However, the existing conventions of a dominant language may not fit the endangered language nor serve the emblematic function of expressing a distinct socio-cultural identity (see Cahill 2018; Sebba 2007). In the next section, we explore three challenges that arose in our attempt to develop a phonemic orthography for Enggano, while also adopting Indonesian conventions wherever possible. We show that adopting Indonesian orthography wholesale is disadvantageous in several respects, and we describe how we augmented and modified standard Indonesian orthography to make it suitable for Enggano. In doing so, we discuss both how to represent contrasts and how to decide which contrasts to represent (cf. Seifart 2006 on underrepresentation), weighing up the advantage of phonological accuracy against simplicity and learnability for users. Ultimately, decisions were guided by input from the Enggano community.

The documentation project began in 2018, and several field trips were conducted between 2018 and 2020, following standard methods for language documentation and description (see e.g. Austin 2016; Himmelmann 1998, 2006; Woodbury 2003, 2011). During the fieldtrips, co-authors I Wayan Arka and Mary Dalrymple established relationships with community elders in Enggano. They also built relationships with colleagues at Kantor Bahasa Provinsi Bengkulu [Regional Office for Language in Bengkulu Province], including co-author Dendi Wijaya, who were working on their own project to produce materials in the Enggano language.6 They provided training for native-speaker consultants, including co-author Engga Zakaria and primary-school teacher Fitri Yunita, to join the documentation project team and play a central role in data collection and analysis.7 During this time, a sizeable corpus was collected comprising word lists, elicitation, stimuli-based materials, and naturalistic texts. Many of the recordings were initially transcribed and translated into Indonesian by Engga Zakaria. Subsequently, the materials were translated from Indonesian into English by colleagues at Udayana University, Indonesia. In 2020, at the start of the COVID-19 pandemic, international travel ceased and all fieldwork moved online. Co-author Charlotte Hemmings joined the documentation project and began to work with the rest of the co-authors (henceforth the project team) to analyze recordings during regular meetings over Zoom.

During the online meetings, the project team worked with the corpus to better understand the structure of Enggano. Specifically, we listened back to recordings and asked Engga Zakaria to repeat slowly in order to clarify the sounds of words and their internal structure. In particular, we looked for minimal pairs in order to establish a list of contemporary Enggano phonemes, and we identified related forms in our word list in order to understand the morphophonological processes in Enggano. Building on Yoder’s (2011) analysis and the findings of these regular meetings, we established the phonemes in Tables 3 and 4 for contemporary Enggano.8

The voiceless consonants /p/, /k/ and /ʔ/ are represented using the symbols ⟨p⟩, ⟨k⟩ and ⟨’⟩. Although Standard Indonesian orthography uses the letter ⟨k⟩ for both [k] and [ʔ], we represent this phonemic distinction orthographically, since it has a relatively high functional load: e.g., it allows users to distinguish between common words like /məʔ/ the relativizer (cf. Indonesian yang) and /mək/ ‘many’, and morphemes like the 1SG possessive suffix /-ʔ/ and the 1PL.INCL possessive suffix /-k/ (see section 4.1 for further discussion on the importance of representing phonemic contrasts). The phonemic status of [t] and [r] are not entirely clear. Kähler (1940) treats both as allophones of /d/. However, [t] and [r] both occur frequently in our corpus, and there are some near minimal pairs suggesting they may have phonemic status in contemporary Enggano: e.g. [i top] ‘below’ vs. [dop] ‘land/earth’; [di] ‘tongue’ vs. [ari] 2pl, [kitahaʔ] ‘stab’ vs. [karahaʔ] ‘my body’. Consequently, we write these as ⟨t⟩, ⟨r⟩, and ⟨d⟩ respectively.9 The voiced stop /b/ is written as ⟨b⟩ and the nasals /m/ and /n/ are written as ⟨m⟩ and ⟨n⟩. Note that the voiced stops /b/ and /d/ are sometimes realised as [m] and [r] word-finally, which results in variant pronunciations, e.g. [jub] vs. [jum] ‘house’ and [hiɨd] vs. [hiɨr] ‘woman’. The variants with [m] and [r] are particularly associated with younger speakers.

The phoneme /h/ has several allophones: [ç] following the oral high front vowel /i/, [x] following non-front high oral vowels and [h] in all other contexts. In Kähler’s work, these variants are represented orthographically ([ç] and [x] as ⟨x⟩, and [h] as ⟨h⟩), but we treat this as allophonic and hence write all as ⟨h⟩. The phoneme /ç/ is variously realised as [s], [t], [ç] and [tʃ] depending on the speaker. It is not attested word-initially and normally occurs word-finally after /i/. Given that [ç] is also an allophone of /h/, Yoder (2011) treats this sound as underlyingly /h/. However, historically it comes from a different source and hence we represent it orthographically as ⟨c⟩. Note that /ç/ appears to have [t] as a variant realisation in non-final positions, e.g. yic [jiç] ‘word/sound’ → yitik [jitik] ‘our words/sounds’. However, it is not to be considered an allophone of /t/ in contemporary Enggano since /t/ can also occur word-finally following /i/, e.g. i dit [idit] ‘there’. Finally, the vowels have nasal allophones following nasal consonants (see section 4.2) and /i/ and /e/ have glide allophones in onset position (see section 4.3 and Yoder 2011: 22).

Once we had established the phonemic inventory, we were able to identify several orthographic issues that arose in the tension between adhering to the phonemic principle of one symbol per sound and following Indonesian orthographic conventions as closely as possible. We first discussed each of these issues within the project team and developed a list of options, weighing up the advantages and disadvantages. We then consulted with other researchers familiar with Enggano, including Bernd Nothofer, Erik Zobel, and Daniel Krausse, who are working within the historical linguistics strand of the Enggano project, and Yanti Riswara Idris of Kantor Bahasa Provinsi Bengkulu, who was co-ordinating the production of an Enggano-Indonesian dictionary (Riswari et al. 2021). Finally, Engga Zakaria and Dendi Wijaya (henceforth referred to as “the local team”), who were based in Bengkulu and able to travel to Enggano Island, took the options to the Enggano community and explained the motivations for our choices to important stakeholders. The local team sought feedback from community elders via focus group interviews, and from schoolchildren and learners of Enggano via sample tasks from the teaching materials. This allowed us to validate the choice of symbols and begin to socialize the proposed orthography among learners.

In conclusion, this paper has outlined three patterns of Enggano phonology that present challenges when trying to develop an orthography with reasonably shallow orthographic depth (following the principle that phonemic orthographies are most transparent for learners) while also adopting many of the existing conventions of Standard Indonesian (following the principle that staying close to an already familiar orthography is the simplest option and hence most likely to be readily adopted). In each case, it was shown that the symbols and conventions familiar to speakers from Indonesian were at odds with the desire for an orthography that accurately and consistently represented the morphophonological structure of Enggano. This necessitated a multi-stage approach to developing orthographic proposals that involved the following steps:

The outcome of this process is a working orthography that is now being implemented in the documentary corpus, grammar and teaching materials. This will hopefully make the project outputs maximally accessible to speakers of Enggano and aid in the process of language revitalization (see Cahill 2018). The current orthography adopts the symbols in Table 16. It also adopts the convention that predictable and regular phonological processes, such as nasal spreading and glide-insertion between vowels in a vowel sequence, are not represented for the sake of simplicity. However, where variant pronunciations are optional or reflect dialectal/sociolinguistic variation, the convention is to write the word as it was pronounced by the speaker.

The process of orthography development and corpus building is still ongoing, and we may encounter further issues as our project progresses. Once the orthography design and the spellings of particular words have been further tested and approved by community leaders, educators, and other important stakeholders, the next step will be to implement the orthography and promote literacy in the wider community. This may not be straightforward, given that there are already competing orthographies. However, to maximize acceptability among speakers, we will explain the linguistic motivations behind orthographic decisions. We will also highlight the potential of the orthography not only to fulfill practical, pedagogical functions in aiding learners to read and write, but also to serve as an identity-marking function that represents the distinctiveness of Enggano culture (see Cahill 2018). During the implementation, we will remain open to any changes that make the orthography more acceptable to the Enggano community, such as including ⟨eu⟩ as a variant grapheme for ⟨u̇⟩, and we will continue to view this as a collaborative and ongoing process. One factor that may increase the acceptability of the orthography will be ease of use with modern technology including smartphones. Our Keyman keyboard allows for the use of diacritics and combined diacritics using any font.16 As part of a future project, we hope to develop an app that will enable predictive text and auto-correction when typing on smart phones in order to make the orthography easy to use and promote uptake among the Enggano community.

The main lesson that we have learned from this project is how important it is to be open to continually developing and re-negotiating orthographic decisions, either when motivated by linguistic factors or when motivated by extra-linguistic concerns within the community. In the context of language documentation, it is to be expected that the linguist’s understanding of the structure of the language will change over the course of the project. Nonetheless, it is helpful to think about orthographic representation from the beginning of a project, since, as shown in relation to glides, community feedback on orthographic proposals can provide direct insight into language structure. In our project, we were forced by the COVID-19 pandemic to do much of the analysis and discussion online. However, this has turned out to have several advantages that could be borne in mind for future documentation and orthography development projects. Firstly, having regular meetings via Zoom allowed us to connect scholars and language activists in different countries and time zones who could bring their own expertise and experience to the challenge of orthography development, particularly in the initial stages of weighing up the advantages and disadvantages of the different options. Secondly, and more importantly, it meant that we had no choice but to invest time in capacity building and training for the local project team in Indonesia, who played an important role in collecting feedback from the community and facilitating the re-development of orthography proposals. This, we believe, led to a more collaborative approach, where important stakeholders were directly involved in the process and empowered to make informed orthographic choices and communicate these to the Enggano community at large (cf. Czaykowska-Higgins 2009: 24). Working collaboratively and collecting feedback as we go ensures greater community engagement. Ultimately, this is the only way to really achieve the goal of an orthography that is linguistically motivated but also practical, easy to use, and widely accepted.