In his groundbreaking article that defined the field of language documentation, Himmelmann (1998) includes archiving as one of the four key steps in creating what he calls “a language documentation” (p. 171), that is, a collection of transcribed and annotated audiovisual recordings with their accompanying metadata. Woodbury (2003) calls such a collection of primary data and the associated metadata a “corpus” and he includes archiving as one of six criteria that establish the overall quality of a documentary corpus.

Metadata are the supporting contextual, technical, and administrative documentation that help to explain the data, including any keys (e.g., codes, orthographies) needed to understand, analyze, and reuse them (KUNG et al., 2020). Table 1, adapted from Conathan (2011), illustrates the levels and types of metadata commonly requested by repositories. The categories in Table 1 do not come from a particular archive, nor are they reflective of the categories that all archives require. Rather, they constitute important contextual metadata that potential depositors should keep in mind when organizing materials for archiving. (For further reading on this topic, we recommend NATHAN and AUSTIN, 2004; JOHNSON, 2004; CONATHAN, 2011; and KUNG et al., 2020.)

Table 1. Table 1. Metadata categories relevant to language documentation(adaptada de CONATHAN, 2011).

Key reasons to archive language data are to ensure their longevity and accessibility. As we discuss below, digital repositories offer options for replicability and protection against the hazards of fire, flood, loss, mold, insects, etc. that threaten the conservation of physical materials, such as those seen in Figure 1.

Figure 1. Figure 1. Physical copies of language documentation materials: boxes of papers, stacks of audio cassettes, handwritten and photocopied notebooks (Pérez Báez: ailla:257492, DSC_0001; Pérez Báez: ailla:257543, DSC_0052; P. Epps. Photos are All Rights Reserved, used with permission.).

Himmelmann (2006) expounds on the importance of archiving the primary data that result from a research project in addition to publishing the analytical results of the project, that is, the Boasian trilogy of a grammar (sketch), dictionary, and set of texts (WOODBURY, 2003). When the primary data are archived, those data can be reused for additional language documentation work, as well as other types of linguistic analysis. Further, archiving the primary and secondary data allows for the analytical output to be verifiable and reproducible (HIMMELMANN, 2006; BEREZ-KROEKER et al., 2018). Ultimately, archiving primary and secondary data in a digital repository where they are publicly accessible facilitates data reuse, and provides a stable means for citing the data (for recent guidance, see CONZETT and DE SMEDT, 2022), in the form of a persistent identifier. Persistent identifiers such as Digital Object Identifiers (DOIs), Handles, and Uniform Resource Identifiers (URIs) allow creators of the data to receive proper attribution for their work (BEREZ-KROEKER et al., 2018).

Language documentation materials are collected with great effort, time, and (often) money; however, they can represent something much more profound to the speakers: their culture, identity, and self-determination (UNGA, 2007; CARROLL et al., 2020). Thus, it is extremely important to ensure that speakers have access to the documentation. Ideally, copies of the documentation are left in the community from the outset, but otherwise a copy should be repatriated or returned to the community (KUNG, 2020; VAPNARSKY, 2020; R. MILLER, 2021; see §4.3). Putting the data into a digital archive is a form of digital repatriation (KUNG, 2020), assuming communities have relatively unobstructed access to the data. While this is not the only way that data should be repatriated to the community of origin, it is one way to share these materials with the community while also preserving them for future generations of speakers. Nevertheless, it is still the case that many indigenous peoples and communities all over the world do not have (adequate) access to the internet. In these cases, a local, regional, or national archive, library, museum, school, governmental office, or some other location that is accessible to the speaker community should be identified, and a copy of the data should be deposited there (WILBUR, 2014) in addition to being deposited in a more broadly accessible digital repository, which will be better equipped to preserve the digital data for the long term (see below). Regardless of exactly where the data are archived, sharing and repatriating the primary and secondary data and resulting publications supports language maintenance, reclamation, and revitalization efforts by making materials available to speakers and their descendants. While we would hope that the data could be used for language education, maintenance, or revitalization efforts, it is impossible to foresee the actual or exact uses that the speakers and their descendants will make of the data (for recent examples see SPENCE, 2018; VAPNARSKY, 2020; and LUKANIEC, 2022). Nevertheless, the data contain the languages and cultural heritage of these speakers, and they deserve the right to decide how and when to use them (HOLTON et al., 2022).

Throughout this article, we use the term archive to refer to a repository of materials that are saved and preserved by an organization or institution so that they can be reused in the present and the future. An archive can hold analog materials, digital materials (both born digital and digitized) or both. A digital repository is an archive of digital materials or records, and a (digital) language archive is an archive that specializes in linguistic cultural heritage and language materials or data. In the field of language documentation, the most quoted definition of an archive is from Johnson (2004, p. 142): “An archive is a trusted repository created and maintained by an institution with a demonstrated commitment to permanence and the long-term preservation of archived resources”. Johnson’s definition does not mention the digital component, so we offer an updated definition from Trevor Owens, the Head of Digital Content Management at the US Library of Congress. This definition emphasizes the misconception that software designed for digital asset management is the same thing as a digital repository. From this point forward, we use the terms archive and repository interchangeably. Owens (2018, p. 4) writes: “A repository is not a piece of software… A repository is the sum of financial resources, hardware, staff time, and ongoing implementation of policies and planning to ensure long-term access to content. Any software system you use […] to preserve and provide access to digital content is by necessity temporary […] it likely will not last forever […] Institutions make preservation possible” (emphasis in the original). There are some recurring themes across these two definitions: institution, commitment, ongoing, long-term, and preservation. We will touch on each of these themes in this section, but the take-away here is that you cannot simply build a digital archive and forget about it. An archive is an ongoing commitment on the part of the institution or organization that decides to create it.

Fortunately, there are many institutions around the world that have made the commitment required to maintain digital archives. The Digital Endangered Languages and Musics Archives Network (DELAMAN, https://www.delaman.org/) is a network of member archives that specialize in language documentation data. Some of these archives are only digital, while others have physical holdings as well. Some specialize in a specific area of the world (e.g., the Archive of the Indigenous Languages of Latin America – AILLA, https://ailla.utexas.org/ – has a regional focus indicated in its title), others will accept material from anywhere in the world (e.g., the California Language Archive, CLA; https://cla.berkeley.edu/). Some archives will accept only data associated with a particular grant or funder (e.g., the Endangered Languages Archive – ELAR, https://www.elararchive.org/ – mainly accepts materials collected with grants from the Endangered Languages Documentation Programme), while others primarily accept materials in or about a specific language (e.g., Standing Rock Sioux Tribe Language and Culture Institute; https://dev-standing-rock.pantheonsite.io/collections). As part of the generalized increase in awareness of the endangered state of the majority of the world's languages, several Latin American institutions, including in Brazil, also committed to hosting local language archives; see Seifart et al. (2008) for an in-depth discussion of the early stages of this process. Unfortunately, many of these archives have not stood the test of time. Two repositories in Brazil, however, have continued to operate and are discussed here in §3; these are the archives of the Museu Paraense Emílio Goeldi (http://arqling.museu-goeldi.br) and the Museu do Indio (https://www.gov.br/museudoindio/pt-br). Other archives have also been created more recently, such as the Archivo de Lenguas, Culturas y Memorias Históricas del Ecuador, which has institutional support from FLACSO Ecuador (http://languages.flacso.edu.ec/).[8] By navigating via the links above, readers may compare the different online interfaces and capacities currently offered by these repositories.

While the above-named archives cover a wide range of geographic locations and collecting requirements, they might not be a good fit for all language documentation projects. Alternative archives include other types of digital data repositories such as those managed by many universities. Linguistic datasets (but not primary data from language documentation) may be submitted to the Tromsø Repository of Language and Linguistics (TROLLing, https://site.uit.no/trolling/about/). Primary and secondary data, as well as datasets, may be archived in general data repositories such as Zenodo (https://zenodo.org/) and the Harvard DataVerse (https://dataverse.harvard.edu/). Additional possibilities can be found in the Registry of Research Data Repositories (https://www.re3data.org/).

For researchers engaged in documentation projects, there are many advantages to archiving language documentation materials (data). These include simplified format migration, personal organization, accessibility, discoverability, graded access, rights management, collaboration, citability, and long-term digital preservation.

Anyone who has ever had to copy field recordings from an obsolete medium (e.g., cassette tapes) to another format, tried to open an old file in a new version of the program, or endeavored to find a program that can even open the old file, understands the headaches involved in format migration.[9] However, researchers who archive language documentation materials in a digital repository never need to worry about format migration again because it is part of the archival workflows that are overseen by the archive’s personnel.

Archiving also facilitates personal organization. Shortly after returning from the field, most researchers still remember where all of the files they created are located and how they are organized. However, as time passes, memory fades, and it becomes much harder to remember where files are, how they are organized, how they relate to each other, and the relevant metadata (i.e., contextual information; see Table 1) that will allow for their future reuse, either by the researcher who created them or by others. The situation is amplified for digital files, which cannot easily be distinguished and, thus, will need to be viewed or opened to be identified. The solution is to archive language documentation data, as well as the associated metadata, as soon as possible after they are created. Once the data and metadata are archived in a trustworthy repository, they can always be found and accessed again. Researchers no longer need to remember where files are stored (e.g., on which external hard drive, laptop, or cloud storage system), and they no longer need to juggle the work of handling multiple copies and backups of the data. Archiving helps to ensure that the data and metadata are not lost, thrown out, or forgotten about.

Once the materials and metadata are archived, they are accessible to (i.e., they can be accessed by) their creators, as well as by collaborators, other researchers, members of the language community, and anyone else who might need access to the data. Furthermore, digital language archives are constructed in ways that support discoverability of materials, that is, the ability of users to find or discover materials they are looking for. Discoverability is relevant both to searches within the archive itself (via faceted, targeted, or advanced searches, browsing, etc.), and to searches carried out via broader internet platforms such as Google indexing and metadata harvesting by the Open Language Archives Community (OLAC).

While there are digital repositories that claim to be purely open access, all digital language archives have rules for how users may interact with the holdings, and most digital language archives have graded access, by which certain materials may be restricted to particular users and/or conditions of use (see also §2.5). Regarding the rules of use, all digital language archives have some sort of terms or conditions that the online visitors must agree to before they can access the media files. Many digital archives require a user to create a free account and log in before they can access the media files. However, the catalog information (i.e., the metadata) is usually publicly accessible, meaning that anyone who lands on the archive’s web page can access and read it. Most language archives have some form of graded access to media files, although the way graded access works varies greatly between archives. AILLA uses numbered levels that indicate grades of access; other archives have specialized user roles and some use color coding to indicate who can access specific materials.

Most digital archives that specialize in language documentation data handle rights management similarly; however, exact details vary between archives. In most cases, the original rights holders[10] retain all of their intellectual and/or cultural property rights. The rights holders give non-exclusive licenses to the archive and the archive’s users; details of the licenses vary between archives and according to specific copyright laws of the country where each archive is located. Commercial use of the data is never allowed.

Data archiving facilitates collaboration on many levels. For researchers who are engaged in remote collaboration with the speech community or other researchers, archiving data as they are created or as analyses are finished helps the entire project team stay organized. Also, once data are organized and archived, they can be discovered by other researchers, which in turn can lead to new opportunities for collaboration. Finally, archiving data helps to facilitate the reuse of the primary data for various research purposes, including a great deal of work that is being done in the areas of natural language processing and linguistic typology. Finally, archived data can be cited, and this is crucial for creating reproducible research (BEREZ-KROEKER et al., 2018). Some journals, such as Language Documentation & Conservation, request that the data sets associated with an article be archived and cited appropriately. Thus, researchers can and should cite their own archived data, and researchers who use archived data must cite them as well.

Finally, a trustworthy digital repository has an active plan for the long-term digital preservation of the digital media files and associated metadata. According to the Digital Preservation Coalition (DPC), digital preservation “[r]efers to the series of managed activities necessary to ensure continued access to digital materials for as long as necessary. [...] [It] refers to all of the actions required to maintain access to digital materials beyond the limits of media failure or technological and organizational change.” Long-term preservation is the “continued access to digital materials, or at least to the information contained in them, indefinitely” (DPC, 2015). This means that digital preservation is much more than just backing up files. Digital preservation work includes migrating data and metadata from one format to another and from one software system to another as technology changes. It means ensuring accurate redundancy of data (meaning that there are duplicate copies stored on various media types and in multiple locations); and it means monitoring the health of all of the files in all of the locations on a regular basis. According to Owens (2018, p. 5), “preservation is the result of ongoing work of people and commitments of resources. The work is never finished […] It is not something that can be thought of as a one-time cost.”

Now that we have discussed the advantages of archiving language documentation data, we want to contrast digital archives with other online platforms that are commonly confused with digital archives, such as social media sharing platforms, websites, and cloud-based file storage.

Social media sharing platforms, like YouTube, Vimeo and SoundCloud (see Figure 2), facilitate sharing video and audio files with communities, speakers, and others because their content is easily discoverable via online search engines. They facilitate rights management because they allow the person who uploads the file to choose between traditional copyright or the application of a Creative Commons license.[11] However, most people do not have a sufficient understanding of either traditional copyright or the application of Creative Commons licenses to make an informed decision. These platforms also have their own versions of graded access in that the content may be kept private or made public.

Websites also facilitate content sharing, and code can be added to make the pages discoverable by search engines. While limiting access to content is possible, it can be a technically challenging process. Traditional copyright automatically applies to websites, but the website developer can choose to apply Rights Statements[12] or open licenses, such as those managed by Creative Commons, to the webpages and/or the linked files.

File storage systems, like DropBox, Google Drive, OneDrive, and Box, are good for sharing files with your collaborators while you are still working on them, as well as for controlling access to the files, but they do not have built-in rights management, and they are not discoverable in web searches.

While all these systems are useful for sharing data and might offer some form of content backup, none of them is committed to the long-term digital preservation of files or data. The terms of use of these platforms include their right to discontinue service and delete accounts and their contents. Moreover, some platforms, such as YouTube, retain copies of materials even when files are deleted by their owners/posters, or may be subject to automated downloads by other websites – an obvious problem where sensitive data are concerned (RICE, 2021).

Figure 2. Figure 2. Video on YouTube of a story told in Caquinte by Antonina Salazar Torres (recorded by Zachary O’Hagan, permanently archived with the California Language Archive, DOI: http://dx.doi.org/doi:10.7297/X2Z60M7W)

Prior to the 21st century, the results of language documentation were typically archived at the end of a researcher’s career or even after their death. However, with the advent of born-digital recording equipment, language archives, and data archiving requirements, most researchers today do not wait that long.

Today’s motto – as frequently voiced by archivists – is “Archive early and archive often!” Many language archives recommend progressive or incremental archiving. Under this model, a researcher or research team submits primary data (e.g., audio and video recordings and photographs), always accompanied by the relevant metadata (such as names of participants, date, location, languages used, and descriptive information to contextualize why the media files were created; see Table 1), to a digital repository as soon as possible after the media files are created (see step 1 in Figure 3 below). Secondary data (e.g., annotations, transcriptions, interlinear glossed texts) and analyses or academic output (steps 2 and 3, respectively, in Figure 3) are added later after they are finalized (ROBINSON, 2006; NATHAN, 2013; KUNG et al., 2020). In Figure 3, the steps represent waves of archiving for one recording session. Step 1, which involves archiving the primary data and relevant metadata, should be done as soon as the fieldwork or data collection phase is complete, or even while fieldwork or data collection are still in progress. Step 2, archiving the secondary data and relevant metadata, can be done iteratively as the transcriptions and translations are finalized. Step 3, archiving finalized analyses and academic output, along with relevant metadata, can also be done iteratively and well into the future. This approach recognizes that, for most researchers, or anyone else, there will never be a truly convenient time to archive language documentation data. The more time that passes and the more digital files and physical materials that accumulate between when data collection is done and when archiving is done, the harder and less convenient the archiving process will be, and the longer it will take. Moreover, since everything is fresh in a researcher’s mind immediately after a period of fieldwork or data collection, it is much easier at that point to organize materials accurately and to make sure that the metadata are thorough.[13]

Figure 3. Figure 3. Progressive archiving (graphic by S. Kung).

While archiving is key to the long-term conservation, preservation, accessibility and discoverability of language documentation data, not all data are equally appropriate for archiving, or for archiving in the same way. Some materials may be culturally or personally sensitive to the point that they simply should not be archived at all. Other materials may require restricted access, as noted in §2.3 above, such that only certain people may engage with them. Sometimes such restrictions may pertain in different ways to, for example, an audio file, a video file, or a written transcription of the same event. Examples of potentially sensitive material include esoteric or protected knowledge that is not meant to be shared with particular people or groups of people (e.g., community outsiders, members of other clans, men vs. women), personally damaging speech, and information that could endanger individuals or communities. Language documentation and archiving calls for open and ongoing communication among documenters, speakers/signers, and other community members in order to ensure a fully ethical and informed process, as we elaborate in §4 below.

Finally, as this section has explored, it is essential when archiving language documentation materials to choose a repository that will ensure their long-term preservation and accessibility. Many archives with the necessary infrastructure, including many of those in the DELAMAN consortium, serve an international community, and should not be viewed as proprietary with respect to whatever country might host them. Nonetheless, archiving initiatives are valuable at all levels – local, regional, national, and international. In some cases, it may be advisable to archive materials in more than one location, in order to meet local priorities while also leveraging higher-level infrastructure that offers the most reliable preservation and access (see §3.2 and §6 below).

The Museu Paraense Emílio Goeldi is a research institute associated with the Ministry of Science, Technology and Innovation (MCTI). Located in Belém, in the state of Pará, the Goeldi focuses on research on the Amazon. Its Linguistics Division, part of the Human Sciences Department, has had a precarious history. In 1986, the Linguistics Division had no recorder, no computer and only a few books. As is often the case in the country, there was little investment in infrastructure. However, there was a respected tradition of scientific collections at the Goeldi, and language documentation fit well into that tradition. Interns, fellowship holders and visiting researchers fostered increasing levels of activity within the Linguistics Division, including documentation.

The infrastructure began to improve with the acquisition, through research projects, of good quality cassette tape recorders, as well as microphones, laptops and solar equipment. In 1996, the World Bank’s Centers of Excellence program brought Digital Audio Tape (DAT) players, Hi-8 video recorders, a Hi-8 editing island and professional storage cabinets. On the basis of accumulated experience, Goeldi linguists, including fellows and affiliates, have successfully competed in international documentation programs that are favorable to projects in developing countries. In the early 2000s, projects from the Endangered Languages ​​Documentation Programme (ELDP) and the Dokumentation Bedrohter Sprachen program (DoBeS) provided more equipment and experience. Documentation and procurement of equipment (e.g. solid state recorders with flash memory) continued into the second half of the decade, with support from USAID, UNESCO and the Ambassador’s Fund (US Embassy). MCTI supplied equipment in 2007, including a server as the basis for a digital repository. Documentation (and research) projects continued in the second half of the 2000s, with support from CNPq, DoBEeS and ELDP. This progress was threatened in 2007 by fulminating attacks by a small group of linguists, who opposed ProDocLin (Projeto de Documentação de Línguas Indígenas, the linguistic documentation program based at the Museo do Índio ^[15] ), international documentation programs and digital archives (MOORE and GALUCIO, 2016, p. 40). Fortunately, this opposition did not attract support among Brazilian linguists and progress continued, both at the Museu Goeldi and in Brazil.

Through 2009, the activities of linguists based at the Goeldi resulted in a large set of diverse recording media: cassette tapes, DAT tapes, mini-disks, Hi-8 tapes, mini-DV tapes, etc. Due to time and budget constraints, this collection was simply placed in drawers, without systematization. From 2009 to 2014, linguist Ana Vilacy Galucio coordinated a project supported by the Ministry of Justice (Fundo de Defesa de Direitos Difusos) to digitize and catalog the collection, following the practices of the best international archives. To streamline the process of cataloging and storing recordings and their metadata, scripts were created by Sebastian Drude and Rose Costa, reducing the time required by 70%. Thus the Goeldi’s digital linguistic archive was created.

Subsequent years saw more equipment purchased and more documentation projects conducted. Throughout the evolution of the Goeldi Linguistics Division archive, technology has evolved and formats and physical media have changed. However, the conversion of all recordings to standardized digital form has mitigated problems of longevity and compatibility. Some devices, such as microphones, do not become obsolete, and the various types of recorders (DAT, mini-DV, etc.) that are currently not used directly for documentation are maintained, since they can still be used to access legacy recordings and produce a signal that can be digitized. Original recordings are stored on their original media in professional storage cabinets.

Currently, recordings in digital form are stored on a 32 TB Network Attached Storage (NAS). For security, in case of failure of one of the NAS hard drives, RAID-6 redundancy is implemented, reducing the storage capacity to 22 TB. The contents of that NAS are copied to the new 96 TB NAS, whose capacity is reduced to 72 TB because of redundancy (see Figure 4). To ensure their safety, in case of risks such as fire or lightning, the two NAS must be kept in different buildings. Audio recordings are saved in their original format, i.e. as .wav files. Video recordings are saved in their highest definition format and also in compressed (.mpg) format, which takes up much less space. These compressed files and audio files are cataloged and stored using Language Archiving Technology (LAT) software (see KOENIG et al, 2009). Most of them are backed up on the server of the Information and Communication Technology Sector of the Goeldi Museum.

Figure 4. Figure 4. 96 TB Network Attached Storage (NAS) in the Museu Goeldi Archive (photograph: Denny Moore).

Currently, the Linguistics Area digital archive contains records of 80 indigenous languages​. Of these, 73 are completely digitized, catalogued and stored in the LAT software. The LAT files take up 2.49 TB of space; the temporal duration of these recordings is 1,561 hours of audio and 474 hours of video. In addition to these files, the archive also contains 9 TB of raw recordings, including High Definition video, which take up a significant amount of space. Edited works, mainly for community use, constitute another 1.5 TB. Another 6 TB is occupied by photographs and loose files of the researchers. The metadata of the LAT files on the server can be found at http://arqling.museu-goeldi.br. The recordings, however, are currently not available for download, pending reorganization of the archive and resolution of access issues.

To contribute to the development of indigenous language documentation in Brazil, the linguists at the Goeldi disseminate widely, through intensive training, knowledge of the technology and methodology involved, including suggestions for equipment. This training is also carried out among indigenous groups. Moreover, the archive offers digitization and storage services for legacy recordings (see §5 below). For example, the Goeldi team is currently working with an anthropologist to digitize his collection of 115 cassette tapes of two Nambikwara dialects. The recordings, some from the 1970s, include music, a dictionary, conversations, and minimal tone pairs. The recordings are being digitized and, in collaboration with the anthropologist, will soon be catalogued and deposited in the digital linguistic archive. He will receive copies in .mp3 format, which take up little space and can be returned to Nambikwara communities. Cassette tapes can be deposited in the archive or returned to the researcher. People interested in this service should contact the Goeldi Linguistics Division (linguistica@museu-goeldi.br).

Alongside the Goeldi Museum, few other Brazilian institutions are involved in archiving initiatives. Currently, the principal effort is based at the Museu do Índio in Rio de Janeiro, a component of the Fundação Nacional do Índio. The Museu do Índio and the Museu Goeldi coordinate the technical and methodological aspects of their respective archives. Over time, the Museu do Índio has systematically increased its infrastructure and has developed impressive digital storage and backup capacities. The Museu do Índio has also been conducting active projects to document indigenous languages ​​and cultures, following the best international practices. Its digital archive contains approximately 10 TB of linguistic documentation, including the languages ​​of 25 indigenous groups. It also contains 9 TB of cultural documentation from 27 indigenous groups, which includes recordings of linguistic interest (Thais Tavares MARTINS, Chefe do Serviço de Referências Documentais, Museu do Índio; p.c. to Denny Moore, 2020). Outputs of the ‘Indigenous Sonorities’ project (focusing on cultural preservation via audio documentation) for five indigenous groups occupy 1.69 TB. The total content of the archive occupies approximately 50 TB. Copies of all recordings are returned to the communities involved, but Internet access is still being resolved (see https://www.gov.br/museudoindio/pt-br).

Various well-organized linguistic documentation projects, led by Brazilian linguists, are ongoing and are producing significant amounts of recordings, steadily increasing the need for professional archiving. For example, the Documentation Center of the Federal University of Amapá in Oiapoque has documentation projects with the Karipuna and Galibi-Marworno (coordinated by Gelsama Mara Ferreira dos Santos) and with the Palikur (coordinated by Elissandra Barros da Silva). Data are currently being stored on microcomputers and external hard drives. The project with the Galibi-Marworno has 13.5 GB of audio (.wav) and 403 GB of video (.mov, .mp4), as well as 51 GB of photos and 15 GB of edited video. The Palikur project has 20 GB of audio, 6 TB of video (.mp4) and over 100,000 photographs in RAW format, for a total of 10 TB (G. M. F. DOS SANTOS and E. B. DA SILVA, p.c. to Denny Moore, 2020).

Clearly, the demand for language documentation archives is growing in Brazil, as elsewhere, and is presently far above the current capacity. One possible solution to the need for more capacity is the creation of regional digital archives. These would not only increase the current capacity, but they would also have the advantage of relative proximity and visibility to the indigenous peoples of the region. This could facilitate both further recording and indigenous groups’ access to those recordings. A natural place for archives of this type is indigenous training programs at universities, for example at the Universidade de Amazonas, the Universidade Federal de Goiás, or the Universidade Federal do Amapá in Oiapoque. In these programs, indigenous students receive computer training while maintaining contact with their respective groups, thus facilitating a productive cooperation between the archive and the groups in question. The infrastructure for an archive of this type can be bought all at once or, as in the case of the Goeldi, built up in stages, in tandem with research or documentation projects. Indeed, the components are relatively affordable: The cost of a 48 TB NAS (Network Attached Storage) was approximately R$20,000 in the middle of 2021 (with the dollar at R$5.80). For digital language documentation, a complete set of excellent portable semi-professional equipment (camcorder, tripod, digital audio recorder, laptop, three types of microphones, video light, carrying case, batteries, cords, adapters) was approximately R$20,000. At least two kits would be needed to equip a regional documentation center and archive. In addition to infrastructure, training in documentation technology and methodology is required to ensure quality, which is always a challenge. One issue to be resolved is the need for an archiving software that is reasonably simple and user-friendly.

Ultimately, the archiving options presented in this section represent different levels of capacity regarding the major goals of preservation, conservation, and accessibility, in keeping with the observations presented at the end of §2 above. While smaller-scale archiving initiatives such as those proposed here may not meet DELAMAN standards with respect to some of these goals, they are certainly a step in the right direction, and in some cases they may offer more options for accessibility even where they are less developed in their capacity for long-term preservation. Moreover, as noted above, researchers should be aware of the possibility of archiving bothat a local or regional level and also in a higher-infrastructure archive. This possibility is illustrated by the Paresi and Enawene Nawe case studies described in §6 below.

Discussions of ethics and ethical best practices have figured prominently in the language documentation literature. In these discussions, particular attention has been paid to collaboration (GLENN, 2009; LEONARD and HAYNES, 2010), community involvement and engagement (YAMADA, 2007; CZAYKOWSKA-HIGGINS, 2009; SAPIÉN, 2018; BISCHOFF and JANY, 2018), and the applicability of ethical models to different parts of the world (DOBRIN, 2008; HOLTON, 2009; PÉREZ BÁEZ, ROGERS and ROSÉS LABRADA, 2016). The intersection of archiving and ethics has also received considerable attention (e.g., Macri and Sarmento (2010) and Innes (2010), both of whom explored ethical issues related to archives). In these discussions, informed consent has played a key role, with multiple researchers questioning how ‘informed’ the consent really is (GRINEVALD, 2006; ROBINSON, 2010). An overview of some of the issues relating to informed consent may be found in Dwyer (2006, p. 43-48).

Informed consent is a legal – at least in many places, including Brazil – and moral obligation. It is, therefore, essential that community members understand what their participation entails and, crucially, what the consequences of digital archiving really are. One potential challenge to informed consent with respect to archiving is that of familiarity with archives and the Internet – and in some instances, computers – on the part of the community and/or the speakers/signers themselves. As Robinson (2010, p. 189) asks, “can we obtain truly informed consent [for archiving] if the consultant and the community have never seen a computer or heard of an archive or the Internet?” There are two additional issues here. Firstly, research is by default open-ended. Linguists generally gather a corpus based on particular research questions but those questions, or rather the answers to them, may lead to new questions. Thus, when you gather consent at the beginning of a project, if the goals of the project change or new research questions arise, does the initial consent hold? This is particularly difficult to anticipate with archived data, which, as we will see below, can be reused by others. Secondly, the descendants and communities of tomorrow may need these materials for revitalization purposes, as described for example in Bomfim (2017). In what follows, we exemplify these issues and propose one approach, among many possible ones, to address some of these concerns.

This case study reflects on the experience of co-author Rosés Labrada in explaining archiving to members of the Mako communities along the Ventuari River in Venezuela during his doctoral project (2012-2015) focused on the documentation and description of Mako, a Jodï-Sáliban language (ROSÉS LABRADA, 2015).

In 2011, Rosés Labrada undertook a trip to several Mako communities along the Ventuari River and its tributaries in order to gather community consent for the project as a whole. As part of this process and attending to local protocols, village meetings in several communities were held, which resulted in an invitation to return to work in two communities: Arena Blanca and San José de Yureva. However, no data were collected on that first trip. The documentation project thus started in July 2012, and Arena Blanca was selected as the site of the first field stay due to its composition as a largely homogenous Mako-speaking community. ^[16]

Initial preparation for a community-wide discussion about ethics and informed consent took place on July 14, 2012 in a meeting with two community members who were also the school teachers in Arena Blanca at the time and who had agreed to act as translators for the meeting. A clear outcome of the meeting with the schoolteachers was the realization that the concept of archiving was likely to be difficult for the outside linguist to explain in ways that would be clearly understood due to the complexity of some of the concepts and technicalities behind them. Of significant concern was the fact that, at the time, there was no connectivity for the Internet or phones, there were no computers in the community, and both the teachers had limited experience with this technology.

The community-wide meeting on July 15 was well attended by many community adults, who engaged in discussion of the project and potential ethical issues posed by recording and archiving audiovisual materials. Because this was a meeting to seek permission to record, the meeting itself was not recorded. However, the permission seeking process advanced significantly during this meeting:

1. recording with both video and audio was accepted;

2. taking pictures was also accepted but with the caveat that pictures of unclothed children or topless women were to be avoided as the community had abandoned those traditional practices;

3. keeping the data beyond the duration of the project and sharing it outside the community was also accepted.

Nevertheless, Rosés Labrada sensed that the concept of archiving remained “fuzzy”, particularly in two areas: (1) the implications of the online sharing of materials that identify specific individuals and (2) the options for sharing or restricting access to those materials or the names of the individuals. Thus, the linguist took steps to make sure that the community really understood what was involved in the archiving process. His goal was to show them what the collection in the archive would look like, but the challenge came from the lack of Internet access in the community.

The solution adopted was to take screenshots of an initial archival deposit to show what the archive’s website looked like and what it would be like to navigate through it. In preparation, Rosés Labrada and one of the schoolteachers visited the AILLA and ELAR websites while in Puerto Ayacucho, the capital of Amazonas state, and the schoolteacher agreed to deposit one of his recorded stories to start a collection that was to be double-archived in both archives. ^[17] After the deposit of a folder with all the items corresponding to one single story, both archives supplied screenshots of what the collection and the deposit looked like.

A second community-wide meeting was held on November 4, 2012 in Arena Blanca after Rosés Labrada’s return with these screenshots. The discussion that ensued was better informed in a number of aspects. First, one request was that originals stay in the community. This request allowed the linguist to provide further explanation about the context of digital documentation: that specific recordings were born digital, that originals and copies could be identical, and that copies could be easily made—which would not necessarily have been the case if Rosés Labrada had recorded directly on cassettes or CDs/DVDs, which the community was more familiar with.

A second issue that came up was the question of access for community members and a distinction between those “who know” – meaning those who know how to use computers and the Internet – versus those “who don’t know.” The general consensus was that those who knew could potentially go to the Internet and use these archived materials but that for those who did not know, making copies in DVDs and CDs was needed. A third issue concerned the content of the recordings themselves; a community member expressed a preoccupation regarding recordings that may contain crude jokes or curse words and who could listen. For these, it did not seem to be a problem that people in other parts of the world could listen to these recordings. Rather, this particular community member was concerned about the potential reactions of Mako people from other nearby villages who might listen and not understand that these jokes were truly meant in jest. Finally, there was discussion of videos of traditional activities as being harmless, when compared to some other videos that the community was aware of, such as violent movies. This provided an opportunity to explain further that individual speakers had agency to restrict access both at the time of recording but also into the future, and that specific access provisions could be enabled and changed at any time in the future. Overall, this discussion reassured Rosés Labrada that community members in Arena Blanca had a clearer understanding – when compared to the initial meeting in July 2012 – of some of the ethical issues around informed consent, archiving and access.

This small case study illustrates three main points: 1) it is crucial that documentary linguists explain archiving in ways that communities and individual participants can understand; 2) communities and participants should be able to revisit their decisions; and 3) linguists should try to anticipate future uses of the materials and make provisions where possible, while being fully aware that we cannot foresee the future. Ultimately, ethical choices are not necessarily defined in black and white, and what works in one specific context may not necessarily work in another (HOLTON, 2009; GASSER, 2017). However, as long as linguists aim to uphold principles of respect, reciprocity, responsibility, and relationships, as advocated by K. Rice (2006), we should be able to avoid some of the potential ethical pitfalls that could arise as part of the archiving process.

As the Mako case illustrates, having and maintaining access to language documentation materials tends to be a key priority for communities, not only for those people who are directly involved in documentation, but also – and sometimes even more so – for their descendants (e.g., DWYER, 2006, p. 59; VAPNARSKY, 2020; R. MILLER, 2021). Yet effectively returning language documentation materials to communities can be a complex and multi-stage process, and must take into account the different capacities community members have to access these materials. These capacities may be constrained not only by a lack of access to the internet, through which archived materials in many digital repositories can be viewed and downloaded, but also by the skills required for navigating online interfaces (as well as the languages in which these interfaces might be presented, such as English) and by basic computer literacy. For members of some communities, access even to static technologies such as CD and DVD players may be limited, as may be the capacity to read printed materials in the community language (or otherwise).

Enabling access may, thus, require creative solutions. One model that has met with considerable success is the “distributed digital audiovisual archive” or “jukebox archive” (BARWICK, 2004; BARWICK et al., 2005; O’MEARA and GONZÁLEZ GUADARRAMA, 2016). This initiative, originally piloted in Australia, involves setting up a computer in a community center or other neutral location where community members can easily access it, and does not require a connection to the internet. The jukebox computer contains language documentation materials – audio and video recordings, as well as (potentially) transcriptions, photographs, etc. – in easily movable file formats (such as .mp3), as well as the capacity to burn CDs/DVDs and to download files onto a flash drive, phone, or .mp3 player. Community members can thus take copies of materials home with them, and can also add to the jukebox by uploading their own. Another useful model relies on a local wifi transmitter called a ‘Raspberry Pi’, which is effective in contexts where people have smartphones and the associated skills, but have limited or no computer literacy (e.g. THIEBERGER, 2019).

Of course, ethical considerations are relevant to all these initiatives – and not only to the process of enabling and maintaining access to documentary materials, but also in considering how and whether community members wish to limit their access and use (see the sections above; also DEBENPORT, 2010; MACRI and SARMENTO, 2010). Some recorded material may be viewed as potentially damaging, dangerous, offensive, or otherwise private, and understandings of who should or should not have access to what material may involve complex and intersecting categories, associated with particular conceptions of outsider/insider, gender, clan, neighbor, relative, etc. – which may be much more complex than the blanket term ‘community’ implies. As pointed out above, decisions about access, like other facets of documentation and archiving, must be continually informed by ethical principles.

Researchers who work or have worked with speakers of indigenous languages are often in personal possession of rich documentary collections including field notes, sound recordings, photographs, and film, which may be stored in their offices or homes in sub-optimal or risky conditions (e.g., damage from humidity, insects, flooding). These legacy materials – even when they are not especially numerous – are of high linguistic, cultural, historical, and personal value for individuals and larger groups. For under-documented languages, legacy materials may constitute the only early records of language use, to say nothing of the other aspects of life that they often capture. Unlike widely spoken languages such as Portuguese, for which historical records are readily at hand, historical records of many indigenous languages often persist only by way of diligent efforts to archive legacy materials. As Austin (2017, p. 23, emphasis added) stresses, “For projects interested in documenting, describing or revitalizing languages, especially endangered languages, historically existing materials (whether digital or analog) like tape recordings made in earlier times or written materials collected years or even centuries ago may exist and may represent important sources of information, indeed, in some cases, the only information available.” Linguists have a professional responsibility to be aware of the existence of relevant legacy materials, and, when possible, work toward preserving them for and in collaboration with communities, for linguistics and related disciplines, and for posterity generally. This allows for the subsequent philological analysis of the documentary record, “recogniz[ing] the documentary filiation which characterizes all linguistic data as they are successively recorded, interpreted, and analyzed” (GODDARD, 1973, p. 727). At the same time, linguists have a responsibility to ensure that their engagement with legacy materials is carried out in an ethical fashion (O’MEARA and GOOD, 2010; see also §4 above).

The preservation of legacy materials involves certain key steps: locating the materials; appraising their physical condition; organizing them in a basic way; transporting them to an archive or other (temporary) safe location; carrying out any necessary conservation remediation; cataloging them; and, ideally, digitizing them (perhaps not necessarily in this order). Locating legacy materials is often one of the more difficult steps, requiring either inadvertently being made aware of their existence and location, or a good deal of sleuthing. In general, locating legacy materials is made easier by having a deep familiarity with the histories of research and other activities such as exploration in particular regions across many disciplines. More concretely, one should pay attention to descriptions, often in the methodology or similar section, of early publications (and unpublished works, when available).

For example, the late anthropologist Gerald Weiss (see O’HAGAN, 2021 for background), in a section of his PhD dissertation titled “Design and Method,” had the following to say about the output of his fieldwork in Ashaninka communities along the Tambo River of Peru from 1961 to 1964 (WEISS, 1969, p. 6):

The techniques employed in the field for obtaining information were standard, adjusted only to the particular requirements of the information being gathered. A journal was kept; temperature, humidity and precipitation readings were recorded daily in c information obtained from observation and interrogation were accumulated on four-by-six slips or some other convenient form – in duplicate, with date, place and informant’s name indicated on each slip; specimens, photographs, and tape recordings were taken of everything possible.

Co-author O’Hagan was familiar with Weiss’s work in the context of his own fieldwork with speakers of the related Caquinte language, but in previous readings of his dissertation had skipped over the section “Design and Method.” From it, however, he learned of the existence – at least at a certain point in time – of field notes, index cards (cf. slips), biological specimens, photographs, and sound recordings. Gerald Weiss became a professor of anthropology at Florida Atlantic University; a Google search in March 2021 revealed that he was still affiliated with that institution as an emeritus professor. Furthermore, a phone call at the same time revealed that all the above-listed materials – and more from postdoctoral research – were kept at his home in Florida (see ANWAR, 2021 for more details).

The basic organization of legacy materials can be difficult if the objects in question (e.g., field notes, tapes) have little metadata associated with them, and if specialists who might be able to read or understand the spoken version of a particular language cannot be consulted. Additionally, some objects may be in a format that is not easily accessed, for example, as is often the case with the analog reel-to-reel tapes that were commonplace from the 1950s to the 1970s, an especially valuable period in the documentation of many languages. Depending on the specific situation, at this juncture it may be beneficial to collaborate with community members who can assist in the interpretation of the materials, or to first have them digitized, either by an external professional audio technician ^[18] or by an archive as part of a donation. Regardless, the main goal of this step of organization is to produce a basic inventory of objects (how many notebooks, tapes, etc.) before they are transported, so that there is a record that can be corroborated to ensure that all the materials arrive at their intended destination. If metadata are lacking, other descriptions of the physical objects can be given (e.g., “two cassette tapes with red design on cover”).

The goal in the transportation of legacy materials is for the materials to be sent to a location that either (on a temporary basis) has relatively better storage conditions (e.g., one that is less humid) or facilitates the transfer of them to an archive, if not to an archive directly. The moment of transportation can be one of the most perilous for archival materials. Ideally, especially delicate materials like tapes can be transported personally, for example, in carry-on luggage. ^[19] For other materials, it is important to keep them dry (e.g., with plastic coverings) and separate from others that might damage them. If materials are to be shipped, make sure to use high-quality cardboard boxes (with ample tape and bubble wrap), and, when financially possible, opt for airmail over ground transportation, as this usually involves less repeated handling of boxes. In general, at this stage it is best to be in contact with an archive that can provide you with additional guidance. Archives will also be the best informed to advise you on cataloging and digitization. We emphasize that digitization of analog sound recordings and film is in most situations relatively urgent, as the original tape degrades over time, and in most situations the prior storage conditions of the materials are not at an archival standard. Humidity and water damage are especially common.

Legacy materials often preserve the voices and knowledge of people who preceded other documentation projects by many decades, and who in some cases represent the last first-language speakers or semi-speakers. In such cases, legacy materials may be a key source of information for community efforts in language revitalization and in maintaining or reclaiming cultural heritage, as well as for scholars’ efforts to understand the breadth and depth of human expression.

Once materials have been archived, many years may pass before they are utilized by others. For example, linguist Catherine Callaghan made a series of recordings of Sarah Ballard speaking Bodega Miwok (Miwokan; California) in 1960, while the former was a graduate student in linguistics at the University of California, Berkeley. She donated these and other recordings to the California Language Archive ^[20] in 1979 (BALLARD and CALLAGHAN, 1960). Forty years later, in 2019, linguist Andrew Cowell made time-aligned transcriptions of these recordings in ELAN, which he archived with the same repository (COWELL, 2019+). ^[21] The two collections are linked in the archive’s digital catalog, and so can easily be related to each other when consulting one or the other. Similarly, linguist Gladwyn Kingsley Noble, Jr. made recordings of speakers of Wapichan and Atorad (Arawak; Brazil, Guyana) during a single field trip to Guyana in 1965. After wending their way through the hands of different academics, the 13 tapes were donated to the California Language Archive around 2006 by Manjari Ohala (see GEORGE et al., 1965). The recordings of Atorad, formerly thought to no longer have any first-language speakers, were utilized for a preliminary phonological description of the language by O’Hagan (2018), which was elaborated on by E. Miller (2021) as part of an undergraduate honors thesis. In 2021, the texts included in the recordings were translated by remaining speakers now resident in Wapichan communities, and served as a basis for a workshop dedicated to Atorad language in 2022 (K. RYBKA, p.c. to Zachary O’Hagan).

Finally, we stress that legacy materials can be utilized on an ongoing basis, and that usage requires interpretation – a creative process of understanding of “what those materials meant to their creators, what new meanings they might take on in the context in which they are being used, and what roles [contemporary actors] themselves as persons might play in the materials’ circulation and reception” (DOBRIN and SCHWARTZ, 2021, p. 23). One especially productive example of ongoing work with legacy materials is the biennial Breath of Life Archival Institute for Indigenous California Languages (see GEHR, 2013) held at the University of California, Berkeley, organized by the Advocates for Indigenous California Language Survival (AICLS) in conjunction with the Survey of California and Other Indian Languages, which houses the California Language Archive. The Institute brings indigenous people to the Berkeley campus to collaborate with volunteer linguists in the interpretation of archival materials, many of which date back to the early 20th century. This model has been successfully expanded to other locales (BALDWIN, PÉREZ BÁEZ and HINTON, 2018) and it may be fruitful to consider whether a similar initiative in Brazil could increase the accessibility of existing language materials.

The Paresi language documentation project started in 2006 and was developed via Brandão’s doctoral and postdoctoral research on the language. The Enawene Nawe language documentation project began in May 2019, and ended in December 2022. The goals of both projects were to organize a large corpus of audio and video recordings, spanning a variety of linguistic genres, and transcribed in the relevant indigenous language ​​and translated into Portuguese.

The documentation of Paresi was initiated at the request of the Rio Formoso community, who were interested in recording their traditional culture. In the Enawene Nawe community, Brandão also got in touch with a speaker of the language who invited her to visit them to present the project proposal, and she obtained permission from the communities to record material within a non-profit model. Communities benefited from the production of DVDs, CDs and USB drives for accessing traditional stories and songs, and from the training of indigenous teachers in linguistic documentation. The Paresi material also served as a basis for the elaboration of a reference grammar of the language, which was defended as a doctoral thesis (BRANDÃO, 2014).

The project participants included indigenous teachers and elders who were knowledgeable about traditional culture. Some teachers worked on recording, and others on transcription, translation, and metadata organization. One of the Paresi participants also worked on video editing. A key Paresi collaborator, Jurandir Zezokiware, participated in the documentation project with the Enawene Nawe community; his presence was very important in establishing a relationship of trust with this community. He assisted during trainings and collected linguistic data. The work also involved undergraduate students at the Universidade Federal do Pará (UFPA), who visited the Paresi and Enawene Nawe communities to learn about fieldwork.

Both projects allowed for high quality equipment to be purchased for the Paresi and Enawene Nawe via financial support from the ELDP. This equipment included a Zoom digital recorder, Shure head microphones, a digital video camera, an external microphone for the camcorder, a tripod, and other materials. Various cultural and speech events were documented throughout the projects. The Paresi communities selected traditional stories, songs, blessings, traditional festivals, indigenous games, etc. for documentation, and Brandão and her team also recorded different dialects or varieties of the Paresi language. The Enawene Nawe preferred to focus on recording traditional stories (e.g. their origin story, the account of the origin of cassava, a story about spirits, and others). More information about the two projects is provided in Brandão and Zezokiware (2018).

Figure 5. Figure 5. Paresi men wearing traditional clothing (PAB-200712-AP-RC-PontePedra80.JPG, photograph: Rose Costa); Enawene Nawe telling stories (UNK-20190500-AP-treinamentos-136.jpg, photograph: Ana Paula Brandão).

The primary data – audio and video recordings – were recorded in WAV and MTS formats, respectively. Each session was recorded in audio and video. Later, the MTS files were converted to MP4 for archiving, to yield a more compact format for storage. The file names begin with the ISO code for the language (PAB for Paresi and UNK for Enawene Nawe), followed by the recording date (in YYYYMMDD format), the abbreviations of the names of the person who recorded and of the primary speaker, and the keyword of the session in which the file is included; for example: UNK-20200114-WE-YI-Kolito.mts. These documentation projects have generated digital collections involving more than 150 hours of Paresi recordings and 37 hours of Enawene Nawe recordings, which are organized into seven categories, as illustrated in Table 2.

Table 2. Table 2. Paresi and Enawene Nawe digital recording.

The secondary data produced in these projects include annotations and metadata associated with the recordings. Annotations, consisting of transcriptions and translations into Portuguese, were made using the ELAN program (EUDICO Linguistic Annotator 2020) and Microsoft Word. Lexicons for both languages ​​were also developed, and several Paresi texts were interlinearized using the FLEx program (FieldWorks Language Explorer 2019); examples from these texts were used in the analysis presented in the reference grammar. The transcriptions and translations of the texts were made by the speakers. Most of the transcriptions were produced during the period when the lead researcher was not in the communities and, in the case of Paresi, they were later reviewed with the speakers. Paresi speakers received training in the use of ELAN, while Enawene Nawe speakers had little interaction with this program and preferred to transcribe in Microsoft Word using a computer or cell phone. The notes from the Enawene Nawe recordings are still being organized, such that all material will be transferred to ELAN and can later be published in a digital collection.

Files in ELAN have at least three lines of annotations: the transcription, the translation, and the notes. More detailed information about the glosses of Paresi or Enawene Nawe morphemes (i.e. the interlinearization) is organized in the FLEx program, as it allows automatic insertion of glosses (once they are already entered in the database), unlike ELAN. Neither language has an established orthography that is used consistently in indigenous schools. Therefore, transcriptions were made in the orthographies that the speakers know and, in the case of Paresi, they were systematized according to the orthography that Brandão proposed for the language.

The metadata were initially organized in an Excel spreadsheet, then entered into the Lameta program (HALTON et al., 2021) to create IMDI files (ISLE Metadata Initiative) for each session. Each session groups together an audio file, video file, and annotation in ELAN (EAF format) or elicitations in PDF. The IMDI files contain information about the recordings, such as the ‘actors’ (people involved), subject, content description, and keywords. There are also sets of photographs organized into the following categories: community, craft, school, daily life, researchers, training, people, and work.

All data from the Paresi project are stored in the archive of the Museu Paraense Emílio Goeldi, but are not yet available to the public. Twenty hours of material collected during the period of financial support from ELDP (2011-2012) were also stored in the archive associated with this institution, the Endangered Language Archive (ELAR). ^[23] Another thirty hours are stored at AILLA, ^[24] with some materials available online. Part of the material collected in the Enawene Nawe project was organized in ELAR ^[25] and will also be made available in the Goeldi archive.

The archives referenced here employ a set of graded access codes, which indicate the different levels of access available for different files. ELAR has the following codes: O for free access; U for materials that can be accessed by creating an account in the collection; and S for materials that are restricted and can only be accessed with the depositor's permission. Files marked with the access code S may contain personal or sensitive information about the speakers, such as conversations and life stories, or may be currently under analysis by the depositor. AILLA's graded access, in contrast, is not based on user roles, but rather it utilizes numbered levels to indicate when (if ever) or how a file may be accessed. Some archives require users to complete a free registration on the website before any materials may be accessed.

These archives are available in English (ELAR) or in English and Spanish (AILLA), but they do not yet have access information in Portuguese, which makes it more difficult for Paresi and Enawene Nawe people to access them. In the future, Brandão intends to make these collections available in regional digital archives near the communities, potentially housed in universities (e.g. the Federal University of Mato Grosso [UFMT]), intercultural colleges, and/or indigenous schools. Speakers could thereby have physical access to the server where the materials are stored, as well as easier access via the internet.

To conclude this section, the two documentation projects presented here have enabled the creation of Paresi and Enawene Nawe archival collections, which are available both to the academic community and to indigenous community members. Importantly, these collections are also a way of safeguarding traditional knowledge, as has been made even more urgent through the loss of many indigenous elders during the COVID-19 pandemic. It is crucial that language documentation materials that have not yet been deposited in archives be archived to ensure their long-term preservation. An important next step is to make these collections increasingly accessible to indigenous peoples in Brazil. In this way, indigenous teachers will be able to use the materials in community classrooms and for academic work within indigenous colleges.