Taking the Potholes out of the Road: Prepublication Recording as a Means of Editing in Oral Preference Cultures

2.1 HearThis with single voice recording

For both projects, HearThis on computers with headsets was the primary recording tool. The Mborena Kam team had received more formal education and as a result, their recording tended to go much faster and was completed by a handful of men who quickly learned to handle the equipment and do their assigned books/chapters. The staff of Pioneer Bible Translators provided the technical support in transferring all the files back and forth from the various computers and keeping track of the status of each recording. Notes were also added in Paratext about possible revisions suggested by the recording team, with frequent send/receives in order to keep the text current with the revisions that the translation team was making.

In contrast, there were over thirty people who volunteered to do the recording in Apal. Some readers only completed one chapter. Others completed entire books. Many of those doing the recording were women, some of whom had just barely learned to read when they started to do the recording process on computers. The volunteers also included a group of children who recorded Luke.

HearThis is not a perfect tool, but over the years I have seen improvements in the program. One feature that was added was to blank out surrounding sentences so that the reader can only see the sentence that should be recorded in that particular location. A feature was also added that allows you to shift blocks of sound files rather than having to manually renumber them. When a published New Testament is being recorded, the text does not change, but in prepublication recording, sentences might be added, combined, or divided. Those changes can create havoc in the recorded data.

2.2 Check and correct recordings until they match the written text

Pioneer Bible Translators did not have a resident translation advisor among the Mbore people who could provide quality control for the Mborena Kam team, so the recording volunteers listened to each other’s recordings and identified any recordings that did not match the text. This was not 100% successful, as can be seen from the quantity of inaccurate recordings found in their recording of Mark (see section 3.1 below).

In the Apal group, I was the quality control person at the initial stages. Because of the overall lower level of education in the group, most recordings required that 30% to 50% of the sentences had to be redone multiple times before being sent to the leaders for listening.

2.3 Leaders listen to chapters or sections as units and then sentence by sentence

With the Mborena Kam team, we used printouts and an external monitor during the listening process. All of their leaders were literate, though those from the minority dialect had more difficulty reading the translation. Many of their discussions were about those dialect issues and how to spell words so that both dialects could read the printed translation well. Many of the leaders attended all sessions and listened to the entire New Testament, but at times there were substitutes sent so that their village would still be represented by someone.

In the Apal group, most of the leaders were just functionally literate and could not fluently read the Scriptures. We did the checking in the village using an external monitor that we could point to when there was a need for discussion. I often wondered if playing the MP3 file of the entire chapter was worth the time, but, if nothing else, it allowed them time to sit and smoke outside the house. When I thought they were not listening, one of them would yell, “Mark that sentence because it is wrong.” One time a leader at a distance came running back and said, “That has to be changed. I heard it and ran as fast as I could.” Another time, a person whom I thought was totally asleep sat up straight and said, “What was floating around in the water? That sounds like Noah is floating in the water instead of being in the big canoe that was floating on the water.” When the translator was shown the verse, he said, “Yes, that verse is wrong. We will have to correct it.” So, even though people appeared not to be listening, I was often wrong in my evaluation of what was happening.

In both groups, the section headings and footnotes had to be read and checked because they were not recorded. This gave the teams more reading practice and occasionally they found an error, or something was ambiguous enough that they sent it back to the translation team to be revised.

2.4 Errors in recordings sent back to recording team

Sometimes there were errors in recordings that were simply reading errors. I could understand this in the Mborena Kam group. However, in Apal I had listened to each sound file a minimum of four times, and, despite that, there were times where a reader had corrected the part of the sentence that I had asked him to correct but in the process had made another error in a different part of the sentence. The leaders caught these errors and sent them back to the recording team.

Sometimes the person recording the verse had followed the actual printed text, but when the leaders thought about it more, they decided they did not like the translation choice. One example of this in Apal was the word for “trumpet” (such as in Rev 8.2), which was translated as a word that means “bamboo bugle.” The person recording the word pronounced it exactly as it was spelled according to the rules that we had applied to all loan words. It was approved in several of the books that were recorded first, but about midway through the process, they heard it and said, “We do not say biugelɨ. We say bihulɨ.” Everyone agreed and as a result we had to find and change both the text and the recordings of the thirty-two sentences with that word in it. Revising the translation itself just takes a few minutes, but in the prepublication recording these simple changes are more time-consuming.

2.5 Errors in translation are discussed and changes made with translators’ agreement

The recordings often highlighted errors in the actual translation. Most of the errors were grammatical issues with sentence connectors and verbal tense/aspect issues, especially for a certain class of verb forms in Apal. Sometimes, even though the translation had been checked by a consultant, there were errors or ambiguous pronouns still left in the translation. In the Mborena Kam session, these issues were discussed and worked through with the national translator who was present at the session and could make the changes on the computer. Before approving the changes, the consultant asked questions to confirm the accuracy of the changed text. In the Apal sessions, the leaders would suggest changes that were only for the sake of naturalness, but all suggestions were taken back to the main national translator who would decide what to do. If the leaders rejected a passage as not having a clear meaning, the national translator would revise it and it would be brought back to other checkers for a final comprehension check. We found that many of the difficult statements that had been translated literally, such as “let the dead bury their own dead” (Matt 8.22) and “You strain out a gnat but swallow a camel!” (Matt 23.24), had to be greatly modified. The leaders asked that the “meaning” of the passage be put in the translation. They could give an accurate back-translation, but when they listened to these passages, the passages just did not make sense.

2.6 Final revised, rerecorded texts are checked and approved by designated teams

In the Mborena Kam translation, all recordings were brought back to the main team of leaders for their final approval. Because the Apal leaders live in widely dispersed villages and were busy, it was decided that a different group of local people who were very detail-oriented would listen through the final rerecordings and approve or reject them. This team could check 200–400 sentences in a day, depending on how many times they had to listen to each sentence. By the end of the process, we had easily rerecorded more than 4,000 sentences that the leaders had rejected for one reason or another.

2.7 Distribution and testing

The Mborena Kam recording was initially distributed via recordings on SD cards at the dedication, but later the team had the opportunity to produce a better-quality recording that was done by Faith Comes by Hearing at a recording studio. This recording was distributed to the churches on devices that could play the sound files for a large crowd.

The Apal recordings were distributed as soon as the first recordings passed the first level of quality control. Each person recording the Scriptures received an Audibible that contained all the recordings completed to date. Whenever they brought their unit back to me, I would reload it with a newer, more complete copy of the recordings. In addition, the sound files were loaded into the Apal Scripture app, which has been available since 2015. The Audibibles were used, but as soon as thieves learned that the Audibibles contained SD cards, they were stolen. I thought that the MegaVoice units, which do not have SD cards, would solve that problem, but I have had those units brought back with all the wires and the little battery gone. Most of the units, however, come back because they have worn out.

The input from the listeners has had a direct impact on the translation in many ways, including key terms. For example, the term used for “crucify” in the noncentral dialect was changed earlier this year to a term that was more acceptable throughout the language group. They had read the word we had previously used since 2010 in checking sessions and in the first printed copies of Mark in their dialect, but it was not until leaders “heard” the recordings that they announced that the word had to be changed. These men from two different villages would never have sat in on a checking session, but their voices were heard because they are fluent speakers of the language who listened to the recordings.

3.1 Mborena Kam changes

In the process of checking the recording of Mark, 1,119 changes were made to the text. Some sentences had multiple changes in them. Of the 1,159 sentences in the translation of Mark, only 301 sentences were marked to be rerecorded. Most of those sentences (216) were rerecorded due to the fact that they corrected a grammatical error in the written text. Most of the changes to the written text (823), however, were spelling issues (Table 1). The men recording the text knew what the text was saying and pronounced it correctly, but the written text was misspelled. There were only eighty-five sentences that were rerecorded because they were read incorrectly (Table 2). There were also many cases where the text was changed to match the recording because the recording was viewed as more accurate or more natural than the written text.

OPEN IN VIEWER

Table 2.

Reasons for rerecording in Mborena Kam Mark.

Reason for rerecording	Number of sentences in Mark	Percentage of the 1,159 sentences in Mark
Rerecorded due to reading error	85	7%
Rerecorded due to changed text	216	19%
Total sentences rerecorded	301	26%

No translation is ever perfect, even one that has been consultant-checked, but we caught and corrected several errors that were classics. For instance, when the people listened, the translation for rising on the third day made it sound as if Jesus would rise on Monday rather than Sunday (Mark 9.31), and the face of the rich young ruler “swelled up in anger” at Jesus (Mark 10.22)—a common misunderstanding of the model translation, Tok Pisin. Some of what appeared to be exegetical errors were actually reading errors, such as the one that resulted in there being seven instead of twelve baskets (Mark 6.43). Overall, it was a sound translation that was improved through the prepublication recording process.

The initial Apal recordings were done by the national translator working in his own dialect, which is the noncentral dialect of Apal. It was an experiment to see if there was any benefit in prepublication recording. The text had gone through multiple comprehension-checking sessions, and it had been consultant-checked, but I knew that it still had issues due to the language variation among the villages that speak that dialect. As a result, there were still spelling issues, but most of the issues involved grammar, naturalness, and information load. There were also a significant number of changes that corrected minor issues in the translation. In the second and third rounds of the checking of the material in this dialect, the team finally decided that they needed to change the spelling of a number of reduplicated verbal forms to reflect their actual pronunciation.

Because a different process was used, all of the changes in this initial round of recording Titus and Jude were due to the text being changed. If the national translator had not tried to record the text, it is likely that the text would not have had as many changes during the initial recording process. As he tried to record the text as it was written, his tongue would say something different, and he would stop and ask if we could change the text. Of the 372 sentences in the Apal translation of Jude and Titus, 126 had to be rerecorded because they were revised in the recording process (Table 3).

Throughout the recording process as we cleared away the bigger issues in Apal, smaller issues came to the fore. When I began working in the area in 1985, the word for “fence” in the dialect I learned to speak was cima. But over the years, there had been a change and now the team all agreed that we needed to change all examples of cima to lua—all 134 occurrences. But thankfully, they identified that dialect shift before we had gone very far in the recording.

In the 1980s, anyone that was “characterized by money” was rich, but as the society shifted more towards a monetary culture, the term “many” was added to the term for rich. While checking the final book of the Apal Scriptures, the checkers said that “characterized by many (pieces of) money” just means he has a lot of coins. So, they said that we needed to change all seventy-one occurrences of “many (pieces of) money” to “big money,” checking to make sure it worked in the context, and then rerecord each of those sentences. That change took a significant amount of time and attention to detail. These are just a few of the many global changes that were made as the team grew in their ownership of the translation and their desire for it to be accurate and natural.

Generational differences in pronunciation of words were a source of stress to the older generation. They repeatedly told me, “Tell them to say simɨ [‘softly’] because it is not a SIM card.” They all knew that in fast speech the ɨ at the end of words was often swallowed, but the leaders were adamant that the young people doing the recording could and would clean up the pronunciations of a list of words that, while not totally wrong, were not the best. When I pointed out the issues to the recording volunteers, they all willingly rerecorded a large quantity of occurrences of words they had mispronounced. They had heard enough criticism from others that they wanted to clean up the problems. These problems included some terms that occurred repeatedly throughout the tabernacle accounts of Exodus, which, when said incorrectly, ended up being swear words.

4.1 The recordings themselves are recognized as a testimony to God’s work

An Apal literacy teacher was listening to his MegaVoice player and said, “I love to listen to Lutɨ and Hegalɨ. Neither one of them ever went to school to learn to read. They prayed and God taught them to read. When I listen to them, I am reminded of God’s power.”

In addition to the women, the children took up the challenge of recording the book of Luke, the first book of the Bible published in Apal, and the one that they had used in Sunday School and Bible studies. The youngest recording volunteer was about twelve when he started the process. His recordings are a patchwork quilt of mainly his little-boy’s high-pitched voice with some corrections done in his teenage voice, which is much lower. Despite that fact, no one has objected to this patchwork of voice quality. The leaders all say they enjoy listening to him read.

4.2 The children have listened to the recordings

I know that the recordings are being used because when I walk my dog in the mornings and evenings, I often hear people listening to Scriptures while they are cooking or doing other chores. These include the Apal speakers’ children who do not speak this endangered language. One day as I passed by a house, three children yelled alialaŋ (which means “OK”) about a second or so before alialaŋ came through the speaker of the Audibible. They had listened to it so many times that they knew when that word would be heard in the recording.

4.3 People have read and heard Scriptures that they have never heard in any language

During the recording process, Ana had stopped, and her head was down while she was recording the final chapters of Revelation. I waited a bit and then asked if there was a problem. She said, “No, I was just thanking God for his word.” I asked if she had ever heard or read this book in any language and she said, “No, all of it is new.” As one of the preachers told me, “All we do is preach about repenting and the fact that the end is near!” So the recordings have been a Bible survey course for many people. One of the preachers said that he takes his phone with him to the garden and whenever he rests, he pulls out his phone and listens to Scriptures or reads if there is no recording available. Many people put their MegaVoice players on and go to sleep listening to Scriptures. I am normally all for listening to Scriptures, but when I was teaching a Scripture course at a distant village and sharing a house with several teenage girls, I had to wake them up and tell them to turn off their player because I could not go to sleep listening to a recording that probably had a translation error in it.

4.4 Even translators benefit from the recorded Scriptures

Prepublication recording and editing greatly improves the quality of translations for oral preference cultures. As the main national translator for the Apal (the only one who really developed the ability to edit on paper) has been listening through the recordings, I find that he is asking a different set of questions. He now has a tablet with Paratext Lite and his questions written in Apal almost sound as if he is asking me comprehension-checking questions. He often asks, “Who said/did this?” and “What does this ‘this’ refer to?” He has also been much more sensitive to the information load. While just looking at the text on paper, he can easily figure it out, but when listening, especially if the reader did not pause in the correct place, the result can be confusing. He has marked those places and written in suggested revisions.