• Free translation. A speaker and I make a first pass through each recording, with the speaker providing a free translation of about a sentence at a time. I write down everything the speaker says, sometimes unable to distinguish her translations from her comments and explanations. I don't record these sessions (but they could be recorded). However, the original free translation is always preserved as a document separate from the interlinear glossed text.
• Easy transcription. On my own, I transcribe the original text and translate each word, assisted by the speaker's free translation. Inevitably, there are places in a recorded text that I cannot transcribe or translate on my own. In a 30 minute recording of Sahaptin, there might be about 50 such places, depending on the quality of the recording. I mark these for later review with a speaker. The better the linguist knows the language (as a second language learner), the quicker this task goes.
• Difficult transcription. I play the marked passages in the sound file for a speaker, who repeats what the speaker on the tape said. If needed, the speaker also retranslates the passage.
• Formatting and first proofreading. I apply LaTeX document formatting (covington style package) to the text to line up text and word glosses. Remarks by the speaker in step (1) that turn out to be cultural or situational comments are moved to footnotes. After the formatting is complete, I listen to the entire recording again and make any corrections I can hear. I print two copies of the document.
• IGT proofreading. A speaker and I listen to the entire recording again with the hard copy in front of each of us. Usually on every page we find a couple of missed syllables and occasionally sentences that might not be punctuated correctly (this will vary according to word order and intonational cues).
• Lexical checking. For Sahaptin texts, because we have an (incomplete) dictionary, I create a recording list of the words in the texts that are not in Beavert and Hargus 2010. We record Virginia's pronunciation of these words, and depending on their lexical category, we also record a sentence illustration (sometimes based on the text sentence). This step helps verify the translations and transcriptions of new words in texts.
• IGT distribution. I make corrections to the text, print it out again, and give the speaker who told the text (or their nearest relative) a copy.