When you go into the field, and record a fieldwork session, you can either record the whole session, or you can record each utterance. For example, suppose that you are interested in subject pronouns, and you want to transcribe the full paradigm. One way to do this is to press the start button on your recording device at the beginning of the session, and the stop button at the end. That way, you will have recorded the whole session. The other way is to record each individual sentence and word that you elicit.
In this blog post, I will enumerate some advantages of each method. If you have two audio recording devices, you could do both, and in certain situations using two recording devices might be the optimal solution.
Recording by Utterance
You can use Praat at the time of the recording to make sure you have an accurate transcription (e.g., checking voicing of a stop consonant, or checking the pitch of a vowel).
You can play back the sound file several times (just after recording it) to help in transcribing, instead of making the consultant repeat an utterance over and over.
You can play back the sound file at the time of transcription to make sure the sound quality is acceptable (e.g., there is no background noise degrading the recording, the recorder is working properly, the consultant is speaking into the mic, the cables are all plugged in properly).
Since you label the sound files as you create them, there is no need to cut up the long sound file for a session and label all the utterances of the session. Hence, recording by utterance can save hours of time in organizing your data.
Recording by Session
Sometimes consultants will say interesting things during a session that you do not prompt them to say. These unprompted utterances can be quite revealing and useful. When you record the whole session, you are certain to not miss any new and interesting data that comes up.
The consultants may use their language to speak to each other during the session. In effect, by recording the whole session, you are recording spontaneous oral texts that you can transcribe later.
By recording the session, you record all the context leading up to the utterance. What did the linguist say to prompt the utterance? What kinds of questions and remarks did the consultant have about the task? What kinds of information did onlookers contribute? This kind of contextual information can be quite useful when analyzing data.
If you have not worked with a particular consultant before, recording the whole session may be the most unobtrusive way of getting a recording. You just turn on the recording device, and forget about it. For a beginning consultant, they may be more comfortable and less self-conscious this way.