A short while ago, ATD (the Association for Talent Development, formerly ASTD) posted an article by Jennie Ruby, titled Voiceover Scripts that Engage Learners. Jennie's main point is that most voiceover scripts, what we call narration, for eLearning is too long, too boring, and it turns off the student. Our attention span is short and most eLearning pages consist of a static screen. The solution is to factor both proportionality and conversational tone into the development of voice-over narration:
Basically, don't spend a lot of time voicing over a static screen. She suggests the amount of time spent on the screen may differ based on the content. A software demonstration should not linger on a static screen for more than maybe seven seconds, but compliance training can have more extensive voiceover descriptions even if the learner doesn't like it.
Consider aiming to have something move on the screen, illustrate the point, or change in some way about every two sentences. You can achieve the right proportions by either adding visual elements or shortening the voiceover script.
The narration should be short using plain-language and active voice. This is where I think we run into trouble relying on subject matter experts. They know their topic well, but are accustomed to talking with like-minded, experienced co-workers. We can try to write "plainly" but when they review it, they edit the narration as if they are writing to an experienced co-worker. We have to learn to be willing to pushback to get them to make the writing more conversational. I recommend reading the whole thing.