Many of our clients have this annoying propensity to request we have a narrator read on screen text. I am trying to build the case for why this is not a good approach and need some suggestions for how to present the case to clients in a kind, gentle, persuasive way.

My basic premise is that adults do not need to be read to. Additionally I think it hinders comprehension/cognitive processing.

Any compelling thoughts out there?

That is annoying.

I've always been taught that people are more likely to read the screen and not listen to the narration because people read far faster. The audio becomes just background noise inhibiting learning. So if no one is listening, why spend precious resources on creating all that audio? That time, effort, and money could be used to add something that actually enhances learning.

Check out Chapter 6 in Clark, R.C. & Mayer, R.E. (2003). E-learning and the Science of Instruction.

This chapter explains the Redundancy Principle, which essentially says that presenting words in both text and audio narration can hurt learning.
More specifically, avoid narrating onscreen text when words and pictures are presented simultaneously at a fast pace. Consider narrating onscreen text when there are no pictures, the learner has ample time to process the pictures and words and/or the learner is likely to have difficulty processing spoken words.

I am very familiar with Chpt. 6 of Clark and Mayer's book, but it seems they talk about graphics text and audio and not so much about text and audio. I will look again at that chapter to make sure I am not missing some good research on my issue.

I agree with everything that's been said here, but I also just ran into an issue that may be the exception.

We have a lot of people with low literacy that are required to take some online courses. In those cases, we've found that it is undeniably helpful to provide an optional audio narration (the key there being "optional"). I had a report of someone taking an hour to get through 2 or 3 small screens of text without the audio. It should take no longer than 3 minutes with audio.