I wonder if the quality of speech coming from the cell phone has anything to do with the amount of processing required. When people can't hear things very well, they start piecing together the dropped parts of the conversation by using some sort of contextual implication. You know what the subject is, so you have a good chance of surmising the dropped words due to context.
I would think something similar could be possible for talk radio as well. I think if you listen to one talk show host consistently enough, you develop a better ability to understand what is being said, but a new talk show host can take some getting used to.
Just some thoughts.