I apologize to Steven for not making myself clear when replying to his paper--his rebuttal of my comments shows that he missed the point altogether and that has to be my fault. (Old trainer's adage--if the learner doesn't understand, you haven't explained it properly.)
Here's a scenario which may help. I run Steven's experiment with a group of known poor readers. Just to be thorough, I run a second trial with a group of visually impaired subjects. To be even more thorough, I get hold of a group of newly arrived immigrants from Slovenia who have only been learning English for two weeks and, through a translator, get them to agree to take part.
The groups are presented with a pencil and paper post-test and can't read well enough to answer the questions. I scrub that test and repeat it using audio questions and write down the verbal answers from the students. Apart from the new immigrants who seemed to perform poorly all round, the groups perform better when learning from audio input. I proudly conclude that "media matter." Steven, what's wrong with my conclusion? Wouldn't you agree that the skewed abilities of the sample groups are more significant than the media?
And please don't say "Clark was no better." Your experiments must stand in their own right to be worthy of serious consideration. I suggest that controlling extraneous variables is at the heart of any worthwhile research and I don't think you did that. Will you not concede that your experiment was flawed and that your conclusion is unlikely to have any validity?