http://www.amazon.com/gp/product/073520019X
After a lot of practice I ultimately came to the conclusion that it doesn't really work. I could 'read' the pages really fast, but it's not what I'd call reading. It's skimming. Sure, the speed of fixation and regressions limit your theoretical maximum reading speed, but — for me at least — the real bottleneck was how fast I could understand what I was reading. You may be able to look at 10 pages of words per minute, but understanding 10 pages of material per minute just doesn't really work.
Another thing I learned is that there's no good metric for what constitutes 'retention'. It's usually measured by making a list of how many 'things' you can remember from what you read. It's arguable whether remembering a list of words from what you've read relates in any meaningful way to whether you comprehended what you read. Having the word 'scientific' in the title of the article is misleading at best.