It's not that meaningful to assign a single number, true. I gave my speed for "normal" text--comments, blogs, newspaper articles, "light" books (business/best-sellers), fiction (if I have to/feel like reading it quickly). When I read scientific papers, the speed drops considerably until I am used to the terms used in the field.

Thanks a lot for that comment though, I have less incentive to try training it further now... I am pretty surprised that anything could decrease significantly from trying to train it though. I would suspect other effects at work (like now she is reading a different kind of text, or had previously never measured herself, or the training was nonsense, etc). Any idea what caused the decrease?