I wish I could find a better resoure. I will have another look in my own time after work.
http://www.digicaminfo.btinternet.co.uk/antishake.htm
I will be interesting to see your you photographs
EDIT: I have just found out that it is dependant on Focal length but not subject distance, I will have a read tonight!
Stuart
Well, this gets interesting, as it looks like I may well have blown my theory right out of the water, at least partially anyway.
I had a look through my old shots with 450D and IS kit lens, and had ones going right down to 1/10 or so, mostly at 18mm, but some at 35mm or more. They were sharp... well, as sharp as a kit lens could be, but I am still fairly certain there wasn't any motion blur.
I tried some tests with the 1Ds and 17-40 in a room lit simply by a lamp, and was able to expose it correctly at f/4, iso 800, 1/4. I took two shots at these settings and to my amazement, the pictures came out surprisingly sharp. At 100%, there is evidence of a tiny sliver of camera movement, but still the pictures were sharper than the kit lens. I upped the shutter speed to 1/8 and maxxed the ISO, and I noticed a sharpness increase.
I really can't be bothered uploading samples... so you'll just have to take my word for it... what can I say, I have steady hands
My conclusion is that the IS on my previous camera/lens combo was doing its job, especially when you consider the crop factor which can make shake even more apparent. (1/10 @ 35 ish mm x 1.6 is pretty good considering) It does indeed seem, as you suggested, that less of a difference is observed at these very short focal lengths, but that could just be because a shorter focal length appears to minimize shake anyway, and my test isn't exactly fair because my current camera is full frame and has huge juicy pixels compared to the last.
I will however maintain that I think IS on short focal length lenses, as well as longer lenses, can be incredibly useful for shooting still subjects, and I'd still quite like it if my 17-40 was equipped with IS
