Well I think he means that you need an image big enough so that when you turn the pixels into rasters, you can actually tell what the picture is. The bigger you go, the harder it is to tell what the hell is going on and it just looks like a bunch of miscolored blobs on paper.
