Theoretically, if you retain the informational dimensions of the image (say 1000 by 1500 pixels) and change the dots per inch density, nothing else changes except the final physical measures. Divide 1000 (pixels) with 100 (pixels per inch) and you get 10 inches. Dividing 1000 with 250 you get 4 inches. This will affect the printout. Now, if you change the dots per inch density and want to retain the physical measures of 10 inches you'll be short of pixels. Those must be interpolated. While the software doesn't know the real-life situation, it will simply multiply what is existing (instead of 1 muddy brown pixel you'll get 4 or 6 or 12 exactly similar ones, in a big uniform cluster). You get more pixels, but in fact no more information Using dedicated software, e.g. some fractals-based stuff, it's possible to cheat the Maker of the Universe and create some really newborn pixels, based on programs or programmers considerations. Simple, eh? Peeter who is not very sure, but has experienced some pixels personally