Let me try to sort this out a bit. When you scan an original, you wind up with a file that has certain pixel dimensions. That's an absolute number of pixels vertically and horizontall, NOT pixel "per" anything.
Think of it like this: suppose you're redoing your bathroom and you want the tiles to form an image of your school mascot. You sketch out the mascot, divide the image into squares, and decide that the least number of squares that will represent the mascot is, let's say, 40 x 40 tiles.
Now you give that sketch to your tile guy. He knows he's going to need to buy 40x40 or 160 tiles. (He also knows he's going to make a mint off this job!)
What you DON'T know yet is how big the tiles are going to be. If you decide to use little tiny quarter-inch tile chips, your design will come out taking up 10 inches by 10 inches on the floor. If you use 12-inch tiles, your design will come out taking up 40 by 40 FEET on the floor.
Notice that both designs will contain the same amount of fine detail -- there's only so much you can represent when you divide the mascot into a grid of 40 x 40 tiles.
So you can see it's the count of tiles (or pixels) that will determine how much detail is in the image; in this case, it's 40 x 40, regardless of size. The size of the tiles... quarter-inch, one inch, twelve inches, whatever -- will determine how big an image you get from a given pixel count.
So how does that apply to film scanners? My old Canon 4000 35mm scanner is rated at 4000 pixels across the height of a 35mm frame. In other words, when the film goes through, it gets divided up into 4000 little slices vertically. The length of the frame is about 1-1/2 times the height, so (assuming the scanner uses square pixels) it'll also slice it into about 6000 pixels along the film's length.
So, scanning a 35mm negative with this scanner will yield a file with dimensions of 4000 x 6000 pixels, period. We haven't specified "per inch" or per anything, because all this number tells us is how many "tiles" make up the image. It's later on, when we decide what size "tiles" to print, that we'll determine how big the final print will be.
It's at the printing stage that the "per inch" numbers get important. Suppose I'm going to print on one of the Epson Stylus Photo printers. These have impressively high-sounding numbers for how many ink droplets per inch they lay down. But since the printer has to combine several droplets to make a patch of color, the actual number of patches per inch is much lower. I've found through experience with my Epson printers that there's not much point in assuming you'll get more than 240 or 250 ink patches per inch. You can set the machine to print a higher number of patches per inch, but it won't translate into extra detail.
(These ink patches are sometimes called "printer pixels," but more commonly they're just called "dots" -- which is where you get specs such as 240 dots per inch or 300 dots per inch.)
So, I've got my 4000-pixel-high scan file, and I'm going to print it on my Epson printer that yields, say, 250 ink dots per inch. That means that each 250-pixel band of my scan file will lay down a band of printer dots one inch wide. Since I've got 250/4000 or 16 such bands, it means my 4000-pixel-high scan file will print at 16 inches high at a printer resolution of 250 dots per inch.
If I want to make a bigger print, I've still got only 4000 pixels to start with. So I have to tell my printer to make larger dots. That's no problem, except eventually the dots get so big that if you look closely at the print, they're visible to the naked eye. When that happens, people say the image is "pixellated" or "has got the jaggies," and for a print you're examining at a close distance, it's going to happen when you tell the printer to print about 100 ink dots per inch, or less. That 100-dot setting will make a print that's 40 inches tall -- 4000/100 -- but it won't stand up to close examination because the viewer will see dots instead of smooth tones.
Of course, if the print isn't going to be examined at a close distance -- and many huge prints are examined only from a distance -- then you can get away with much lower printer dots-per-inch numbers and make impressively large prints from your same 4000-pixel-tall image.
Does that make it thoroughly clear? Or are you thoroughly confused now?
