Pixel Interpolation
Take a close look at the following two images:
The bottom on is pixellated, you can see individual pixels giving curves a stair-step pattern. It is most notable around the eyes and on the man's hairline. Despite the fact that the image on top looks better than the one on bottom, they both have the same amount of detail captured from the original photo, since both are from the same scan. Both images are cropped and resized to the same size from the same starting scan, only in the top image, I used a good pixel interpolation algorithm.
Let's go back to the ridiculously pixellated example of the cabinet card:
As you can see, the pixels are presented as a series of blocks. So diagonal or curved lines will get represented as a series of stair-step patterns (or saw-tooth patterns, depending on which imagery you prefer). Pixel interpolation algorithms attempt to smooth out this pixellation patterns by filling in partial pixel areas. Take a look at the photo done with bicubic spline interpolation, but still at 5 ppi:
You can see that the sharp edges are smoothed out to form curves, and there is greater variation in shades. This has the same information as the pixellated photo above, it is just that the photo editor smoothed out the sharp edges. The same thing has happened in the photo at the top of this page.
Without the interpolation, the crop from the 300 ppi scan would have been unacceptable. With the interpolation, however, it is much better. In fact, the higher resolution scans for the most part are not adding additional detail, they simply act to smooth the edges on enlargement like interpolation does. So why not save your file size, and let the software do the smoothing for you.
Some interpolation algorithms are better than others. The bicubic technique is pretty good. The Lanczos filter, which I used for making the print, is one of the best. Picture Window and Photoshop will handle either technique as well as other techniques. Most other high-end editors will give you at least decent if not excellent options. Q-image, an inexpensive printing program, has about the best interpolation mechanisms I have seen, although Genuine Fractals, a Photoshop plug-in is the industry standard. Links to both programs are available on the links page.
Note: at the risk of repeating myself, interpolation does not add detail to the image, it just smoothes the ragged edges due to pixellation. Therefore, if you don't have the detail in the original scan, the best interpolation algorithm in the world will still not add in the necessary details.