Basically NTSC standard works by showing 29.97 frames every second. FILM however is only 24 frames per second. So first they slow down to 23.976 frames per second, which is 4/5 of 29.97. Next step is making it so that every 4 frames the have an extra 5th. To make this as smooth as possible they do this: they pick the first 2 frames and leave them alone. Then, they divide the 3rd frame of the original stream into odd and even fields, which are alternate lines of the original entire frame. They pick the bottom lines and put them in a frame along with the top lines of the second frame, and this is the 3rd frame of the telecined stream. Next the 4th frame of the telecined stream, which is the top lines of the 3rd original frame along with the bottom lines of the 4th original frame. Then they put the 4th frame of the original stream as a unique 5th frame.
What you get is basically this, as wikipedia shows pretty well:
http://en.wikipedia.org/wiki/File:32pulldown.svgUpon viewing on an interlaced screen, you won't see the lines and the motion will be rather smooth (although depending on a scene a jarring motion during the telecined frames might be noticeable - still less jarring than just doubling a frame).
Keep in mind that the order of the field matching might be different than the one explained above, and it might change multiple times in a single stream.
Of course, what you do when IVTCing is dropping the bottom fields of the 3rd telecined frame and the top fields of the 4th telecined frame, and reconstruct the original 3rd frame with the other field, having a 23.976 stream in output. Depending on your source you could even go the extra step of speeding back up to 24fps, but I believe that on digital sources as anime it's already made in 23.976 to begin with, at least in recent times (proof: if you try comparing the opening song that you could find in an OST CD vs the opening animation speeded up to 24fps, the op sequence will sound slightly faster compared to the actual opening song in the CD, although there likely won't be any noticeable difference in the pitch).