Not a terribly helpful answer. I'm not sure I 100% understand it, but I'll give you what I think I know.
Pools assign work units, similar to the current block being worked on, however at difficulty 1 (or higher depending on your hashing power), which are called shares. The idea is that you solve your mini-block, and submit it back to the pool, to show your proof of work, out of every X shares submitted, one will also solve the "real" block being worked on (by chance).
So you are in essence solving an easier problem, that can potentially solve a more difficult one, and everyone doing this over and over and over eventually do find a block.
Sounds reasonable, thanks. I'm actually looking for some articles that go into more detail. It looks like it really depends on how the pool operator decides to do it so it seems to vary from pool to pool.