Referential integrity's a bit of an involved subject to get into on a bulletin board, but primarily it's there so that only valid data can be entered into the database.
Lets say you're working with a chain of stores, so you have a table called stores, with intStoreId as the Primary Key. Each store has it's own id.
Now you have an inventory table that all the stores access, so if a customer comes in and says "Do you have Product X", the salesperson can go the computer, look it up and say "No, I'm sorry, but Store Y has one."
Your inventory table may contain the Product ID (presumably from a product table), the Quantity, and the Store ID. The Product ID and the Store ID combine to make a joint key.
This all comes back to referential integrity when you're adding new information into the inventory table. By enforcing the foreign keys (in the inventory table the Product ID and Store ID are both foreign keys) the database management software (in your case Access) ensure that only products that already exist in the product table are added, and they're only listed against stores that already exist.
So the referential integrity makes it impossible to add products that don't exist, or (and in an inventory scenario probably a worse situation) adding a product to the inventory of a store that doesn't exist.
Does that make things any clearer, or did I just make things even more confusing?