This is how the index works in simple terms.

When you add a row to a table the rowid of that row plus a few other info is store in a separate area which is dedicated to keep it. Lets say you define a unique index. When you insert a valid row the rowid + the key of the index is stored in the index region ( i forgot what it's called physically ). So a repository which says for this key so and so is the rowid is created. When you try to fetch out the row containing that key SQL engine looks into this region finds out the rowid and moves to that location.

Things work pretty much the same way for all the type of indexes. For a non unique index it will be a single key and a number of rowids. That means for this key(or group of keys) all these rowids are possible. SQL engine takes this info and goes to find out those particular rowids.