in a normal table with his correspondong indexes the data and indexes are seperated. with index organized table the data is actually stored *inside* index, in a normal b tree index what it contains it's the index and the rowid which points the location of data, so basically oracle does more I/O with normal b tree indexes. With indexed organized table instead of storing index and the rowid the btree index actually stores the index AND the data so in one I/O Oracle can retrieve both index and data therefore reducing I/O, in some way the way this work is similar to a cluster

MS SQL Server 2000 has a similar (or same) feature called clustered indexes which microsoft claims with this feature they beat TPC benchmark, if you are familiar with SQL Server 2000 may be this might help you a bit to understand indexed organized table, also the main page of dbasupport there is an article about indexed organized tables