I'm guessing you mean to say denormalized table. An example would be for phone number. Any person can have a home, cell, work, and fax numbers. Given that the is a one to many relationship between person and phone number, you could store phone number as a separate table from person with a primary key consisting of person id, and phone number type. But doing so will require a join to bring back person and phone number, so instead you can denormalize the table by making 4 extra columns in your person table. One for each type of number that you track.