In this video we discuss normalization at the level of "Second Normal Form". When we moved to First Normal Form by flattening the file to eliminate an multi-valued fields, we typically introduced some redundancy. If we resist (for now) the temptation to introduce an overall primary key for the table, we might have a composite primary key -- a combination of fields which as a team uniquely identify each row of the table. Field B is "functionally dependent" on Field A if knowing Field A means that we can determine (maybe just by looking something up) Field B. To arrive at Second Normal Form, we split up a table into multiple tables by taking at properties (fields) that depend on only a part of the composite key into a separate table. It must remain possible that a JOIN query would reproduce the original table.