A relational database can be thought of as a logically structured collection of normalized relations (tables) that contain attributes (columns) with some number of tuples (rows) that contain data. In terms of a Microsoft Excel workbook, each worksheet is a relation, the columns that contain data in each worksheet are attributes, and each row that contains data in each worksheet is a tuple.
If designed properly, each relation should be normalized such that, at minimum as a best practice, the data is in third normal form (3NF), and each relation should have a unique identifier (primary key) that identifies each unique tuple in the relation such that any relation that shares that unique identifier can be related to the other relations that also share that unique identifier.
For more information, please see other technical explanation pages under this topic as well as the Bookstore Database Design Project.
Daniel M. Packer 