Spreadsheets for critical applications, such as financial reporting, are widely created and used by many people with no expertise in programming or software development. It is well known, however, that creating spreadsheets is an error-prone process. Several methodologies have been designed to reduce these errors. In this thesis we characterise the patterns and functional relationships among the formula cells and the corresponding data cells that commonly occur in spreadsheets, and show how the patterns occurring in a given sheet can be generalised to produce a template structure representing the family of spreadsheets of which the given sheet is a member. Finally, we show how this generalisation can be translated into an L-sheets program from which instances of this family can be generated.