Efficient use of intelligent information systems requires tools that can assist not only in finding information that can be deduced from the databases, but also in inferring missing information using similarity and statistical based measures adapted from artificial intelligence. Inductive dependencies provide an expressive way to achieve this goal. Relationships that are not functional can be written precisely, and a delta measurement makes explicit the reliability of the relationship. Using inductive dependencies and appropriate delta functions one can express a wide spectrum of relationships and manipulate them with some degree of certainty. We discuss the utility of inductive dependency as a knowledge mining formalism that can be used in discovery in databases and study properties of delta functions that capture the functionality of relationships.