Ever since Codd introduced so-called “null values” to the relational model, there have been debates about exactly what they mean and their proper handling in relational databases. In this paper I examine the meaning of tuples and relations containing “null values”. For the type of “null value” representing unknown data, I propose an interpretation and a solution that is more rigorously defined than the SQL NULL or other similar solutions, and which can be implemented in a systematic and application-independent manner in database management systems.