It is often desirable to represent entities in a database whose properties cannot be deterministically classified. We develop a new data model that includes probabilities or confidences associated with the values of the attributes. Thus we can think of the attributes as random variables with probability distributions dependent on the entity the tuple purportedly describes. We study two sets of issues, one dealing with the proper model for probabilistic data and the other dealing with the choice of operators and language necessary to manipulate such data.