A new method to link records across datafiles and to detect duplicate records within files leverages a Bayesian approach incorporating prior information about the data collection processes of the datafiles in a flexible manner.