The development of large-scale cognitive models introduces significant computational challenges. Large declarative memories are a case in point. It is not computationally feasible to load a large declarative memory into the process space available for execution of a cognitive model. Fortunately, computer science provides us with relational databases to support access to large external stores of information from within an executing process. This paper motivates and describes the interfacing of the ACT-R cognitive architecture with a relational database to support large declarative memories within ACT-R models.