Efficient activation of rules is a fundamental issue in active database systems; choosing the suitable rule activation technique is therefore an important task. We have developed a technique, called join pattern indexing, to support incremental update of rule-derived data. In this paper, we compare join pattern indexing with discrimination networks (Rete and TREAT) for data-derivation rules. A performance study based on a stochastic model indicates that join pattern indexing is more efficient than discrimination networks in many cases. © 1996 Kluwer Academic Publishers,.