Many extensions to the basic STRIPS action representation have been proposed. Many of them amount to the use of filter conditions, preconditions which do not become subgoals. The purposes to which f'dter conditions have been put and the justifications for their use have seldom been made clear. In this paper we clarify and examine the various arguments that appear to lie behind their use. We conclude that there are two major flaws in their use: their interpretation is inconsistent with commonly used planning techniques, and their use often leads to incompleteness. Moreover, their use generally addresses special cases of more general problems, and is not extensible to the general case.