This paper examines the issue of developing semantics for agent communication languages. In particular, it considers the problem of giving a verifiable semantics for such languages -- a semantics where conformance (or otherwise) to the semantics could be determined by an independent observer. These problems are precisely defined in an abstract formal framework. Using this framework, a number of example agent communication frameworks are examined with respect to the problem of verifying conformance. A discussion is then presented, of the various options open to designers of agent communication languages, with respect the problem of verifying conformance.