We present a framework for reasoning about secrecy in a class of Diffie-Hellman protocols. The technique, which shares a conceptual origin with the idea of a rank function, uses the notion of a message-template to determine whether a given value is generable by an intruder in a protocol model. Traditionally, the rich algebraic structure of Diffie-Hellman messages has made it difficult to reason about such protocols using formal, rather than complexity-theoretic, techniques. We describe the approach in the context of the MTI A(0) protocol, and derive the conditions under which this protocol can be considered secure.