Since Simmel's seminal essay on secret societies, little systematic work on the topic has been attempted even though the empirical basis for such work has greatly improved. The present paper uses six case studies with unusually good structural detail in analyzing one aspect of secret societies: their organizational structure. In contrast with Simmel, I argue that secret societies under risk must be differentiated from others; that societies under risk have a wide range of structural forms; and that the major sources of structure are found in preexisting social structures rather than in psychological factors. Risk enforces recruitment along lines of trust, thus through preexisting networks of relationships, which set the limits of the secret society's structure. Structure can still vary considerably, depending primarily on the centralization of control of recruitment, in turn dependent on the control of key resources.