During the last decade there has been a lively debate about the degree to which sociobiological theory is untestable due to the highly speculative and ad hoc nature of sociobiological explanations and, usually by inference from the explanatory cases, predictions — especially those concerning human behavior. In this paper, I examine this problem within the framework of the semantic conception of theories. I shall argue that a semantic conception of theories provides a more thoroughgoing analysis of this problem than the more familiar syntactical conception of theories. In particular, a semantic conception more accurately represents the relation between a theory and phenomena which, in this case, results in a clearer understanding of: the nature of the problem of the testability of sociobiological theory, the difference in explanatory success of the theory when applied to insects and to humans, and the kind of additional knowledge required in order to apply the theory successfully to human behavior.