Little academic attention has been accorded to terrorism hoaxers—i.e. those perpetrators who use lies, benign materials and/or empty threats to give the impression that a terrorist act is or has been underway. This dissertation harnesses under-utilised terrorism events data to build a theory of hoaxes in pursuit of a dual aim: to provide a robust substantive answer to the empirical puzzle of why hoaxes are used, but not by all groups, and not all the time; and to evaluate the degree to which existing data can demystify the hoax phenomenon. The starting point is a rationalist framework for terrorist groups' strategic logics, which emphasizes the relative costs and benefits of hoaxes in relation to serious terrorism activity. In the empirical theory-building chapters, probit regression and qualitative comparative analysis (QCA) are used to identify various organizational conditions that differentiate hoaxers from non-hoaxers, thereby indicating which strategic logics are plausibly at play, and in which contexts. A statistical cluster analysis demonstrates that there are five broad classes of hoaxing terrorist groups, which differ from one another along motivational, structural, and campaign contextual lines. While the unit of analysis throughout is the terrorist group, these analyses rely on cross-national terrorism events databases—predominantly ITERATE and the Monterey WMD Terrorism Database—to identify which groups never hoax, and which groups sometimes do. In the dissertation's final section, earlier findings are tested against a new sample of terrorism perpetrators derived from the recently-released Canadian Incident Database (CIDB). Although the Canada-centric data reveals a biased under-reporting of hoax activity in the cross-national datasets, a QCA analysis of its perpetrators reveals roughly similar conditions differentiating hoaxers from non-hoaxers. The CIDB's comprehensive events coverage is further exploited to test whether these organizational indicators and their associated hypothesized mechanisms hold, when campaign activities are evaluated at the event-level. A fine-grained analysis of event sequencing in Canada's most prolific terrorism campaign (that of the Front de libération du Québec) corroborates a range of proposed strategic logics. The observational nature of available data is thus limited in its ability to clarify hoaxers' strategic logics, which are both over-determined and equifinal.