Whilst computer-mediated communication (CMC) can benefit users by providing quick and easy communication between those separated by time and space, it can also provide varying degrees of anonymity that may encourage a sense of impunity and freedom from being held accountable for inappropriate online behaviour. As such, CMC is a fertile ground for studying impoliteness, whether it occurs in response to perceived threat (flaming), or as an end in its own right (trolling). Currently, first and second-order definitions of terms such as im/politeness (Brown and Levinson, Politeness: Some universals in language use, Cambridge University Press, 1987; Bousfield, Impoliteness in interaction, John Benjamins, 2008; Culpeper, Reflections on impoliteness, relational work and power, Mouton de Gruyter, 2008; Terkourafi, Towards a unified theory of politeness, impoliteness, and rudeness, Mouton de Gruyter, 2008), in-civility (Lakoff, Civility and its discontents: Or, getting in your face, John Benjamins, 2005), rudeness (Beebe, Polite fictions: Instrumental rudeness as pragmatic competence, Georgetown University Press, 1995, Kienpointner, Functions of Language 4: 251–287, 1997, Kienpointner, Journal of Politeness Research 4: 243–265, 2008), and etiquette (Coulmas, Linguistic etiquette in Japanese society, Mouton de Gruyter, 1992), are subject to much discussion and debate, yet the CMC phenomenon of trolling is not adequately captured by any of these terms. Following Bousfield (in press), Culpeper (Impoliteness: Using language to cause offence, Cambridge University Press, 2010) and others, this paper suggests that a definition of trolling should be informed first and foremost by user discussions. Taking examples from a 172-million-word, asynchronous CMC corpus, four interrelated conditions of aggression, deception, disruption, and success are discussed. Finally, a working definition of trolling is presented.