Why do animals make false alarms; are false alarms truly deceptive; and if, just as in Aesop's fable 'The boy who cried wolf' , animals can learn to ignore false alarms, why doesn't deception break down? I investigated these questions in a population of habituated and individually recognisable fork-tailed drongos (Dicrurus adsimdis), in the South African Kalahari Desert. Drongos either self-foraged, when they hawked and gleaned small insects, or followed other species stealing large terrestrial prey that hosts excavated. Stealing food from other species enabled drongos to capture prey otherwise unavailable to them and accounted for over 20% of their biomass intake. This was of greatest benefit during the morning and on colder days when payoffs from stealing remained stable while those from self-foraging declined (Chapter 3). Drongos stole food using two strategies, either by physical attack or by producing false alarm calls in response to which hosts fled to cover, enabling drongos to fly down and collect abandoned food. False alarms increased overall success, and were produced when stealing small food items unprofitable to gain by physical attack, or when stealing from larger species more likely to defend food (Chapter 4). Drongos produced both their own alarm calls and mimicked alarm calls of other species in their false alarm repertoire. Playback experiments on two host species, pied babblers (Turdoides bicolor) and meerkats (Suricata suricatta), confirmed that these false alarm calls were deceptive because they were as effective at alerting hosts as true alarms given to approaching predators (Chapter 5). Further playbacks showed that hosts were more likely to be deceived by mimicked false alarm calls, including mimicry of the host species alarms, than by a drongo' s own alarms. In addition, host species habituated to repeated playback of the same alarm but resumed their response when a new alarm call was played, and drongos naturally changed their alarm calls when hosts failed to respond to a previous false alarm. Therefore, by employing vocal mimicry to vary their alarm calls drongos were both more likely to deceive hosts, and to maintain deception. Consequently drongos evaded the frequency dependent constraints which typically limit payoffs from deception when species produce only one signal (Chapter 6). The drongos sophisticated communication strategy raises questions for future research regarding the mechanisms leading to the production of complex behaviour.