Yellow-bellied marmots, Marmota flaviventris, were reported to produce qualitatively different alarm calls in response to different predators. To test this claim rigorously, yellow-bellied marmot alarm communication was studied at two study sites in Colorado and at-one site in Utah. Natural alarm calls were observed and alarm calls were artificially elicited with trained dogs, a model badger, a radiocontrolled glider and by walking towards marmots. Marmots 'whistled', 'chucked' and 'trilled' in response to alarming stimuli. There was no evidence that either of the three call types, or the acoustic structure of whistles, the most common alarm call, uniquely covaried with predator type. Marmots primarily varied the rate, and potentially a few frequency characteristics, as a function of the risk the caller experienced. Playback experiments were conducted to determine the effects of call type (chucks versus whistles), whistle rate and whistle volume on marmot responsiveness. Playback results suggested that variation in whistle number-rate could communicate variation in risk. No evidence was found of intraspecific variation in the mechanism used to communicate risk: marmots at all study sites produced the same vocalizations and appeared to vary call rate as a function of risk. There was significant individual variation in call structure, but acoustic parameters that were individually variable were not used to communicate variation in risk. (C) 1997 The Association for the Study of Animal Behaviour.
