Many species produce alarm calls that vary according to situation. Theoretically, alarm call structure could covary with predator type and could communicate potentially ''referential information, or calls could covary with the degree of risk a caller experienced when it emitted a call. Using similar methods, I studied the ways in which Olympic (Marmota olympus), hoary (M. caligata), and Vancouver Island marmots (M. vancouverensis) communicated situational variation. I observed both natural alarm calling, and I artificially elicited alarm calls with simulated terrestrial and aerial predators. I used playback experiments to study marmots' responses to different alarm call variants. All three species produced four roughly similar but distinctive loud alarm vocalizations that could be categorized by their relative shape, duration, and whether calls were quickly repeated to create multi-note vocalizations. In addition, the Vancouver Island marmot produced a fifth loud alarm call-the kee-aw. Call micro-structure varied as a function of the distance the caller was from an alarming stimulus and the type of alarming stimulus. Two lines of evidence suggest that all three species had alarm calls associated with the caller's risk (i.e. they were not referential). First, marmots often changed call types within a calling bout: there were no unique stimulus-class specific vocalizations. Second, marmot responses to alarm calls were graded: marmots did not have unique responses to different call types. These three close taxonomic relatives with superficially similar calls, communicated risk differently.