This study is an analysis of the functional and adaptive significance of "caw" calls used by common crows (Corvus brachyrhynchos) in contexts involving danger. Seven tame birds, including three siblings and several birds familiar to the siblings (i.e., within their sight and hearing), were the subjects of playback experiments. Four types of caws associated with danger were played back as test calls: screams, mixed caws, inflected alarm caws, and alert caws. These four test caws were recorded from each of six different individuals (senders) and played back to the experimental subjects (receivers). The vocalizations of the three sibling receivers were tape recorded during playback trials, and three types of response caws were scored: mixed caws, alert caws, and long caws. By counting caws in each 10 s interval during the 1 min before, 20 s during, and 20 s after playback, the following functional interrelationships among cawtypes were found. Mixed caw responses were elicited immediately by, and only by, screams and mixed caws. These caws are used in harassing a predator, and seem to function in part to assemble and coordinate a mobbing group. Alert caw responses were suppressed over the 20 s during which caws were played back for all types of presentations except those of alert caws, but reached a peak just after playback. Alert caws seem to be multifunctional, probably indicating either mildly threatening objects or cessation of danger. Long caw responses were suppressed during the 20 s presentations of all cawtypes, but reached a peak after scream and mixed caw playbacks. Long caws seem to indicate either cessation of danger or continuation of normal activities. By further subdividing numbers of responses on the basis of social familiarity between senders and receivers, the effect of social relationship on responses was examined. The senders were either siblings of, familiar to (heard and seen only from a distance), or unknown to the receivers. The receivers did not respond more to the voices of senders from any particular social category (G goodness-of-fit tests) although the data were not conclusive. The results are discussed in terms of the information encoded in mobbing and alarm calls, the functions of those calls, and the sources of natural selection which may have shaped the evolution of mobbing and alarm vocalizations.