The attack behavior of captive Canada geese (Branta canadensis canadensis) in winter was studied. An analysis of responses deals mainly with the head and neck units, and utilizes evidence from several independent sources. The general nature of agonistic behavior under certain experimental conditions is described. A wild-reared goose from one flock when introduced to another integrated flock generally approaches this flock (approach may become extinguished rapidly if this animal is consistently attacked on each introduction). The integrated flock reacts characteristically to this approach, and obviously at a certain distance can discriminate between a member of its own group and an introduced stranger. As the spacial arrangement of the flock birds changes a particular group of males assembles, and from this group usually one male attacks the non-group goose. Attack behavior was arbitrarily divided into the following stages depending on the presence or absence of locomotion: approach, pre-attack, locomotory attack and post-attack. Each phase was characterized by a particular neck position, and the greatest variation in this regard occurred just prior to locomotory attack. To this extent only was there a typical chain of responses from flock approach to the termination of an attack. A particular combination of components (mouth open, neck tightly coiled, and head aimed at the opponent) appear as threshold elements in the attack response. At times when attack probability is demonstrably low, the reciprocal of these components (i.e., mouth closed, neck extended horizontally and head directed away from the rival) frequently appears. The latter component group also appears ill-adapted for the delivery of a powerful blow by head or wings. It is suggested that complex responses consist of several behavioral components, and that any analysis of a response or between responses must incorporate all of these components. To interpret on the basis of a single component may, and often does, lead to over-simplification and possibly distortion. The coil-down neck position appears in the attacking bird and at times in the bird to be attacked. Using all components in an evaluation, the latter seems to be a protective response, and hence, there does not appear to be causal affinity between these responses. In locomotory attack, two morphologically dissimilar neck positions are recorded. The probability of their occurrence is correlated with the neck position found prior to locomotory attack. The proximity between rivals at different stages of attack was measured. Neck positions that appear to represent relatively attack intensity proved to have relatively low mean separation distances between rivals before attack and vice versa. Proximity measures after attack do not indicate any clear trend. Wing-flapping appears frequently in locomotory attack for both captive and wild geese, whether on land or water. Flapping appears to increase the probability of contact between rivals, and may represent an increase in the intensity of the response. As clearly a threshold element in locomotory attack, it seems to make its appearance under conditions of high stimulus contrast.