This paper presents a simulation study of ship cluster merging and separation characteristics using a ship cluster behavior model. In order to simulate merging and separation of ship clusters, it is necessary to establish a cluster membership recognition rule. Four kinds of such rules (course-dependent, distance-dependent, inclusive and exclusive rules) were introduced and the characteristics of ship clusters' merging and separation behavior were tested using each rule. Time changes of the ship cluster's mean distance and the shortest distance were considered, and the course-dependent rule turned out to be the best decision making strategy. Using this rule, various encounter conditions and resultant cluster behavior and configurations were evaluated. It was found that collision avoidance is easier at the time of merging than separation, indicating that merging action should be planned carefully while considering the separation behavior. Also tested in this study was the effect of a speed adjustment function that controls the distance between a given ship and the one ahead of it. The introduction of this new function was found to contribute to more compact, yet evenly distributed ship clusters.