Multi-agent systems offer capabilities that single agent systems cannot. Chief among these capabilities are robustness, scalability, modularity, and the ability to distribute the agents throughout their environment. Along with these capabilities, however, comes the challenge of coordination the agents so they are able to achieve their goals effectively and efficiently. Our work focuses on the ... [Show full abstract] issues involved with the scalability of coordination between both individual agents and teams of agents. We are specifically interested in scaling up multi-agent systems in three ways. The first is to scale up the capabilities of individual agents. When using behavior-based systems, this generally equates to adding new behaviors, often in the form of composite behaviors. The second is to scale up the number of agents in a team. The third is to scale up to having teams of teams, or meta-teams, to arbitrary levels. Having teams of teams is different than simply adding all of the agents to a single, larger team because it allows us to have a team hierarchy. This allows us to distribute subgoals to individual teams so the meta-team can collectively achieve its overall goal. With homogeneous teams, this distributed approach allows us to make coordination and cooperation tractable. With heterogeneous teams, this additionally allows us to assign distinct types of subgoals to appropriate teams. With a hierarchical approach to teams, we should be able to arbitrarily extend the number of levels present in the hierarchy to such a degree that inter-team coordination is not only possible, but highly desirable. However, when scaling in any of these three ways, the complexity of the resulting interactions can quickly become prohibitive. As a result, we are interested in techniques that are able to mitigate this complexity. Since, at a conceptual level, the problems encountered are very similar for each type of scaling, we believe a parsimonious approach to all three is warranted. Our approach is to abstract sensor information through priorities. We demonstrate this approach using flocking at both the team and meta-team levels.