In this paper we consider the output synchronization problem for heterogeneous networks of linear agents. The network's communication infrastructure provides each agent with a linear combination of its own output relative to that of neighboring agents, and it allows the agents to exchange information about their own internal observer estimates. We design decentralized controllers based on setting the control input of a single root agent to zero and letting the remaining agents synchronize to the root agent. A distinguishing feature of our work is that the agents are assumed to be non-introspective, meaning that they possess no knowledge about their own state or output separate from what is received via the network. We also consider the problem of regulating the agreement trajectory according to an a priori specified reference model. In this case we assume that some of the agents have access to their own output relative to the reference trajectory. (c) 2012 Elsevier Ltd. All rights reserved.