Current handoffs are not designed to achieve multiple desirable features simultaneously. This weakness has resulted in handoffs that are seamless but not adaptive, or adaptive but not secure, or secure but not autonomous, or autonomous but not correct, etc. To face this limitation, we propose a new kind of multipurpose handoff that simultaneously is seamless, autonomous, secure, correct, and adaptive, where each desirable handoff feature can be associated to one specific purpose. A cognitive handoff is a multipurpose handoff that trades-off different objectives to reach its intended goals, makes decisions considering information from the entire handoff environment, and exhibits good performance in any handoff scenario. The main purpose of this dissertation is to create a model-based framework intended to understand, develop, and evaluate cognitive handoffs. Using a holistic approach, we produce a new taxonomy of handoff scenarios, organize handoff variables into context domains, and make structured definitions of many desirable features. Then, using foundations on problemsolving, functional decomposition, and model-based design, we develop a methodology for systematically building cognitive handoffs. Applying such methodology, we generate a cognitive handoff functional architecture and strategies for evaluating the performance of multi-objective handoffs. Finally, changing to a reductionist approach, we execute particular models that integrate the functional architecture. As a proof of concept, we take the handoff control state-based model and develop a case study about a particular type of multi-objective handoff named correct handoff. We craft a virtual instrument that measures the performance of the correct handoff algorithm. A statistic analysis and probability models provide experimental evidences that support the worthiness of multi-objective handoffs.