Over the last two decades, the complementary properties of symbolic and connectionist systems have led to a number of attempts at hybridizing the two approaches to leverage their strengths and alleviate their shortcomings. The fact that those attempts have generally fallen short of their goals largely reflects the difficulties in integrating computational paradigms of a very different nature without sacrificing their key properties in the process. In this paper, we propose that biological plausibility can serve as a powerful constraint to guide the integration of hybrid intelligent systems. We introduce a hybrid cognitive architecture called SAL, for “Synthesis of ACT-R and Leabra”. ACT-R and Leabra are cognitive architectures in the symbolic and connectionist tradition, respectively. Despite widely different origins and levels of abstraction, they have evolved considerable commonalities in response to a joint set of constraints including behavioral, physiological, and brain imaging data. We introduce the ACT-R and Leabra cognitive architectures and their similarities in structures and concepts then describe one possible instantiation of the SAL architecture based on a modular composition of its constituent architectures. We illustrate the benefits of the integration by describing an application of the architecture to autonomous navigation in a virtual environment and discuss future research directions.