The document proposes an interpretation-driven model of syntax to address shortcomings in current grammar formalisms from the perspective of developing a system that can understand natural language. It discusses theoretical shortcomings related to how linguistic knowledge is represented and implementation shortcomings regarding computational processing. The proposed model is motivated by four principles: satellite structures, feature structure unification, feature structure aggregation, and incrementality. The model is implemented in a cognitive architecture to parse and interpret language input using pairwise pattern matching of feature structures.