Schemas in Problem Solving explores a new theory of schema development and studies the applicability of the theory as a unified basis for understanding learning, instruction, and assessment. The theory’s prescriptions for teaching are direct, and its application to assessment suggests new directions for tests. After examining the roots of the theory in earlier work by philosophers and psychologists, Marshall illustrates the main features of her theory with experimental evidence from students who are learning to recognize and solve arithmetic story problems. She describes individual performance with traditional empirical studies as well as computer simulation. The computer simulation reflects a new approach in modelling cognition. Marshall’s model links neural networks with symbolic systems to form a hybrid model that uses pattern matching of sets of features as well as logical step-by-step rules.