This paper presents a series of simulations performed with the AMBR model that demonstrate how deduction, induction, and analogy can emerge from the interaction of several simple mechanisms. First, a case of deductive reasoning is demonstrated when a problem is solved based on general knowledge. The system represents the target in different ways depending on the goal, and different solutions are generated. Second, the constructed solutions of the problems are remembered and later on used as a base for remote analogy. Finally, on the basis of the analogy made, a generalized solution of the class of problems is induced. One important characteristic of the model is that representation of the task, problem-solving, and learning are not viewed as separate modules. Instead, they are different aspects of one and the same joined work of the basic mechanisms of the architecture.