Do computer simulations really cognize?

Abstract Several common arguments against the claim that computers can cognize are based on the assumption that simulation can never be replication. In this article, simulation is defined as the construction of a device that represents the causal capacities of a system of interest. The relationship of simulation to theories and to replication is analysed, from which it is clear that it is possible for a simulation to be a replication if the device used can not only represent but also instantiate the same capacities as the system. Determining whether this is the case for computer simulations of cognition will require more coherent hypotheses than the symbol-processing hypothesis, and more cogent tests than Turing-type tests.