A semantic network model is structured so that usually there is only one node in the network to represent each individual. A series of experiments were performed to determine under what circumstances subjects would show unitary memory for individuals. The experiments were principally concerned with the speed with which subjects could retrieve the facts and make inferences from them. Subjects learned facts about individuals which could be referred to by two labels. The semantic network model predicted that subjects would integrate facts learned to one label with facts learned to the other. Evidence for such integration was found, but only when considerable effort was taken to encourage the subjects to develop a unitary impression of the individual. The situation was also investigated in which the subjects did not learn of the identity between the two labels until after the facts were learned to each label individually. There was evidence that subjects set up two nodes to represent the. individual, one for each label. There was also evidence that, upon learning of the identity, subjects chose to abandon one of the two nodes and to start a process of copying information from the to-be-abandoned node to the preserved node.