A multinomial model of the “Who said what?” paradigm (S. E. Taylor, S. T. Fiske, N. J. Etcoff, & A. J. Ruderman, 1978) explains the pattern of participants' assignment errors by means of the joint operation of several processes. Specifically, memory for discussion statements, person memory, category memory, and 3 different guessing processes can be accommodated by the model. The model's ability to disentangle these processes is validated in a series of 5 experiments. The model thereby enables a more refined use of the “Who said what?” paradigm in testing theories of social categorization. This is demonstrated in a 6th experiment in which the validated model is applied to the study of the effects of cognitive load on categorization.