We propose a model in which researchers maximize their number of times they are cited in later papers. The equilibrium is inefficient, because researchers distort their effort toward writing on popular topics. This inefficiency is smaller when citations in higher-ranked journals carry a higher weight, or when citations in time-remote papers carry a higher weight. We also discuss (a) the relation between the quality of papers and the number of citations, (b) the relation between efficiency and the empirical distribution of citations, and (c) the response of researchers motivated by citations to anticipated progress in available research methods.