The citation rates of scientific papers, long used by numerous sociologists of science to measure the influence of individual scientists and the diffusion of knowledge, are shown to be partly affected by the various structural characteristics of these papers. Based on an analysis of 221 scientific papers in three cocitation clusters, between 15 and 35 percent of the variation in citation rates is found to be a function of those papers' textual and nontextual characteristics. The citation rates of papers in the Burkitts-Lymphoma and Heavy Quark Potential clusters are shown to be heavily dependent on abstract characteristics such as readability and number of uncommon words. The citation rates of DNA cluster papers are observed to be significantly affected by both the number of references and figures found in teh body of those papers. Of particular note is the fact that while the readability of abstracts is shown to decrease the citation rates of Burkitts-Lymphoma papers, the opposite is true of Heavy Quark Potential papers.