Scientific methods and models are interdependent. That the techniques one uses determine which findings one gets, is evident. But equally important is the influence of our a priori expectations; they may cause us to choose inadvertently those methods that are most likely to yield results that appear to confirm an already preconceived picture of reality. The conceptual models and methods of solid tumor cytogenetics are to a large extent inherited from leukemia and lymphoma cytogenetics. We illustrate how this may bias the generation and interpretation of new findings, especially when carcinomas are investigated. These malignant epithelial tumors much more often harbor cytogenetically unrelated clones than do hematologic or mesenchymal neoplasms. Carcinoma cytogenetics is therefore extremely susceptible to selection differences, making the results heavily dependent on which sample is processed, how it is disaggregated, how and for how long the cells are cultured, and on how the analysis is performed and the results presented. This calls for more efforts to be directed toward establishing also the phenotypic nature of those cells that are being karyotyped. As one cannot yet quality-grade most clonal chromosome changes in any reliable manner, meaning that one cannot determine to what extent each aberration or each clone contributes to the neoplastic process, statements about the "true" karyotypes of tumor parenchymas should be viewed with suspicion. A complete carcinoma karyotype may be much more complex than extrapolations from the analysis of a few cells may lead one to believe.