Whether cancer cytogenetics can be reduced to the molecular genetics of cancer cells is a question that must be addressed in three domains, focusing on its ontological, methodological, and epistemological dimensions. The possibility of ontological reduction hinges on whether chromosomes have other important constituents than molecules. Although this must obviously be answered in the negative, it should be emphasized that both cytogenetic and recombinant DNA investigations provide us with very selective pictures of genomic organization. This is of concern because the higher order packing of DNA and its joining with other molecules to form chromosomal structures give rise to emergent properties, functional features that become manifest only at higher levels of complexity and that may not be deducible from the base pair composition of the DNA. A position of extreme methodological reductionism would in our context be that the best research strategy is always to investigate the genetic changes of tumor cells at the highest possible resolution level, as alterations of genes and, ultimately, as changes in DNA primary structure. There are two fundamental differences between cytogenetic and molecular genetic techniques that make this stance untenable. First, whereas cytogenetic investigations are open-framed (all chromosome aberrations are revealed), molecular genetic analyses are highly specific (only those aberrations are revealed that one tests for). Second, whereas the molecular approach determines the genotypic constitution of an idealized, average tumor cell, cytogenetic analysis is of real, individual cells. These may not necessarily be representative of the main population of the tumor, but at least whatever karyotypic differences exist between them are detected. Heterogeneity and clonal evolution within the tumor can thereby be assessed.(ABSTRACT TRUNCATED AT 250 WORDS)