It is shown that the dyes used to produce banding patterns on chromosomes, quinacrine and Giemsa, are bound to DNA, and not to non-histone protein, the other chromosomal component remaining after acetic acid fixation. Studies on fixed nuclei and on extracted DNA in gelatine films show that the amount of dye bound is not affected by whether the DNA is native or denatured, and is not directly related to the amount of DNA present. Quinacrine is bound to the DNA ionically. With Giemsa, a new magenta compound is formed in situ, consisting of two molecules of methylene blue and one of eosin; this compound is attached to the chromosome by hydrogen bonds. Both quinacrine and the magenta compound formed from Giemsa appear to be attached to DNA molecules at two separate points, and the available evidence suggests that the amount of dye bound is related to the concentration of the DNA. It is suggested that the dye molecules bridge longitudinally separated sites brought into close proximity by folding of the DNA, and that the spatial arrangement of sites in the chromosome is influenced by non-histone proteins. It is concluded that chromosome banding is thus a consequence of the reduction of dye binding in those regions where the DNA chains become sufficiently dispersed to prevent bridging by the dye molecules. Possible indirect effects of base composition and repetition on dye binding at certain chromosomal sites are discussed.