Previous work has shown that nucleosome repeat lengths, and hence linker DNA lengths, are preferentially quantized to a set of values differing by integral multiples of the helical twist of DNA. An explanation was proposed in which this preferential quantitation is due to twist constraints on linker DNA arising from nucleosome-nucleosome interactions in folded chromatin. Here we report the results of a study, using ethidium intercalation, designed to test whether twist constraints do indeed exist. Electron microscopy reveals that ethidium intercalation causes decondensation of dinucleosomes. Direct measurement of the free energy of intercalation by fluorescence spectroscopy reveals competition between chromatin folding and ethidium intercalation. Results from other laboratories establish that these effects of ethidium are due to ethidium-induced changes in the twist of linker DNA, and not to a variety of other possible effects. We conclude that twist constraints on linker DNA do exist. These may explain the observation of preferentially quantized linker DNA lengths. Implications of these results for mechanisms of nucleosome phasing and the mechanisms of drug action are discussed.