The gel electrophoresis mobility shift assay is widely used for qualitative and quantitative characterization of protein complexes with nucleic acids. Often it is found that complexes persist within electrophoresis gels for much longer than expected on the basis of their free-solution lifetimes. Volume exclusion, direct interaction with gel matrices and the reduction of water activity by the gel have been proposed as mechanisms enhancing the stability of complexes during electrophoresis. We have used the well-characterized interaction of the E. coli cyclic AMP receptor protein (CAP) with lactose promoter DNA to test these proposals. We found that the activity of water within polyacrylamide gels differs little from that of the buffer in which they were cast and that the dependence of the dissociation rate constant on water activity is too small for osmotic stabilization to contribute significantly to the lifetimes of CAP-DNA complexes. In addition, we found that a cross-linked gel matrix is not required for the stabilization of CAP-DNA complexes, that comparable stabilization is produced by three dissimilar polymers (linear polyacrylamide, dextran and polyethylene glycol), and that these polymers stabilize complexes more effectively than equivalent weight concentrations of their cognate monomers. While these results challenge the notion that direct interaction with the gel matrix contributes to the stability of protein-DNA complexes, they are all features expected of excluded volume mechanisms.