Microchip separations can achieve high resolution at speeds much faster than conventional capillary electrophoresis, due to the ability to inject a very small sample plug defined at the intersection of two channels. To achieve these benefits, separations must be carried in conditions where resolution is controlled both by sample plug size and by diffusion. Thus, to design an optimal microchip separation system, it is necessary to determine the diffusion coefficients of the species being separated. We propose a novel method to determine diffusion coefficients in separation systems. Since we have the ability to observe species at any point of the separation channel, we can measure the widening of a peak as a function of separation distance. This allows diffusion to be measured at a constant electric field, as opposed to varying the field to achieve different diffusion times at a fixed observation point. This distinction is important in cases such as DNA separations in polymeric sieving matrices, where the diffusion coefficient has been reported to be field-dependent. In this work, glass microchips with channels 10 mum deep and 30 mum wide were used for separations, at distances varying from 3 to 15 mm. Samples used were fluorescent molecules, single-stranded DNA oligos, and dsDNA. For the DNA samples, diffusion coefficients between 3.6 x 10(-7) and 4.2 x 10(-8) cm(2)/s were observed. For double-stranded DNA, we found that the diffusion coefficients increased strongly as the electric field was increased.