A digital microfluidic (DMF) device was applied to a heterogeneous sandwich immunoassay. The digital approach to microfluidics manipulates samples and reagents in the form of discrete droplets, as opposed to the streams of fluid used in microchannels. Since droplets are manipulated on relatively generic 2-D arrays of electrodes, DMF devices are straightforward to use, and are reconfigurable for any desired combination of droplet operations. This flexibility makes them suitable for a wide range of applications, especially those requiring long, multistep protocols such as immunoassays. Here, we developed an immunoassay on a DMF device using Human IgG as a model analyte. To capture the analyte, an anti-IgG antibody was physisorbed on the hydrophobic surface of a DMF device, and DMF actuation was used for all washing and incubation steps. The bound analyte was detected using FITC-labeled anti-IgG, and fluorescence after the final wash was measured in a fluorescence plate reader. A non-ionic polymer surfactant, Pluronic F-127, was added to sample and detection antibody solutions to control non-specific binding and aid in movement via DMF. Sample and reagent volumes were reduced by nearly three orders of magnitude relative to conventional multiwell plate methods. Since droplets are in constant motion, the antibody-antigen binding kinetics is not limited by diffusion, and total analysis times were reduced to less than 2.5 h per assay. A multiplexed device comprising several DMF platforms wired in series further increased the throughput of the technique. A dynamic range of approximately one order of magnitude was achieved, with reproducibility similar to the assay when performed in a 96-well plate. In bovine serum samples spiked with human IgG, the target molecule was successfully detected in the presence of a 100-fold excess of bovine IgG. It was concluded that the digital microfluidic format is capable of carrying out qualitative and quantitative sandwich immunoassays with a dramatic reduction in reagent usage and analysis time compared to macroscale methods.