The phenomena of nonvolatile switching in amorphous-hydrogenated-silicon thin-film devices are explained by a theory based on the presence of charged inclusions in the film, having originated from the initial forming treatments. The inclusions almost penetrate the film, and are able to move under applied fields of sufficient magnitude, and in a direction determined by the applied polarity. The differences between analog and digital switching are ascribed to differences in the homogeneity of the inclusions, being affected by the nature of the top contact.