Experimental measurements of misfit dislocation velocities, obtained from in situ transmission electron microscope observations of strained GexSi1-x/Si(100) heterostructures, are compared with predictions of the diffusive double‐kink (or ‘‘kink pair’’) model of dislocation propagation. Good agreement is observed between experiment and theory for buried strained layers with applied stresses in the range of hundreds of MPa. For very thin uncapped strained layers, the diffusive double‐kink model does not describe experimental data well. In these structures better agreement between experiment and theory is obtained if we model single‐kink nucleation at the epilayer free surface. We compare our experimental data to those of other groups, and show how our modeling can reconcile apparently disparate trends deduced by these other groups.