The usefulness of monitoring electrical resistivity changes in a stressed crust as a means of studying crustal behaviour during dilatancy prior to earthquakes is considered. Examination of published laboratory data shows that much of the large decrease of resistivity of stressed rocks prior to failure is accounted for by 'surface conductivity' in electrical double layers at rock–fluid interfaces in microcracks. Resistivity changes in the crust might be observed before other dilatancy-induced precursory phenomena are seen because of the strong effect of the surface conductivity. Measurements of resistivity changes alone with time cannot be used for predicting the time of occurrence of an earthquake because of uncertainties in factors such as pore fluid pressure and degree of saturation of newly formed cracks. However, observations of time-dependent resistivity (or impedance) variations might provide information on crustal behaviour during dilatancy, since as cracks begin to open in some preferred direction the impedance anisotropy of the crust should change. Such changes would be seen in the magnetotelluric impedance tensor elements. Information on crustal behaviour during dilatancy is required if a means of predicting the time and location of earthquakes using observations of time-dependent, dilatancy-induced precursory phenomena is to be developed.