An attempt is made to revisit the main theoretical considerations concerning temperature effects ("Joule heating") in electro-driven separation systems, in particular lab-on-a-chip systems. Measurements of efficiencies in microfabricated devices under different Joule heating conditions are evaluated and compared to both theoretical models and measurements performed on conventional capillary systems. The widely accepted notion that planar microdevices are less susceptible to Joule heating effects is largely confirmed. The heat dissipation from a nonthermostatically controlled glass microdevice was found to be comparable to that from a liquid-cooled-fused silica capillary. Using typically dimensioned glass and glass/silicon microdevices, the experimental results indicate that 5-10 times higher electric field strengths can be applied than on conventional capillaries, before detrimental effects on the separation efficiency occur. The main influence of Joule heating on efficiency is via the establishment of a radial temperature profile across the lumen of the capillary or channel. An overall temperature increase of the buffer solution has only little influence on the quality of the separation. Still, active temperature control (cooling, thermostatting) can help prevent boiling of the buffer and increase the reproducibility of the results.