The relative efficiency of freeze-fracture replicas versus thin sections for the visualization and quantification of gap junctions in biological tissues has been evaluated. Both methods may underestimate gap-junction number--thin sections for reasons of tissue resolution and freeze-fracture replicas due to the mechanics of the fracturing process. Freeze-fracture misses gap junctions in regions of plasma membrane which are highly contoured, such as the overlapping basal cell processes of Drosophila imaginal wing discs and the interdigitating lateral membrane plications of intercalated discs in cardiac tissue. If the missed gap junctions are relatively large, as they are in both of these examples, freeze-fracture significantly underestimates the total gap-junctional area. Thin sections may miss small gap junctions, but in tissues which contain a range of gap-junction sizes the lost junctions constitute a relatively small fraction of the total junctional area. In neoplastic imaginal wing discs, thin sections were as efficient as freeze-fracture replicas in identifying even the smallest gap junctions. Although freeze-fracture may be the better technique for the qualitative and quantitative documentation of small gap junctions in tissues with relatively flat to gently contoured plasma membranes and thin sections may be the superior method for gap-junction quantification in tissues containing a range of gap-junctional sizes and highly contoured cellular processes, the data suggest that a combination of the two approaches should be utilized whenever possible.