Many estimates of seismic risk depend crucially on how well the date of faulting events can be determined from the stratigraphic position of fault strands exposed in exploratory trenches. However, fault strands cut by trenches may seem to die out where they are in fact only poorly expressed, or they may actually die out. Such ambiguity can lead to misinterpretation of the time of the most recent displacement and also of the recurrence interval between faulting events. To investigate the frequency of occurrence and other characteristics of strands that falsely appear to die out or that actually do die out, more than 1200 fault strands were analyzed. For strands associated with individual faulting events, 45% could not be visibly traced in the trench walls to the ground surface that existed at the time of faulting; for reverse and strike-slip faults, the number exceeds 70%. Thus, any apparent upward termination of a fault strand that is thought to indicate the date of a faulting event requires critical examination and corroboration from other evidence. Nonvisibility (i.e., dieout upward, obscure segments, and dieout downward) is more common on strike-slip and reverse faults than on normal faults and is more common in soil horizons than in clay and gravel. Nonvisibility occurs on strands that have had large (> 1 m) as well as small displacements.