Trenching-a widely used method for evaluating fault activity-has limitations that can mislead investigators. Some segments of fault strands in trench walls may not be visible, and this nonvisibility can lead to incorrect interpretations of time of most recent displacement and recurrence intervals on a fault.

We examined the logs of 163 trench exposures and tabulated data on more than 1,200 fault strands to investigate three categories of nonvisibility: (1) strands with obscure (invisible or poorly visible) segments, (2) strands that die out upward, and (3) strands that die out downward. About 14 percent of all the strands have obscure segments. Of the 143 strands on which it is possible to recognize dieout up (limited to strands for which position of ground surface at time of faulting is known), 45 percent do die out upward, and the fraction exceeds 70 percent for strike-slip and reverse faults. Thus a fault strand overlain by an apparently undisturbed deposit is not necessarily older than the deposit. More than 30 percent of all the strands die out downward, providing more evidence that fault strands can end for reasons other than being covered by deposits younger than the fault.

Analysis of trench-log data revealed various relations between geologic factors and nonvisibility of fault strands. For example, fault type affects the incidence of nonvisibility, which is generally most common on strike-slip faults, less common on reverse faults, and least common on normal fau Its. The type of material penetrated by the fault also influences nonvisibility, which tends to be more common in soil horizons and sand, and less common in gravel. Dieout down is weakly influenced by fault displacement, decreasing in frequency with increase in displacement; the frequencies of obscure segments and dieout up do not vary consistently with fault displacement. Frequency of obscure segments generally decreases with increase in length of obscure segments, and frequency of dieout up generally decreases with depth of dieout up. Length of obscure segments and depth of dieout up are typically less than the effective thickness of associated beds. On the basis of few data, obscure segments seem to be more common on faults with younger, rather than older, ages of latest displacement. 

Our study revealed additional relations not directly related to nonvisibility. For example, the median widths of faults crossed by the trenches vary by fault type, strike-slip faults being narrower than dip-slip faults. In the shallow and mostly unconsolidated materials cut by the trenches, fault widths show only an erratic and, at best, weak relationship to fault displacements. Hanging walls are deformed more frequently than footwalls in dip-slip faults, but both walls are deformed at more than 30 percent of the exposures.

We tabulated several phenomena that may indicate faulting or provide evidence of prehistorical earthquakes. Rotation of pebbles was identified in 41 percent of the exposures having gravel in the fault zone; type of fault has no strong influence on the incidence of pebble rotation. Fissures were recorded at 52 percent of the exposures and were more common in strike-slip and normal faults than in reverse fau Its. Gouge was reported at 1 5 percent of the exposures; fault type has no significant influence on its frequency. Slickensides were noted at 10 percent of the exposures, and fault type has an unknown influence on their incidence. Slickensides in unconsolidated materials were restricted to clay, silt, and gouge. Other mechanical or hydrologic effects related to faulting or earthquakesrubble, breccia, mixing, crushing, polishing, water barriers, c;ind probable liquefaction effects-were reported at fewer than 1 0 percent of the exposures.