To understand how well fossil assemblages represent original communities, paleoecologists seek comparisons between death assemblages and their source communities. These comparisons have traditionally used nearshore, marine invertebrate assemblages for their logistical ease, high abundance, and comparable census data from living communities. For large marine vertebrates, like cetaceans, measuring their diversity in ocean ecosystems is difficult and expensive. Cetaceans, however, often beach or strand themselves along the coast, and archived data on stranded cetaceans have been recorded, in some areas, over several decades. If the stranding record is interpreted as a death assemblage, then the stranding record may represent a viable alternative for measuring diversity in living communities on directly adjacent coastlines. This study assessed the fidelity of the cetacean stranding record in the eastern North Pacific Ocean. The living community in this region has been studied for over 100 years and, recently, extensive and systematic live transect surveys using ship-based observing platforms have produced a valuable source of live diversity data. Over this same period, the U.S. Marine Mammal Stranding Program has collected and archived a record of cetacean strandings along the U.S. Pacific coastline, providing an ideal death assemblage for comparison. Using fidelity metrics commonly used in marine invertebrate taphonomy, I determined that the stranding record samples the living cetacean community with high fidelity, across fine and coarse taxonomic ranks, and at large geographic scales (>1000 km of coastline). The stranding record is also richer than the live surveys, with live-dead ratios between 1.1 and 1.3. The stranding record recovers similar rank-order relative abundances as live surveys, with statistical significance. Also, I applied sample-based rarefaction methods to generate collector's curves for strandings along the U.S. Pacific Coast to better evaluate the spatiotemporal characteristics of the stranding record. Results indicate that saturation (i.e., sampling >95% assemblage) at species, genus, and family levels occurs in less than five years of sampling, with families accumulating faster than species, and larger geographic regions (i.e., longer coastlines) accumulating taxa the most rapidly. The high fidelity of the stranding record, measured both in richness and by ranked relative abundance, implies that ecological structure from living cetacean communities is recorded in the death assemblage, a finding that parallels marine invertebrate assemblages, though at far larger spatial scales. These results have implications for studying cetacean ecology in both modern and ancient environments: first, these results imply that the stranding record, over sufficiently long time intervals, yields a richer assemblage than using line-transect methods, and faithfully records aspects of community structure; and second, these results imply that geochronologically well-constrained fossil cetacean assemblages might preserve ecologically relevant features of community structure, depending on depositional and taphonomic conditions.