A survey of benthic communities in the vicinity of Isla Cancun and Isla Contoy, Territory of Quintana Roo, Mexico, demonstrates that invertebrate death assemblages (i.e., future fossil assemblages) in this region reflect the in-place accumulation of remains of living benthic communities through long periods of time. Samples were collected with a diver-operated suction dredge, and 290 molluscan species were identified and counted. Although nearly three-fourths of the species are known only as dead shells in the samples, almost all individuals occurring alive in any sample are represented by dead shells of their species in the same sample. Worn shells are uncommon, and a count of disarticulated bivalve valves revealed a nearly even left-right distribution. A series of twelve Q-mode cluster analyses, utilizing five different similarity coefficients and data based on both the presence or absence and the relative abundances of species, demonstrates that the same associations of samples tend to occur whether living animals only or dead remains only are considered. A Q-mode analysis based on the relative abundances of 180 species in the death assemblages of all 50 samples results in distinct clusters corresponding directly to each of the major environments sampled. An R-mode analysis of 66 common species in the death assemblages yields three distinct groups of organisms from environments of (a) restricted circulation, (b) open marine circulation and (c) swift currents. These results imply that large-scale transport of shell material and mixing of faunas from different environments are negligible in the areas sampled. Rankings of samples in terms of species diversity and total abundance indicate that most samples occupy roughly the same position in the lists whether only dead shells or only living organisms are considered. Associations computed between sample couplets with Chi-square tests and various similarity coefficients demonstrate that the death assemblages in each couplet are similar, whereas the living communities are not. Histograms of growth ring counts and size-frequency distributions of Chione cancellata in lagoon and strait samples resemble a mortality curve expected for a single generation of a living population of that species. The assemblages of dead shells in the sediments of this region represent "time-averaged communities" (Walker and Bambach, 1971), produced by the averaging of benthic communities during sedimentation as patchily distributed populations migrate across the bottom and leave a record of their mortality behind them in the sediment.