Sensitivity of taphonomic signatures to sample size, sieve size, damage scoring system, and target taxa

The sensitivity of taphonomic signatures to a battery of common sampling an d analytic procedures is tested here using modern bivalve death assemblages from the San Bias Archipelago, Caribbean Panama, to determine (a) the magn itude of methodological artifacts and, thus, the comparability of taphofaci es patterns among studies; and (b) the most efficient and robust means for acquiring damage profiles (taphonomic signatures) of death assemblages both ancient and modern. Damage frequency distributions do not stabilize below sample sizes of 120-150 individuals. Using damage to the >8 mm portion of t he assemblage as a baseline (interior damage only, fragments included), it is found that qualitative trends among environments (higher damage levels i n reefal skeletal gravel versus mud) and the rank-order importance of tapho nomic variables per environment (intensity of damage from encrustation, bor ing, fine-scale alteration, edge-rounding, fragmentation) are robust to mos t methodological decisions. The exception is the use of target taxa: of thr ee genera tested, only one was sensitive to the same suite of environmental differences as the total-assemblage, and taxa had disparate rank-ordering of variables. In contrast to the general robustness of qualitative trends, quantitative damage Levels are affected significantly by methodology. Speci fically, the measured frequency of damage is generally lower for finer size fractions and finer sieve sizes, for whole shells versus fragments, for ta xonomically well-resolved specimens, for infaunal versus epifaunal species regardless of mineralogy, and for interior surfaces versus exterior or tota l surface area of shells. Full frequency-distribution data on states of tap honomic damage are most powerful for differentiating samples, but if single -value metrics are desired, the frequency of high-intensity damage is more powerful-and shows less between-operator variance-than presence-absence dat a or average damage state. To maximize the detection of damage and of betwe en-environment differences in taphonomic signature, and to foster between-s tudy comparisons, the following are recommended: (1) analysis of discrete s ize-fractions rather than broad spectra and, in particular, the separate tr eatment of coarse size fractions (>4 mm); (2) examination of complete assem blages (fragments as well as whole specimens; all species or broad subsets of species rather than select taxa); (3) variables scored independently (e. g., encrustation v. boring) rather than grouped into summary grades; and (4 ) evaluation of rank-ordering of variables in plots of threshold damage pro files as a complement to ternary taphograms.