DRILLING FREQUENCY IN THE FOSSIL RECORD: WHEN IS A ZERO REALLY A ZERO?

Paleoecological studies on molluscan drilling predation commonly included predation frequencies of zero for one or more species in an assemblage. When sample sizes are low (e.g., n < 25), zero frequencies—and predation frequencies in general—can be unreliable and it can be difficult to differentiate between structural (i.e., real) and sampling (i.e., artificial) zeroes. When modeling predator-prey interactions and considering the preferences of predators for their prey, the inability to distinguish between structural and sampling zeroes has the potential to result in faulty conclusions (at worst), or the elimination of data from analyses (conservatively).

Using a simulated community, we establish probabilities for structural and sampling zeroes at low sample sizes. We simulate an assemblage of 10,000 individuals, distributed across 100 species. Each species is randomly assigned a trait value—ranging from 0 – 1—from a truncated log-normal distribution. From this assemblage, we simulate predation of 1000 individuals on the basis of trait values, applying predator preference for certain values. In the resulting “drilled assemblage” structural and sampling zeroes are determined with respect to a prey species’ trait value and the predator’s preferred trait value. That is, a structural zero occurs when a species with a trait value outside of the predator’s preference range is “correctly” not included in the drilled assemblage and a sampling zero occurs when no individuals from a species that would be expected to be consumed based on its trait value occur in the drilled assemblage. Intuitively, the likelihood of sampling zeroes occurring increases with decreasing sample size. By incorporating the probability for sampling zeroes in analyses of drilling predation, paleoecologists can be more confident with the data they include, or omit, and increase the ecological fidelity of their studies.