Many paleoecologically important variables are measured from bulk samples, but several studies have shown that ecologically important parameters like drilling intensity and total taphonomic grade are affected by how a sample is extracted from the enclosing sediments. For example, the same sample can yield different parameter estimates depending on whether it is sieved with a coarse- or a fine-meshed sieve. Here, we extend the work of Kowalewski and Hoffmeister (2003, Palaios) by testing the effects of sieve mesh size on the paleoecologic composition of bulk samples of mollusks from the Miocene of Europe. The size of every specimen in these collections was measured, and an artificial sieving protocol was used (i.e., simulated on a computer). We tested the effects of sieving on the proportions of tiering, motility, and feeding categories, as well as combinations of the three. When mesh sizes varied by an order of magnitude (1-10 mm), the apparent relative abundances of the ecologic categories varied substantially in some (but not all) individual samples. Ecologic composition was most stable when using mesh sizes less than or equal to about 6 mm, which may not be surprising considering that larger mesh sizes excluded the bulk of the size frequency distribution. Kidwell (2001, Science) found that taxonomic congruence between life and death assemblages for modern mollusks was best when using mesh sizes greater than or equal to 2 mm; these results add an estimated upper limit on mesh sizes, for Cenozoic molluscan assemblages at least. For pre-existing data sets composed of heterogeneously collected data, there is a positive result: averaging samples together to produce a mean view of ecologic composition removed the more egregious effects of the size-filtering bias. Although comparisons of the ecologic composition of single samples may be sensitive to mesh size effects, comparisons of regional or global faunas are likely more robust. Measuring ecologic importance using biomass instead of abundance also reduced the effects of the mesh size bias.