Fossil record biases skew inferences of mammal trait evolution

The fossil record is key to elucidating past biodiversity dynamics, yet it remains incomplete and non-uniform, governed by factors such as sedimentation rates, species' range and body size, and sampling effort. How these biases distort inferences on evolutionary trends remains poorly understood. Here, we simulate a future fossil record of present-day land mammals to quantify the extent to which real evolutionary patterns can be recovered from fragmentary fossil data. Species are filtered according to sediment availability, species' range and body size, and spatial and taxonomic sampling biases under three preservation rates, and then systematically pruned from a baseline mammal phylogeny before analyses of body mass and diet evolution through phylogenetic comparative methods. We find that biological and taxonomic filters strongly skew these inferences, while spatial filters do not significantly distort them. Our results provide first-order estimates of how fossil record incompleteness can bias our interpretation of deep-time biodiversity change.