The clearest pattern that emerges from these field studies is that larger trees are more likely to be damaged than smaller trees, but this seemingly simple relationship is made more complex because the rate of increase in risk of damage or treefall is not constant across sites or species. For example, in two sites that have abundant beech (Fagus grandifolia) and hemlock (Tsuga canadensis), the relative vulnerability of these species reversed. Moreover, relative vulnerability among species is a function of size: in small size classes, hemlock is less likely to be toppled than beech, but in larger size classes, hemlock is more vulnerable. Broad generalizations are emerging that shallow-rooted species are more likely to be toppled by wind than tap-rooted species, and pines are often more damaged than hardwoods – but there are clear exceptions. In the most severely-damaged sites, differences among species become obscured, because essentially all trees above some minimum size are damaged or toppled. Finally, within particular sites, anomalous patterns remain (e.g. a large tree of a vulnerable species that is undamaged amidst damaged neighbors) that suggest influences by other non-biological factors, such as soil depth. Much improvement is needed in how tree damage is assessed in the Fujita scale rankings if more accurate assessments of tornado intensity are to be achieved.