The origins of evolutionary innovations have been intensively studied, but relatively little is known about their large-scale ecological patterns. For post-Paleozoic benthic marine invertebrates, which have the richest and most densely sampled fossil record, order-level taxa tend to appear first in onshore, disturbed habitats, even in groups that are now exclusively deep-water (so that present-day distributions are not reliable indicators of original environments). New results presented here show that the onshore-origination pattern is robust to shifts in taxonomic methods and to new paleontological discoveries, and the few available studies suggest that this pattern can also be seen in terms of excursions in morphospace or the acquisition of derived character states, without reference to taxonomic categories. The environmental pattern at high levels contrasts significantly with the origin of low-level novelties (such as defined genera and families) in crinoids, echinoids, and bryozoans, where first appearances tend to conform to their clade-specific bathymetric diversity gradients. This discordance seems to eliminate potential driving mechanisms that simply scale up within-population genetic or ecological processes. Little is known about the factors that promote the onshore-offshore expansion of orders across the continental shelf, or that drive some clades to abandon ancestral habitats for an exclusively deep-water distribution. The origin of evolutionary innovation must ultimately reside in developmental changes, but the onshore-origination bias could emerge from two different dynamics: the pattern could be primarily genetic and developmental, i.e., innovations truly arise onshore; or primarily ecological, i.e., innovations arise randomly but preferentially survive onshore. Whatever the ultimate driving mechanisms, these macroevolutionary patterns show that theories of large-scale evolutionary novelty must include an ecological dimension.