We use new macroevolutionary rate estimates to resolve the dynamics of severe versus background extinction through the history of a major, globally distributed, Paleozoic zooplankton clade, the graptoloids. Our data span one of the “Big Five” mass extinctions, the Late Ordovician Mass Extinction (LOME), and several secondary, severe extinction events. We use cohort survivorship curves to derive both “instantaneous” rates and smoothed rates based on “natural” time bin intervals that honor the structure of the data. We avoid the approximation of many approaches that average rate estimates within essentially arbitrary time bins. We find that 63 % of graptoloid extinctions lie within intervals classified previously as “background” extinction; only 7 % lie within the LOME, and the remainder lie within the spans of 15 other secondary extinction events spread through the Ordovician and Silurian. Extinction rate magnitudes define a continuous, unimodal distribution. Background extinction in the graptoloids is not stochastically uniform but includes many more high-rate pulses than expected under a null model of uniform, memoryless extinction. Our results support the inference of pulsed extinction in the marine realm, with pulses occurring on timescales much finer than the standard age divisions of the Ordovician and Silurian periods. The LOME and secondary extinction events are not characterized by instantaneous extinction rates that are higher than so-called background. Instead, extinction events are distinguished from background by increased duration of their component, short-lived pulses of elevated extinction, and the LOME represents a protracted interval with multiple such pulses and little time for faunal recovery. Our results are consistent with the notion that, whereas a mass or severe extinction may have an exceptional or singular initial trigger, the effects of that trigger propagate out to global-scale species loss via a complex web of processes that are common to many extinction episodes and may take significant time.