As the ongoing biodiversity crisis begins to affect our species negatively, the need to understand the processes underlying extinction grows. Past biotic crises can test methods for detecting warning signs and recoveries. This study uses rank-abundance curves (RACs) to examine community response to rapid global warming during the Paleocene-Eocene Thermal Maximum (PETM), an appropriate analogy for anthropogenic climate change. RACs are an excellent tool for identifying stress because the curves reflect resource-partitioning within a community, and resource-partitioning changes predictably with stress. Following stress, resource-partitioning is typically uneven (producing convex-down RACs), and becomes more even (convex-up RACs) as the community stabilizes and recovers. A community experiencing continual stress may never stabilize, and the community structure will reflect high stress (convex-down RACs). An extinction event is a period of high stress, and as such the community response could be quantified by RACs. Equally importantly, RACs can document increasing environmentally-induced community stress that may precede extinctions, thus serving as a potential "early warning system." To map ecological effects of the PETM on deep-ocean benthic communities, RAC-derived curve-area and kurtosis values were calculated for high-resolution (cm-scale) benthic foraminiferal assemblage data through the onset of the PETM at ODP Site 690 (Thomas, 2003) and for lower-resolution ostracode assemblage data spanning the PETM at ODP Site 689. Both communities are relatively stable before the PETM, but each responds similarly to the event. Site 689 ostracode communities show increasing stress prior to the Carbon Isotope Excursion (CIE) marking the event and consistently high stress during the CIE. Site 690 benthic foraminiferal communities are stable prior to the CIE, increasingly stressed through the CIE and subsequent benthic foraminiferal extinction, and then stabilize afterwards. Pre- and post-extinction kurtosis values are significantly different (t-test, p<0.01 for both community types) and highlight the utility of RACs for identifying stress responses before major taxonomic loss. Thus, our approach may have great potential for examining other extinction events including the current crisis.