Under stress, corals and foraminifera may eject algal symbionts ('bleach'), which can increase mortality. How bleaching relates to species viability over warming events is of great interest given current global warming. We use size-specific isotope analyses and abundance counts to examine photosymbiosis and population dynamics of planktonic foraminifera across the Paleocene-Eocene Thermal Maximum (PETM, ~56 Ma), the most severe Cenozoic global warming event. We find that, unlike modern bleaching-induced mass mortality, populations of photosymbiont-bearing planktonic foraminifera increased in relative abundance during the PETM. Multigenerational adaptive responses including flexibility in photosymbiont associations and excursion taxa evolution may have allowed some photosymbiotic foraminifera to thrive.This dataset contains new records of size-specific stable isotope compositions and relative abundance changes in three clades of planktonic foraminifera from three ocean drilling sites (ODP Site 1209, DSDP Site 401, and ODP Site 690). We also include relevant published datasets used in the corresponding paper. Published high-resolution (~1-10 kyr) bulk isotope records provide a robust framework and inform us on the overall shape and timing of the Palaeocene-Eocene Thermal Maximum (PETM), whereas published multispecies planktic and benthic foraminifera provide a range of “expected” values for a given foraminifera size. We intentionally limited our compilation to high-resolution records that provide 1) a generic and/or specific-specific determination (i.e. we generally exclude “bulk” foraminifera isotope data, unless part of the original compilation), 2) a defined range of foraminiferal size, although often only defined by a soft limit i.e. “larger/smaller than X µm”, 3) a continuous sampling resolution that resolves the shape of the PETM, in turn allowing for a data comparison across all sites and across all defined PETM time bins.All published datasets included in our compilation are well known in the palaeoceanography community. Many of these datasets have often been cited and reused in subsequent research, and persistent copy-errors are not uncommon. We used the original datasets and metadata given in the articles themselves. Original data and metadata is classically represented in tables or in the corresponding “Material and Methods” sections, published as supplementary information, or published in online databases such as Pangaea.de. We designed our compilation in a way that the data for all three sites (DSDP Site 401, ODP sites 690 and 1209) were presented in a uniform way, aiding internal comparisons and allowing further compilation work.