Despite decades of use as a paleoceanographic proxy, considerable uncertainty still surrounds the temperature dependence of benthic foraminiferal δ18O. Widely applied paleotemperature equations may mix non-equilibrium foraminifera with equilibrium synthetic calcite, resulting in temperature sensitivities that are too large. Warm-water foraminiferal calibrations may give temperature sensitivities that are too small for very cold waters. Here we combine new core top measurements from the Florida Straits and the Arctic Ocean with published data to derive new δ18O:temperature relationships for three groups of benthic foraminifera. We derive a quadratic equation for Cibicidoides and Planulina that agrees well with equilibrium synthetic calcite, and that should be applicable over all oceanographic temperatures. We find that Uvigerina is not at equilibrium and is isotopically heavier than Cibicidoides and Planulina by 0.47‰, in contrast to the historically used 0.64‰. Hoeglundina elegans is further enriched and appears to be slightly heavier than equilibrium aragonite. Finally we discuss the implications of the Florida Straits observations for the hypothesis that benthic foraminifera precipitate their shells from a pH-dependent mixture of bicarbonate and carbonate ions.