Knowledge of basic data variability is essential for the interpretation of any proxy-based paleotemperature record. To evaluate this for δ18O stable isotope paleothermometry based on early Paleogene fish otoliths from marginal marine environments, an intra- and interspecific stable O and C isotope study was performed at a single locality in the southern North Sea Basin (Ampe Quarry, Egem, Belgium), where shallow marine sands and silts are exposed. The age of the deposits is early late Ypresian (ca. 50.9 Ma) and falls within the early Eocene climatic optimum (EECO) interval. In each of four fossiliferous levels sampled, the same three otolith species were analyzed (Platycephalus janeti, Paraconger papointi and “genus Neobythitinorum” subregularis). Intrataxon stable isotope spread amounts on average 2.50-3.00‰ for all taxa and is present in all levels. This implies that each sample level comprises substantial variability, which can be attributed to a combination of temporal and taphonomic effects. More importantly, intertaxon offsets of 4.60‰ in δ13C and 2.20‰ in δ18O between the mean values of the three otolith species are found, with “N.” subregularis representing more positive values relative to the other species. We hypothesize that freshwater influence of coastal waters is the most likely cause for these discrepancies. Similar analyses on two coastal bivalve species (Venericardia sulcata and Callista laevigata) corroborate this hypothesis. Accordingly, δ18O values measured on “N.” subregularis otoliths probably represent a more open oce- anic signal, and therefore seem well-suited for δ18O stable isotope paleothermometry. This study highlights the importance of investigating data variability of a biogenic carbonate paleotemperature proxy at the species level, before applying paleotemperature equations and interpreting the outcome.