The minor element composition of calcium carbonate (CaCO<sub>3</sub>) biominerals from marine calcifying organisms leaving a sedimentary record has been used for decades to reconstruct various biogeochemical parameters. Advancing geochemical proxies and understanding their underlying mechanisms is essential for climate reconstructions, environmental research, and investigations of biomineralization processes. Despite considerable success of proxy applications, limited mechanistic understanding still restricts their full potential. The problem is often summarized by the term "vital effect", i.e. minor element partitioning due to biological activity. The element partitioning from the calcifying fluid into the biomineral, however, is usually described in terms of inorganic precipitation of a mineral from an aqueous solution of inorganic ions. Although this assumption is central to many partitioning models it has not been tested because the calcifying fluid of classic proxy archives such as foraminifera, molluscs, and coccolithophores has not been successfully sampled for element analysis. The calcifying fluid of fish otolith formation (endolymph), by contrast, was sampled and chemically analysed accompanied by corresponding otolith data. However, previous datasets have not been compared to inorganic partitioning coefficients to test this assumption. In this study, we address this gap using published data from four fish species and six elements. Our results indicate that the final stage of otolith minor element incorporation is influenced by organic matter in the endolymph fluid and therefore cannot be considered purely inorganic. Our conclusion questions a central assumption of many minor element partitioning models. This does not imply that existing models are questionable, but that they share a common oversimplification. By removing this oversimplification all kinds of different models can be improved. Our study contributes broadly to the understanding of biogenic CaCO<sub>3</sub> geochemistry, and it is relevant to the majority of existing models.