Traditionally models of oceanic pelagic ecology, which also lay at the heart of general circulation models used for climate change simulations and of models describing coastal ecosystem dynamics have employed descriptions of plankton that assume fixed Redfield elemental compositions as inputs for rectangular-hyperbolic (Monod-type, Holling type II) descriptions of resource-limited growth and predation kinetics. The performances of Redfield–Monod and variable stoichiometric models are compared with theoretical expectations and experimental data for descriptions of multi-nutrient limited phytoplankton growth and predator–prey interactions. Serious deficiencies are revealed in Redfield–Monod implementations; such constructs have outputs and/or structural logic contrary to empirical biological knowledge. For example, Redfield–Monod models often employ nutrient limitation as a significant factor controlling phytoplankton growth, and yet biologically such nutrient limitation is associated with significant variation in elemental stoichiometry. One could argue that reliance on such dysfunctional descriptions is unacceptable, especially in an era when increasing political play is made of model simulations and predictions. Biological studies should examine the consumption and fate of all major nutrients, not just of the limiting nutrient, in order to furnish modellers with the mechanistic understanding, and data, to enable a refined description of the organisms responsible for biogeochemical cycling. Modellers, in turn, should embrace existing experimental and phenomenological observations and update their models accordingly. Simplifications in model structure can then be made from a sound knowledge base rather than by making potentially incorrect a priori assumptions.