Traditional habitat simulation approaches (e.g. PHABSIM) have historically been criticized for conceptual limitations in their underlying assumptions and implementation. Despite these concerns, habitat simulation methods have been instrumental in setting instream flows in thousands of streams and rivers, and remain the most widely implemented detailed instream flow methodology. Because models are simplified representations of complex biological processes, errors are inevitable; a deeper concern are structural biases that cause systematic over- or under-estimation of model predictions. Several recent studies suggest that standard habitat simulation modelling incorporates assumptions that may systematically underestimate instream flow needs. These include i) systematic biases in velocity Habitat Suitability Curves, caused by territoriality in drift-feeding salmonids displacing subordinate fish to low velocity micro-habitats at high densities, thereby inflating the fitness value of low velocities; ii) failure to account for the effects of declining flow on production and delivery of drifting prey to drift-feeding salmonids (i.e. habitat simulation models only account for flow effects on available habitat, not prey flux, which may decline with flow more quickly than available habitat); and iii) use of focal velocities to constructing habitat suitability curves, which systematically underestimates velocity preference associated with the broader foraging area of a drift-feeding salmonid. Collectively, these factors suggest that current and historic instream flow prescriptions using habitat simulation methods may underestimate optimal rearing flows for juvenile salmonids.