Aims: Banksia attenuata is a resprouting species growing in deep sand, while B. sessilis is a fire-killed species occurring in shallow sand over laterite or limestone. We aimed to discover the ecophysiological basis for their different distributions by exploring their investment in deep non-cluster roots and shallow cluster roots, and their cluster-root functioning. Methods: Deep-pot (1 m), shallow-pot (400 mm), hydroponic experiments and phosphorus (P)-extraction experiment were carried out. Biomass allocation, cluster-root exudation, plant P and leaf manganese (Mn) concentrations were measured. Results: Banksia attenuata allocated more biomass to deep roots and less biomass to cluster roots than B. sessilis did in deep pots. The two Banksias released similar carboxylates in all experiments, with similar carboxylate-exudation rates in hydroponics. The carboxylate amount per unit cluster root of B. sessilis grown in shallow pots was greater than that of B. attenuata, and B, sessilis acquired more P than B. attenuata did in limestone substrate. Conclusions: Greater investment in deep roots for water uptake accounts for the presence of B. attenuata in deep sand, and vice versa for the absence of B. sessilis. A larger investment in cluster roots, which released greater amounts of carboxylates, likely accounts for B. sessilis occurring over limestone. Trade-offs in investment and cluster-root functioning support the species’ distribution patterns and life histories. Leaf Mn concentration was a good proxy for the plant capacity to acquire P.