The Biotic Ligand Model (BLM) integrates geochemistry, physiology, and toxicology to predict the site-specific toxicity of metals relative to water chemistry. We provide a brief history of BLM development, noting that (i) modern versions are done by data-fitting, not by experimentation; (ii) no BLM has yet been developed to predict the toxicity of the most studied metal (Cu) to the most studied fish (zebrafish); and (iii) no BLM has been based on a tropical species. We have returned to experimental basics: 96-h LC50 and 3-h gill binding assays (with radiolabeled 64Cu) in soft water, where individual components were manipulated separately. These yielded Log Ktox values for Cu toxicity, Log Kgill values for gill Cu binding, and LA50 (3-h gill accumulation predictive of 96-h LC50). Key novel findings in zebrafish included (i) a different mechanism of Cu toxicity; (ii) high LA50; (iii) high protection against both gill binding and Cu toxicity by dissolved organic carbon (DOC) and strong cations (potency order Mg2+ > Ca2+ > Na+ > K+); (iv) the BLM constructed using Log Ktox values predicted acute Cu toxicity to adult zebrafish better than one based on Log Kgill values, but the performance of both was superior to generic BLMs for other fish which are in widespread use.