PURPOSE: Research and development of various accelerator-based irradiation systems for boron neutron capture therapy (BNCT) is underway throughout the world. Many of these systems are nearing or have started clinical trials. Before the start of treatment with BNCT, the relative biological effectiveness (RBE) for the fast neutrons (over 10 keV) incident to the irradiation field must be estimated. Measurements of RBE are typically performed by biological experiments with a phantom. Although the dose deposition due to secondary gamma rays is dominant, the relative contributions of thermal neutrons (below 0.5 eV) and fast neutrons are virtually equivalent under typical irradiation conditions in a water and/or acrylic phantom. Uniform contributions to the dose deposited from thermal and fast neutrons are based in part on relatively inaccurate dose information for fast neutrons. This study sought to improve the accuracy in the dose estimation for fast neutrons by using two phantoms made of different materials in which the dose components can be separated according to differences in the interaction cross sections. The development of a "dual phantom technique" for measuring the fast neutron component of dose is reported. METHODS: One phantom was filled with pure water. The other phantom was filled with a water solution of lithium hydroxide (LiOH) capitalizing on the absorbing characteristics of lithium-6 (Li-6) for thermal neutrons. Monte Carlo simulations were used to determine the ideal mixing ratio of Li-6 in LiOH solution. Changes in the depth dose distributions for each respective dose component along the central beam axis were used to assess the LiOH concentration at the 0, 0.001, 0.01, 0.1, 1, and 10 wt. % levels. Simulations were also performed with the phantom filled with 10 wt. % 6LiOH solution for 95%-enriched Li-6. A phantom was constructed containing 10 wt. % 6LiOH solution based on the simulation results. Experimental characterization of the depth dose distributions of the neutron and gamma-ray components along the central axis was performed at Heavy Water Neutron Irradiation Facility installed at Kyoto University Reactor using activation foils and thermoluminescent dosimeters, respectively. RESULTS: Simulation results demonstrated that the absorbing effect for thermal neutrons occurred when the LiOH concentration was over 1%. The most effective Li-6 concentration was determined to be enriched 6LiOH with a solubility approaching its upper limit. Experiments confirmed that the thermal neutron flux and secondary gamma-ray dose rate decreased substantially; however, the fast neutron flux and primary gamma-ray dose rate were hardly affected in the 10%-6LiOH phantom. It was confirmed that the dose contribution of fast neutrons is improved from approximately 10% in the pure water phantom to approximately 50% in the 10%-6LiOH phantom. CONCLUSIONS: The dual phantom technique using the combination of a pure water phantom and a 10%-6LiOH phantom developed in this work provides an effective method for dose estimation of the fast neutron component in BNCT. Improvement in the accuracy achieved with the proposed technique results in improved RBE estimation for biological experiments and clinical practice.