Intercomparison experiments were made between two independent techniques for ozone (O3) reference standards often used as the primary standards in air quality monitoring networks. These techniques include ultraviolet absorption photometry of O3 at the 253.7-nm Hg line and gas-phase titration of O3 with excess NO. For ultraviolet photometry, a well-designed and maintained standard reference photometer (SRP) built by the National Institute of Standards and Technology (USA) was employed. For gas-phase titration (GPT), an existing system was significantly modified by the National Institute for Environmental Studies (Japan) by using gravimetric NO/N2 standard gases, accurate flow measurement systems based on laminar flow elements, and two chemiluminescence NO detectors to minimize uncertainty in the measurements, which had previously been a major shortcoming of this method. Uncertainty in the improved GPT system was reduced to less than 0.4% above 100 nmol mol-1 O3 mole fraction. A series of comparison runs between the two methods over the course of 13 months from August 2004 to August 2005 showed a significant discrepancy, which cannot be explained by the measurement uncertainties attributed to either SRP or GPT in the range of 80--800 nmol mol-1 O3, where GPT was about 2% higher than SRP. This result indicates possible biases in the currently existing O3 reference standards and warrants further studies to identify and characterize possible sources of the systematic discrepancy.