In a severe accident of a BWR plant, zirconium alloy of fuel cladding reacts with steam becoming a heat source for core melting and becoming a hydrogen source for hydrogen explosion. There are a lot of publications giving empirical formulas for energy and hydrogen production due to the zirconium alloy-steam reaction. Most of the formulas are based on a parabolic law, which means the reaction rate and hydrogen production rate decrease monotonously with the reaction time due to the protective oxide layers developed on the alloy surface. However, when the oxide layers are defective for some reason, their effectiveness against the reaction is weakened and the reaction rate will increase. In order to confirm the stability of the oxide layers, an integral-type experiment on the zirconium alloy-steam reaction was set up, hydrogen production rate due to the reaction was measured as a function of exposure time and surface temperature, and then, the oxide layers on the specimen were examined after the exposure. It was confirmed that (1) the analytical results based on Cathcart's formula well explained the experimental results, and (2) the hydrogen production rate did not decrease simply with the exposure time but the decrease in effective oxide layer density resulted in loss of protectiveness and the reaction rate increased.