The paper addresses indoor environment performance of a naturally ventilated residential building during heat waves. For the assessment of an extremely hot indoor environment a heat stress period (HSP) is defined. By the HSP a quantification of potentially harmful extreme heat exposure of building occupants is intended. The developed probabilistic model of HSP durations is based on the statistical theory of extreme values. The quantification is expressed by the mean return period related to the annual probability of exceedance of a considered HSP duration. This concept allows comparison of different designs; assessment of retrofitting effectiveness or it may serve as essential preliminary input for a health cost-benefit analysis. A dwelling in a typical naturally ventilated residential building represented by two opposite rooms connected by a corridor is studied. The indoor environment is simulated by the ESP-r software tool. Hourly time series of meteorological data describing climate during 14 years at a lowland station in Slovakia are employed.