This paper aims to validate and improve three cross and single-sided natural ventilation models implemented in a whole-building hygrothermal and energy simulation program. The tested models are the British Standard for cross ventilation, the de Gids and Phaff's and Larsen's for single-sided ventilation. Airflow rates obtained by those models have been compared to the measurements performed in two full-scale buildings: one single room house located in a wind tunnel facility and one real three-storey building. Results show a large variation of airflow rates provided by the different models. The Larsen's model can be improved if coupled to the CPCALC algorithm, providing better results for both wind tunnel and on-site experiments. (c) 2013 Elsevier B.V. All rights reserved.