Microbeam radiation therapy (MRT), a so far preclinical method in radiation oncology, modulates treatment doses on a micrometre scale. MRT uses treatment fields with a few ten micrometre wide high dose regions (peaks) separated by a few hundred micrometre wide low dose regions (valleys) and was shown to spare tissue much more effectively than conventional radiation therapy at similar tumour control rates. While preclinical research focused primarily on tumours of the central nervous system, recently also lung tumours have been suggested as a potential target for MRT. This study investigates the effect of the lung microstructure, comprising air cavities of a few hundred mi- crometre diameter, on the microbeam dose distribution in lung. In Monte Carlo simulations different models of heterogeneous lung tissue are compared with pure water and homogeneous air -water mixtures. Experimentally, microbeam dose distributions in porous foam material with cavity sizes similar to the size of lung alveoli were measured with film dosimetry at the European Synchrotron Radiation Facility (ESRF) in Grenoble, France. Simulations and experiments show that the microstructure of the lung has a huge impact on the local doses in the microbeam fields. Locally, material inhomogeneities may change the dose by a factor of 1.7, and also average peak and valley doses substantially differ from those in homogeneous material. Our results imply that accurate dose prediction for MRT in lung requires adequate models of the lung mi- crostructure. Even if only average peak and valley doses are of interest, the assumption of a simple homogeneous air -water mixture is not sufficient. Since anatomic information on a micrometre scale are unavailable for clinical treatment planning, alternative methods and models have to be developed.