Background: The human lung is complex system, comprising branching networks of airways and blood vessels embedded in a highly deformable tissue. The contributors to normal pulmonary function and disease progression are difficult to tease apart, and vary between individuals. Personalised computational models of lung structure and function have contributed to improved understanding of the contributors to pulmonary disease and are becoming useful tools to stratify patient severity [1]. However, personalised models are time-consuming to develop and so predicting how population variability contributes to pathological response is not possible, except in very small samples. Methods: Here we propose a statistical model of the lung which can be used to predict function in both the population average, and for lungs that represent the normal range of variability from this mean. This model is based on a novel framework called SFeaL (Statistical Finite element analysis of Lung) and an Active Shape Model (ASM) concept. The model is parameterised using a training set of 15 healthy non-smoking subjects aged 20-30 years old from a standardised Human Lung Atlas database. Results: The model is able to capture 90% of inter-subject variability in 7 principal components, which represent 3D shape changes. When lung volume is accounted for, the largest variability in lung shape is diaphragmatic, and there are not significant differences in male and female average lung shape (a maximum of ~30 mm surface distance from the average lung). To demonstrate a use of the SFeaL model to predict lung function we simulate tissue deformation under gravity and elastic recoil pressure in individuals, and in the statistically averaged lung. On an individual basis, model predictions of total lung density lie within 3% of CT measured values, and gradients in tissue density in the gravitational direction are 0.010-0.023 g/cm4 compared with measured values of 0.009-0.014 g/cm4. With the statistically averaged model we are able to predict the normal variability in distribution of lung tissue and elastic recoil pressures under a gravitational load. Implications: Prediction of a normal range of lung tissue deformation provides the first key steps to assessment of whether an individual’s lung tissue, which can be observed in computed tomography (CT) imaging, lies within a normal range. This improves upon current methodologies which do not capture individual variability. Local elastic recoil directly influences the patency of pulmonary airways and blood vessels and so the gas exchange ability of the lung so is critical to gas exchange function. Long term we aim to provide statistically averaged representations of pulmonary airways and blood vessels, so predictions of population response to pulmonary disease can be conducted in conditions representative of normal (or pathological) cohorts quickly and accurately.