Pressure swing adsorption is a method for separating gases by selective adsorption. It is being used increasingly in industry where for some applications, including air separation, the number of plants have increased at a near exponential rate in recent years. However, despite the many hundred different plant configurations and cycles, there is still a lack of understanding of the basic process steps and plenty of scope for achieving even better plant performances. This study examines experimentally and theoretically a two bed process for separating oxygen from air using a zeolite adsorbent. A plant was designed and built to incorporate the novel feature of recycling waste gas from the purge and depressurization steps back into the feed line and also to investigate conventional purge, backfill and combined cycles. Theoretical modelling had predicted that large amounts of waste gas could be recycled without loss in product oxygen concentration, but causing an improvement in oxygen yield. This was verified experimentally, demontrating potential energy savings. For each cycle investigated the product oxygen concentration and yield were optimized over a range of product amounts per cycle while continually monitoring the main process variables. Other experimental work included studies of the bed temperature and pressure profiles, the waste gas oxygen concentrations, unsymmetrical operation and supplying purge from different sources. The theoretical aims were to further develop the instantaneous local equilibrium (ILE) model used by Kirkby (1984). The model was made more efficient and developed to include novel options for waste recycling and the previously neglected, but common, design feature of supplying purge directly from one bed to another. The model's qualitative agreement with experiments was verified over a wider range of cycles and the quantitative agreement was improved for some cycles.