The behaviour of Dielectric Elastomer Actuators (DEA) can be predicted using hyperelastic models that are based on strain energy density functions. The parameters used in the hyperelastic models are generally obtained via a uni-axial pull test. However, DEAs are most commonly used in an biaxially stretched configuration. This is an appropriate assumption if the modelled parameters translate accurately to different stretch configurations. We have conducted stress-stretch experiments on silicone membranes in two different configurations; uni-axial and pure shear stretch. Fitting common hyperelastic models, such as Gent, to the experimental data shows that the modelling parameters depend on the stretch configuration. In addition, we show that the Mullins effect, where the stress-stretch response is dependent on the maximum stretch previously experienced by the sample, is predominant in the silicone membranes. This means that the model parameters depend on the loading configuration and the stretch history of the sample making it difficult to predict the behaviour of highlyprestretched DEAs. One way to tackle this issue is to carry out testing as close to the original configuration as possible which is difficult in the case of highly prestretched DEAs. We have created a model that takes into account both the loading configuration and the Mullins effect and used this to optimize the prestretch and stretch of the cell stretching device.