Predicting the performance of antenna arrays designed for MIMO communications is difficult. To determine the benefits that additional antenna elements may provide to a wireless system, the statistical correlation between the signals received from the antennas needs to be evaluated. In 2003 Blanch, proposed a simple method to estimate the correlation coefficient from S-parameters and in 2005 Hallbjorner proposed a modified version including the effect of antenna efficiency. We investigate the validity of these methods. Three different antenna arrays, with four elements each, and with varying efficiency and mutual coupling were designed, manufactured and measured. The method proposed by Blanch was found to be inaccurate whereas the Hallbjorner method produced better results but required determination of antenna efficiencies. S-parameter measurements alone do not provide valid estimates of the MIMO performance of antenna arrays. In order to evaluate MIMO performance, the antenna efficiency is paramount and needs to be measured accurately.