The performance of wireless communication systems can be significantly improved using the modulation diversity technique, which is based on the combination of a suitable choice of the reference angle of a signal constellation with independent interleaving of the symbol components. This technique has been evaluated considering different fading channel models, such as Rayleigh, Rice and Nakagami-m. However, in some specific scenarios, the tails of those fading distributions do not properly fit the experimental measured data, which demands the use of more general channel distributions. This article presents a performance evaluation of the modulation diversity technique for κ-μ fading channels. New expressions for the PEP (Pairwise Error Probability) are obtained using numerical integration, series representation and upper/lower bounds. The evaluation, based on Monte Carlo simulation, demonstrates that the performance gain of the modulation diversity increases as the fading becomes more severe. Communications channels exhibit some degree of time correlation, which cannot be perfectly estimated, affecting the performance of the modulation diversity system. Thus, a performance evaluation of the system, concerning the presence of temporal correlation and estimation errors in the channel is also presented in the article.