In this work, spatial diversity techniques in the area of multiple-input multiple-output (MIMO) diffusion-based molecular communications (DBMC) are investigated. For transmitter-side spatial coding, Alamouti-type coding and repetition MIMO coding are proposed and analyzed. At the receiver-side, selection diversity, equal-gain combining, and maximum-ratio combining are studied as combining strategies. Throughout the numerical analysis, a symmetrical $2\times 2$ MIMO-DBMC system is assumed. Furthermore, a trained artificial neural network is utilized to acquire the channel impulse responses. The numerical analysis demonstrates that it is possible to achieve a diversity gain in molecular communications. In addition, it is shown that for MIMO-DBMC systems repetition MIMO coding is superior to Alamouti-type coding.