In my presentation I will discuss the application of optimal design methods to fMRI experiments. Functional magnetic resonance imaging (fMRI) is a non-invasive neuroimaging method to localize functional areas in the human brain, i.e. to determine which brain areas are involved in performance of a given task, such as recognizing faces, reading text, or distinguishing between two objects. So far, fMRI researchers have not fully benefited from the methods and procedures available from statistical theory of optimal design. A general linear model with autoregressive correlated error can be used to describe the fMRI signal over time in one voxel (small cube of brain tissue) of one subject in an experiment. Different design factors were considered, thereby modeling a variety of what is known in fMRI as blocked experiments. Several optimality criteria were applied, reflecting different research questions. Optimal designs were dependent on the parameters of the error structure and therefore the maximin approach was considered to handle local optimality. Our results lead to some general recommendations for optimizing blocked fMRI experiments and other recommendations which are depending on the effect of interest.