In this paper, a segment of the electroencephalogram (EEG) signal of a subject is viewed as outcomes of a single and a joint two random variables. The Shannon entropy (self information) contained in the single random variable and the mutual information associated with the joint two random variables can be used for the diagnosis of Alzheimer’s disease. It is shown that the self information may provide diagnostic error while the mutual information provides accurate and significant diagnosis. This can be justified by the following fact. The mutual information between a current sample and a delayed one of the EEG signal is a quantitative measure for the information of the current sample contained in the past one. Thus, this mutual information is related to the subject memory, which experiences a problem in Alzheimer’s disease.