Abstract: A memory unit stores, as decision-criterion data for a brain wave to be examined, a mean vector A, a variance vector C and at least one critical value F.sub..alpha. at an appropriate significant level .alpha. obtained from F-distribution, the mean vector A being the mean value of autoregressive coefficient vectors calculated by a data operation and processor unit from respective brain waves of a standard brain wave group consisting of K number of standard brain waves, the variance vector being calculated from the mean vector and the autoregressive coefficients. An L-dimensional row vector X of the brain wave to be examined is calculated by the data operation and processor unit from the autoregressive coefficients of the brain wave to be examined. The distance F.sub.X between the row vector X and the mean vector A is also calculated. Further, the value of the distance F.sub.X is compared with the significant level F.sub..alpha.