In each trial, brain electrical activity at multiple points of a target person is measured. An acquirer of a determination device acquires response matrices for n trials under a first condition and response matrices form trials under a second condition. An analyzer performs canonical correlation analysis on the acquired response matrices to obtain first canonical variable time series. A distance calculator calculates a distance between the trials from the obtained first canonical variable time series to obtain a distance matrix. A determiner obtains a possibility that the n trials and the m trials are classified into two different clusters from the distance matrix and determines whether the first condition and the second condition are substantially different. It is possible to provide to a single target person a first content in n trials and a second content in m trials so as to determine a difference in interest of the single target person. It is possible to provide the same content to a first subject who is the target person in n trials and to a second subject who is the target person in m trials so as to determine whether the two are different or the same.