I will present an evolutionary model for the origin of Andromeda II, a dSph satellite of M31, involving a merger between two disky dwarf galaxies than explains the origin of prolate rotation recently detected in the kinematic data for And II. The simulation traces the evolution of two dwarfs, whose structural parameters differ only in their disk scale lengths, placed on a radial orbit towards each other with their angular momenta inclined by 90 deg. After 5 Gyr the merger remnant forms a stable triaxial galaxy with rotation only around the longest axis. This prolate rotation is naturally explained as due to the symmetry of the initial configuration which leads to the conservation of angular momentum components along the direction of the merger. The stars originating from the two dwarfs show significantly different surface density profiles while having very similar kinematics in agreement with the properties of separate stellar populations in And II. I will also discuss an alternative scenario for the formation of And II, via tidal stirring of a disky dwarf galaxy. While intrinsic rotation occurs naturally in this model as a remnant of the initial rotation of the disk, it is mostly around the shortest axis of the stellar component. The rotation around the longest axis is induced only occasionally and remains much smaller that the system's velocity dispersion. I conclude that although tidal origin of the velocity distribution in And II cannot be excluded, it is much more naturally explained within the scenario involving a past merger event. Thus, in principle, the presence of prolate rotation in dSph galaxies of the Local Group and beyond may be used as an indicator of major mergers in their history or even as a way to distinguish between the two scenarios of their formation.