Direct experience by a skeptical observer in a controlled environment.
Preferably backed up by a solid, provable null hypothesis and some statistics.