One process is not enough! A speed-accuracy tradeoff study of recognition memory