The "conjunction fallacy" revisited : How intelligent inferences look like reasoning errors