Researchers have relied on a very narrow normative view, and have ignored conceptual distinctions—e.g. single case versus relative frequency—fundamental to probability theory. By recognizing and using these distinctions, however, we can make apparently stable "errors" disappear, reappear, or even ...