ABSTRACT
Sigmund Freud was notorious for his peculiar approach to science: He could explain everything, but he could predict nothing. I think just such an approach characterized item bias research 15 to 20 years ago, in its infancy. We could invent elaborate and persuasive reasons for why a particular item was biased, but could not apply this reasoning to new items and predict whether or not they would be biased. At that time we found ourselves in just the opposite position of the Supreme Court justice who, in a celebrated court decision, wrote “I cannot define pornography, but I know it when I see it.” We could define precisely (that is, mathematically) an item that was performing differentially for two specified groups of examinees, but, unlike the Supreme Court justice, we could not recognize it when we saw it.
