Judgements of typicality and judgments of category membership both derive from the same source: resemblance to an exemplar or to a prototype. Typicality and category membership should go hand in hand. It turns out, however, that we can also find situations in which there is no relation between typicality and category membership, and this is surely inconsisitent with our claims so far. Armstrong and Gleitman asked their research participants to do several of the concept tasks we have been discussing. The twist is that the researchers used categorie for which there clarly is a definition - for exmple, the category "odd number". They told them: "We all konw that some numbers are odder than others. What I want you to do is to rate each of the numbers on the list." Participants felt that this was a rather sill task, but nevertheless they were able to render these judgements. The obvious response to this experiment is that particpants knew the definiction of "odd number" but ware playing along in respons to the experimenters' peculair request. PArticipants's judgments about typicality were in this case disconnected from their judgments about category membership: they knew that all the test numbers were in the category "odd number", but they still judged some to be more "typical " than others. There is not always a link between typicality and membership.