Another good example I heard is with colors. Different languages have different words for colors, blue being a specifically unmentioned one in Greek. So if you're calling the sea "wine-red" and the sky white, it just makes sense that you would compare other similar colors with the limited set of categories you have. I mean it seems easy to simulate. You set up a classifier to classify colors but leave out blue. Sure, it is going to poorly predict blues and misclassify them as things like reds, whites, or greens, but isn't that what SW predicts?
