Yeah, the sample was quite clearly handpicked for increased impact (and faster habituation, probably). But the argument is (I believe) about the fact that you can easily handpick such samples from perfectly normal spoken phrases.