Drawing on Sperber and Wilson's (1986) notion of representation by resemblance, this paper argues against Banfield (1982), that echo questions can be used to question more than merely a speaker's words, and that echo questions differ from non-echo questions simply by the fact that they are echoic.