This paper is about 'conversations' and 'texts'. I am going to define these two terms in a fairly pragmatic fashion. By 'conversations' I mean those things studied by conversation analysts. I am tempted to define 'texts' as those things not studied by conversation analysts. But that seems a little too inclusive. So, for my present purposes, a text is either a written statement (e.g. a letter) or an extended spoken statement (e.g. a lecture), many of which take the form of 'readings' of a prepared written statement. In part, then, the underlying inter? est is in the relationship between spoken and written discourse. One way of approaching the analysis of written texts is to examine how far they rely upon interpretative forms which have been shown to be characteristic of spoken discourse. This approach to textual analysis has recently been recom? mended by Atkinson, who proposes 'that an adequate understanding of how texts are produced and responded to may remain elusive so long as the issue is pursued without making close comparative reference to how talk works' (1983: 230). My aim is to illustrate the fruitfulness of this approach (see also, Mulkay, 1985b). First of all, I will show that Pomerantz's (1984a) findings on agreement and disagreement in conversations are equally applicable to a set of letters in which a group of scientists carry out a technical debate. Secondly, and perhaps more unexpectedly, I will show that Pomerantz's (1978) findings on responses to compliments help to reveal the basic structure of Nobel ceremo? nies.