It is a very long time ago, now, since Sonia Marks and I worked together in the Department of French Studies at Sydney University. We were friends as well as colleagues, and we shared a great deal more than mere co-occurrence in the syntax of a university department would ordinarily predict. Our talk together was often about our work, but I suppose we thought then-in the ways determined by the structure of our profession-that the fields in which we. had our special interests-she, in foreign language teaching methodology, I, in semiotics-were quite separate. Yet we also knew that that was not necessarily so, and it was from Sonia that I first learnt that the pedagogy of a foreign language raised questions about teaching and about language that were inevitably obscured by that separation. There was a particular series of discussions, culminating when I had already left Sydney, in which our interests converged more concretely than hitherto. Sonia had developed a research interest in the teaching of reading comprehension. In a paper she read to a conference which I attended, she contested the premisses upon which literary semiotics had theorised the topic of reading. Comprehension of the written word, she argued, had very little to do with the structure of the text, and a great deal to do with the predictions a student brought to the reading task. The question, then, was how to tap into those predictions, to theorise where they came from and how they could be used in the business of reading a text in a foreign language. In a situation in which we standardly think of our students as having no predictions, that is, in a language of which they have little knowledge, in a situation, moreover, in which predictions from the native language environment can so frequently be counterproductive, what are the useful predictions, and how can we help our students mobilise them? It is in the area that these questions opened up for me that I have chosen my topic for tonight, in the