(1) The technique of cloze passages is well known. Passages are specially prepared by deleting words; students have to make predictions from context and complete the gaps. Such exercises are linguistically principled, but nevertheless involve artificial preparation of texts. A real alternative which I used was to take a newspaper article in East African English which contained a large number of Swahili loan words, incomprehensible out of context to an English speaker. These loan words provided real lexical gaps for students to translate into English.
(2) A common situation in which hearers have to predict large parts of a conversation occurs when they hear one end of a telephone call. It is usually possible to predict much of what is said at the other end of the line. It is easy to tape record a telephone call, and to delete one speaker's contributions from the transcript to form a discourse cloze passage.
(3) A short story can be divided into sections and fed to students one section at a time. Their task is to predict what will happen next, and to write the continuation of the story. Again, this involves some manipulation of a text, but forces students to make explicit their expectations in a way which is essentially similar to that involved in an intelligent first reading of a literary text. Any such exercises can provide material for subsequent more formal analysis of the students' own predictions. This will inevitably involve comparison between different students' predictions, and between these predictions and the original. This will inevitably lead also to an analysis of the grammatical and lexical cohesion in the passages, of semantic relations such as paraphrase and entailment, as well as of the macrostructure of narratives and other discourse types. (In Stubbs, 1983a, I discuss in more detail some aspects of the semantic organisation of a literary text and give other examples of such classroom activities, suitable for mother tongue teaching in secondary schools.)